ࡱ> B` RTbjbjipyy $$LSLSLSLS8V$VZZZZ7L80Ghz|)||| ZZg3u&&&| ZZ&|&&r|^6ZV 'KLS,Z0z):)6)6$3;:&uq 6||||$$$d%:$$$:$$$  An Integrated Resource Management and Scheduling System for Grid Data Streaming Applications Wen Zhang1, Junwei Cao2,3*, Yisheng Zhong1,3, Lianchen Liu1,3, and Cheng Wu1,3 1Department of Automation, Tsinghua University, Beijing 100084, China 2Research Institute of Information Technology, Tsinghua University, Beijing 100084, China 3Tsinghua National Laboratory for Information Science and Technology, Beijing 100084, China *Corresponding email: jcao@tsinghua.edu.cn Abstract Grid data streaming applications are novel from others in that they require real-time data supply while the processing is going on, which necessitates harmonious collaborations among processors, bandwidth and storage. Traditional scheduling approaches may not be sufficient for such applications, for they usually focus on only one aspect of resources, mainly computational resources. A resource management and scheduling system for such applications is developed in this paper, which is responsible for enabling their running based on Globus toolkit. An integrated scheme is proposed, including admission control, application selecting, processor assigning, allocation of bandwidth and storage, with corresponding algorithms elaborated. Evaluation results show excellent performance and scalability of this system. 1. Introduction Streaming applications are gaining their popularity recently, and in most cases data are pushed to the computational resources for distributed processing with real-time constraint, so the processing rate must match the data arrival rate. Nowadays, new kinds of streaming applications are emerging with different requirements and characteristics. For example, LIGO (Laser Interferometer Gravitational-wave Observatory)  REF _Ref182636441 \r \h \* MERGEFORMAT [1] is generating 1TB scientific data per day and trying to benefit from processing capabilities provided by the Open Science Grid (OSG)  REF _Ref182728811 \r \h  \* MERGEFORMAT [2]. Since most OSG sites are CPU-rich but storage-limited with no LIGO data available, data streaming supports are required in order to utilize OSG CPU resources. In such a data streaming scenario, data should be pulled rather than pushed to the computational system in the form of streams of tuples, and processing is continuously executed over these streams as if data were always available from local storage. Whats more, data arrival rates must be controlled to match the processing speeds to avoid waste of computational capacity or data overflow. Meanwhile, processed data have to be cleaned up to save space for the subsequently coming data. Such applications are novel in that (1) they are continuous and long running in nature; (2) they require efficient transmission of data from/to distributed sources/sinks in an end-user-pulling way; (3) it is often not feasible to store all the data in entirety for later processing because of limited storage and high volumes of data to be processed; (4) they need to make efficient use of high performance computing (HPC) resources to carry out compute-intensive tasks in a timely manner. Grid computing  REF _Ref196322353 \r \h  \* MERGEFORMAT [3] paves a new way for such kinds of applications, giving birth to the so-called Grid Data Streaming applications. Such applications require the combination of bandwidth sufficiency, adequate storage and processors to guarantee smooth and high-efficiency processing, making them different from other batch-oriented ones. Most scheduling infrastructures available in the filed of grid, such as Legion  REF _Ref181867177 \r \h \* MERGEFORMAT [4], Nimrod/G  REF _Ref181867214 \r \h \* MERGEFORMAT [5] and Condor  REF _Ref181867223 \r \h \* MERGEFORMAT [6], are largely geared to support batch-oriented applications rather than the streaming ones. Some schedulers are developed to support data streaming applications, such as E-Condor, GATES  REF _Ref196407260 \r \h  \* MERGEFORMAT [7], and Streamline  REF _Ref182131562 \r \h \* MERGEFORMAT [8], but they just concern on computational resource allocation, paying little attention to storage and network bandwidth. Pegasus  REF _Ref190156686 \r \h  \* MERGEFORMAT [9] has the most similar motivation with the work described in this paper, but it handles data transfers, job processing and data cleanups in a workflow manner. EnLIGHTened computing  REF _Ref202169507 \r \h [10] and G-lambda  REF _Ref202169520 \r \h [11] project, which provide co-allocated computing and network resources with advance reservation, but they dont concern with specific requirements of Grid data streaming applications. In this paper, an integrated resource management and scheduling system is developed from viewpoint of the resources, including processor, storage and bandwidth, to make efficient use of them and accommodate as many streaming applications as possible to achieve high throughput. This resource management and scheduling system tries to allocate processors, storage and bandwidth synchronously to guarantee such applications to execute smoothly with high efficiency. Based on Globus toolkit  REF _Ref196707636 \r \h [12], this system is able to discover and manage resources geographically distributed and belonging to different management domains in a transparent and secure way. Some key algorithms are proposed, including admission control, application selecting, processor assigning, bandwidth allocation and storage allocation. Evaluation results show excellent performance and scalability of this system. The rest of this paper is organized as following: Section 2 describes the overall architecture and mechanism of this resource management and scheduling system, whose core algorithms are elaborated in the next section; some evaluation results are included in Section 4, and the following section concludes this paper. 2. System Architecture The architecture of our resource management and scheduling system is shown in Figure 1 and its key components include but are not limited to: Client Tool This tool is an interface for users to submit their applications with their requirements in XML format, including the executable, processor types and amount , minimum bandwidth and storage, data source, just like but more than what Condor submission does. It is also capable of monitoring the status of submitted applications and that of the resources in the whole grid. Nowadays, it is carried out in command lines, and in the future a graphical user interface (GUI) will be available. Management Engine The management engine accepts users submissions of applications and put them into the queue, which will be accessed by the scheduler. Its main function is to provide grid supports for streaming applications, such as security, resource discovery and management. The components of Globus toolkit used here include GRAM (Globus Resource Allocation Manager), MDS (Meta-computing Directory Service), GSI (Globus Security Infrastructure), GASS (Global Access to Secondary Storage), NWS (Network Weather Service), GRIS (Grid Resource Information Service), GIIS (Grid Index Information Service) and so on. Scheduler This is the core component in the whole architecture and its key algorithms will be discussed in details in Section 3. It is responsible to carry out admission control, application selecting, processor assignment, and bandwidth and storage allocation. Its instruction will be executed by the dispatcher. Dispatcher The dispatcher is in charge of sending executables with their description files to appropriate processors and invoking a remote component, i.e., application wrapper. This component will interact with the services provided by grid middleware, such as GRAM.   Figure 1. System architecture Application Wrapper This component will parse the description file according to the XML schemas, initialize execution of executables, and start data transmission to specified storage with allocated bandwidth. Also, it will send back the results through dispatcher. Another function is to monitor the usage of storage to determine data transmission status, see more details in subsection 3.5.  The overview of the running mechanism is illustrated in Figure 2. Besides allocation of computational resources as most traditional resource management and scheduling systems do, it also deals with allocation of bandwidth and storage to support real-time data supply, which is required by data streaming applications. Management and scheduling of processors, bandwidth and storage are carried out in an integrated way rather than independently. 3. Key Algorithms This section just elaborates on the key algorithms as the core of this resource management and scheduling system, i.e., the scheduler. Note that although processor assignment, allocation schemes for storage and bandwidth are described and evaluated separately, they are carried out synchronously as integration. 3.1. Admission control When a new job is submitted, admission controller would decide to run it instantly or just keep it in the waiting queue. This decision is made according to the usage status of resources and the requirements of the jobs. Each job can allege its minimum requirement of resources, e.g., it needs some processors, bandwidth and storage. An XML schema is developed for the applications to express their requirements in the manner similar to Resource Description Language (RSL). For each application s, it can declare its minimum requirement of resources like  EMBED Equation.3  where ps stands for the number of processors it requires, so ps=1 for simple applications (i.e., standalone applications) and ps>1 for composed applications (such as a pipeline); bs and sts stand for the required minimum bandwidth and storage respectively. This information will be included in the submission file in XML format. Suppose the running applications in the computing pool form a set, denoted as SR, and the total amount of processors, bandwidth and storage be denoted as P, B and S respectively. Some applications have their special requirements upon processors, for example, applications compiled on X86_64 cannot run on I386 processors, so not every processor is suitable for each application. Suppose those processors eligible for application s form a set, called Ps, and the number of free (not occupied or reserved) processors in it when s comes is denoted as | Ps |. In any one of the following three cases, a new application, sn, would just be kept in the waiting queue for there are no enough resources (suitable and enough processors, enough bandwidth and enough storage respectively) for it.  EMBED Equation.3   EMBED Equation.3   EMBED Equation.3  If an applications minimum requirement can be satisfied according to the current status of computing pool, it will called a potential eligible application (PEA), which means that it may be permitted into computing pool. 3.2. Application selection PEAs form a queue, within which maybe several ones satisfy the admission control policy. A selecting policy must be applied to choose some from the queue and assign appropriate resources for them. Those selected ones will be called eligible applications (EAs). PEAs have different weights, and the higher weights mean that they can be selected with bigger priorities. PEAs will be classified into several groups according to their weights, and in each group, the selecting principle is first-come-first-serve (FCFS). The selecting will be heuristic and iterative: the first coming PEA with the highest weight will be selected, and then the next one till the last one in its group (if there are) will be tested in their arriving order; then it is turn for the group with second highest weight, till all the groups are tested. Notice that the PEAs with higher weight will not be selected prior to those with lower weight necessarily, for whenever a PEA is accepted, the resource status will change and some PEAs will become ineligible. To some extent, this algorithm resembles first-fit (FIFT) with backfilling mechanism. What is more, to avoid that some PEAs starve for a long time, some reservation policy will be adopted. Some resources will be labeled as reserved when they are executing other applications, and as soon as they are free, they will be assigned to the applications which reserve them. Weights of each application will increase as time goes by, to avoid such cases where applications with lower weights will be idle forever. The weights will be a function of time, with the originally set value as their initializations  EMBED Equation.3   EMBED Equation.3  where wi0 is the initial weight of application i and f(t) is an non-decreasing function about time t. A function in case is  EMBED Equation.3   EMBED Equation.3  where di is the increase coefficient and di>0; Ti is the increase period and function floor returns the nearest integer towards minus infinity for t divided with Ti . Then wi will increase by di once a period Ti. Assigning appropriate values for di and Ti, after some time of waiting, the applications with lower weights initially will be endowed a high enough weight to be selected from PEA queues. Combination of reservation policy and increasing weight over time will guarantee each application will be accepted by the computing pool in appropriate time. In one word, the selecting algorithm tries to make full use of resources and keep fairness among applications. 3.3. Processor assignment As soon as EAs are selected, it is time to assign resources for them. Applications may have their own styles, i.e., they may be executed more smoothly on some processors than on others. So it is necessary to assign appropriate processors for applications, and purely random assignment will not work. On the other hand, the processors can be classified into several groups according to their characteristics, their architecture for instance. One application will achieve similar performance on the processors of a group, so it is not necessary to launch it on each processor for trial, but a processor can act as the representative of its peers in the same group. Matchmaking will be carried out to find candidate processors for applications, and applications will be assigned to processors in the matched group to run a short period of time to get its performance information. The applications with higher weights will have higher priorities to find their matched processors, and the processors producing the highest processing efficiency will be selected. 3.4. Storage allocation When new EAs arrive, the scheduler is responsible for allocating bandwidth and storage for them, together with the existing applications in the computing pool. The overall principle for storage allocation is to make full usage of storage to increase robustness while getting ready for new coming applications. If there are only a few applications running in the pool, the storage allocated for each application can be set to a high value. While the applications increase, the allocated storage for each application may be decreased. There must be some margin of storage for potentially coming applications. An iterative allocation algorithm of storage is proposed as following:  = 1 \* GB3 `$ initialization: suppose there are n applications in the pool, to generate n random numbers, ri" (0, 1), i=1,2,& ,n. Calculate each quota, qi as following  EMBED Equation.3   = 2 \* GB3 a$ If qi*TSe"sti, reserve these numbers for initially allocated storage for application i; else, repeat step  = 1 \* GB3 `$ until all of these inequations hold true, where sti is the minimal required storage of application i as mentioned in subsection 3.1 and TS is the total storage available for applications.  = 3 \* GB3 b$ dynamic adjustment: periodically, monitoring usage status of each allocated storage, and those with high occupation percentages will be increased while others will be decreased;  = 4 \* GB3 c$ when a new EA is coming, decrease the amount of the biggest partition of storage;  = 5 \* GB3 d$ when an application is finished, its storage will be divided and allocated to the minimal partitions;  = 6 \* GB3 e$ repeat  = 3 \* GB3 b$,  = 4 \* GB3 c$ and  = 5 \* GB3 d$ until all the applications are completed. 3.5. Bandwidth allocation Bandwidth allocation plays an important role in the whole resource allocating scheme, for appropriate bandwidth is indispensable to guarantee data supply for applications to make them run constantly. To make a flexible allocation scheme, so-called utility functions are introduced and genetic algorithm  REF _Ref182923659 \r \h  \* MERGEFORMAT [13] is adopted to maximize their sum. Different from traditional bandwidth allocation, our scheme is storage aware, i.e., data transmission may be intermittent rather than continuous to avoid data overflow, for allocated storage for each application is limited. When the storage is full of data, transmission will be halted for a while until some data have been processed and cleaned up so that some storage is released for more data. At any moment, the amount of data in storage for each application is affected by data supply and clean-up at the same time, where the former tends to increase the amount while the latter will decrease it. The computing pool is connected to Internet through which the data are streamed to the applications being executed on the processors, and the total input bandwidth, denoted as I, is limited, which is shared by the data streams. The data streams, called sessions, denoted as s, form a set S. Each session will be assigned a bandwidth xs, where xs "Xs, Xs =[bs,Bs] and bs >0, Bs <". bs stands for the least bandwidth required for session s, while Bs is the highest bandwidth available for s from the corresponding data source. Session s will have a utility Us(xs) when its data is supplied at a rate xs, where Us(xs) is called utility function and assumed to be concave, continuous, bounded and increasing in the interval [bs,Bs]. Note that it is not necessary that all the sessions adopt identical utility functions. We try to maximize the sum of the utilities of all the sessions, maintaining fairness among them. The problem can be described as follows.  EMBED Equation.3   EMBED Equation.3  EMBED Equation.3  EMBED Equation.3   EMBED Equation.3  EMBED Equation.3  EMBED Equation.3  EMBED Equation.3  Due to the usage status of storage, there are two possible states for each s at any time, i.e., active and inactive, which indicate a data transmission is on or off. All the active sessions form a set, called SA, and it is obvious that this set is varying because the states of sessions are changing. We just allocate bandwidth for active transmissions, so the constraint (2) may be rewritten as  EMBED Equation.3  EMBED Equation.3 An iterative optimization algorithm is proposed in  REF _Ref192854071 \r \h [14] and its convergence is analyzed, but it is required to be aware of the congestion on the path, which is hard to be satisfied in the wide Internet. According to our situation, we make some modification upon it as following. While s "SA  EMBED Equation.3  EMBED Equation.3 Otherwise  EMBED Equation.3  EMBED Equation.3 Here, xs(k)is the bandwidth for session s"S at the kth step. {k} and {k}are two positive sequences. For the sake of convenience, k and k are usually substituted as a fixed value, denoted as  and  respectively. [] Xs denotes a projection on the set Xs and can be calculated as  EMBED Equation.3   EMBED Equation.3  is the sub gradient of  EMBED Equation.3  EMBED Equation.3 and  EMBED Equation.3  If we define  EMBED Equation.3  formula (7) can be modified as  EMBED Equation.3  And  is the so-called safety coefficient to avoid bandwidth excess, where " (0, 1), i.e., there is some margin from the full use of total bandwidth for flexibility and robustness. Some heuristic algorithms, such as genetic algorithm will be applied to find optimal values for them. A popular utility function can be expressed as  EMBED Equation.3  where s may stand for the sessions coefficient. Essentially, bandwidth allocation here is a kind of additive increase multiplicative decrease (AIMD) algorithm which is usually used in TCP congestion avoidance, but it has some pleasant new characters. It is storage-aware, for the transmission will be stopped if allocated storage is nearly full, so data overflow will be avoided while enough data are supplied; it is processing-aware, for the processing capacity of processors will be inflected in the varying occupation of storage, although it may not aware of the precise values, and such bandwidth allocation is on-demand; of course, it is congestion aware. As mentioned above, the scheduler should make allocation schemes in an evolving way to keep pace with latest situation, i.e., when an application is submitted or finished, or resources increase or decrease dramatically, it will be invoked to make new schemes to correspond to latest conditions. 4. Evaluation A campus computational grid is being established in Tsinghua University (Beijing, China) which holds a large amount of supercomputers, personal computers and other special instruments. Globus toolkit 4.0.1 is being deployed to provide common grid services and a simple Certificate Authority has been established to sign certificates for hosts and users which will be used to establish a secure and transparent environment for data streaming applications. This campus grid is connected to Internet with limited bandwidth, and network file system (NFS) is established to which all the data streams are directed. Applications are submitted at moments complying with negative exponential distribution law, and their requirements of resources are also explicitly expressed. Experiments are carried out for 10,000 units of time and some results are obtained. Resource scheduling is carried out once per 200 units of time to correspond to updated situations. 4.1. Admission control As the resources in the computing grid are limited and each streaming application holds its own requirement, it can be inferred that too many applications accepted by the computing grid will lead to low processing efficiency if no admission control is carried out, and some experimental results verify it, as shown in Figure 3. Higher throughput is achieved in the scenario with admission control than that in the scenario without admission control. Note that in the latter scenario, one processor may have to deal with more than one application at the same time, which offends the assumption made before that one application will occupy a processor exclusively. Inadequate data supply for each application and discount of computational capacity due to competition among applications on one single processor lead to a lower data processing efficiency as a whole.  Figure 3. Throughput with/without admission control Whats more, admission control can make applications finished sooner than without, as demonstrated in (a) and (b) in Figure 4, where the red bars with character P stand for the pending status and pink bars with character R mean for the running status for each application. The numbers of completed applications are 29 and 25 respectively. More importantly, most of the makespans without admission control are longer than their counterparts, which is adverse for the requirements of quality of service. 4.2. Bandwidth allocation Bandwidth is allocated to each running application to guarantee their data supply. Parameters in bandwidth allocation are obtained with genetic algorithm and are applied in each period. This bandwidth allocation is adaptive to the total available bandwidth and requirements of running applications.  (a) With admission control  (b) Without admission control Figure 4. Status of applications in iterative bandwidth allocation To justify our bandwidth allocation algorithm (named iterative allocation), we compare it with the even bandwidth allocation, where the bandwidth is allocated to the running applications equably as shown in Figure 5. In (a), the total available bandwidth is relatively small which equals with that in case of Figure 3, and 25 applications are finished; in (b), the available bandwidth is relative big, so each application can get enough data supply with the even allocation scheme, and the result resembles that in (a) of Figure 4.  (a) Low bandwidth  (b) High bandwidth Figure 5. Status of applications in even bandwidth allocation Figure 6. Throughput for bandwidth allocation schemes Throughputs in the cases of iterative allocation and even allocation are shown in Figure 6, where the numbers at the end of legends stand for the total available bandwidth. Then our allocation scheme is justified that it can achieve high throughput with relatively small available bandwidth for it is processing-aware while the even allocation scheme is not, so in that case some applications may starve while others may be allocated redundant bandwidth. 4.3. Storage usage Data supply in our scheme is storage-aware, i.e., data supply is controlled by the usage of allocated storage, rather than spontaneously. The principle here is just enough data is ok, not the more data the better. Sometimes the data transmission is intermittent, not always continuous. In this way high volume of data can be processed with just reasonable storage, as shown in (a) of Figure 7, where the used storage just varies in a limited scope. If data supply is continuous and available storage is big enough, the occupied storage will be of high volume, which can be observed in (b) of Figure 7. Actually, small storage can achieve high throughput in the streaming applications with well-made data supply and processing scheme, which is the prominent characteristic of such scenarios. Relative big storage is not necessary but rather desirable, for more data can be stored before processed to survive network collapse when no more data can be supplied. 4.4. Processor assignment Processors are assigned to applications, one for a single application exclusively. Here, the allocation resembles job shop problem scheduling, as shown in Figure 8, where the numbers in the horizontal bars stand for the corresponding applications executed on the processors in a certain group. Group 1 and 2 deal with more applications while they hold less processors than group 3, so the average loads of processors are heavier that those of group 3, as demonstrated in (a), (b) and (c) respectively. Because some applications may not be able to be executed on the processors in group 3, they cannot be transferred to the processors in group 1 and 2 to make a load balance.  (a) Storage-aware data supply  (b) Non-storage-aware data supply Figure 7. Storage usage 5. Conclusions Data streaming applications are of the novel types of grid scenarios for own their characteristics, such as requiring of real-time data supply and integrated resource allocation schemes. Different from existing resource management and scheduling schemes that just focus on computing resources, the system proposed in this paper takes computational resources, bandwidth and storage into account simultaneously and make integrated management and scheduling schemes, which are proved to be feasible with excellent performance. Up to now, requirements of quality of service (QoS) for applications have not been paid enough attention, and this desirable character will be the emphasis of further research. Scheduling for pipelined applications will be studied which is more complex with requirement of balance among stages and appropriate data supply. Ongoing work includes the consideration of data sharing scenarios among multiple data processing applications. Also some heuristic scheduling algorithm is under development for refined performance optimization.  (a) Group 1  (b) Group 2  (c) Group 3 Figure 8. Processor assignment for jobs Acknowledgement This work is supported by Ministry of Education of China under the higher education quality engineering project National Open Course Integrated Systems, and Ministry of Science and Technology of China under the national 863 high-tech R&D program (grants No. 2006AA10Z237, No. 2007AA01Z179 and No. 2008AA01Z118). References E. Deelman, C. Kesselman, G. Mehta, L. Meshkat, L. Pearlman, K. Blackburn, P. Ehrens, A. Lazzarini, R. Williams, and S. Koranda, GriPhyN and LIGO, Building a Virtual Data Grid for Gravitational Wave Scientists, Proc. 11th IEEE Int. Symp. on High Performance Distributed Computing, pp. 225-234, 2002. R. Pordes for the Open Science Grid Consortium, The Open Science Grid, Proc. Computing in High Energy and Nuclear Physics Conf., Interlaken, Switzerland, 2004. I. Foster and C. Kesselman, The Grid: Blueprint for a New Computing Infrastructure, Morgan Kaufmann, San Francisco, 1998. S. J. Chapin, D. Katramatos, J. Karpovich and A. S. Grimshaw, The Legion Resource Management System,  "#789P[\]^ghtxzM ʿtjbWMMht[ht[H*o(ht[PJnHo(tHht[h4o(hQhBh4H*o(h4PJnHo(tHhQhBht[H*o(ht[ht[o(#h|b+h]4CJ PJaJ nHo(tH hCJo(hPJnHo(tHhwPJnHtHhwPJnHo(tHhPJnHo(tHhx@PJnHo(tHhfPJnHo(tHh3.PJnHo(tHhfPJnHtH]^M   # gd8 ,WD`,gd8 ,WD`,gdy ,WD`,gd2Xj ,WD`,gdz gd gd"!gdxq!gdt[gdt[gdUO RTM N # * + Y a i v zl_QC6Ch7zh%PJnHtHh7zh%PJnHo(tHh7zhYPJnHo(tHh7zhYPJnHtHh7zh(6PJnHo(tHh7zhMPJnHo(tHh7zhPJnHo(tHh7zhxMzPJnHo(tHh7zhuPJnHo(tHh7zhu1lPJnHo(tHh7zhPJnHo(tHh7zhPJnHtHhRIhaJ hCJhht[ht[o(ht[ht[H*o(  " ' 5 > ^ ` a    ' ] ǼDz}rg\gN@h7zh>APJnHo(tHh7zh%PJnHo(tHh\PJnHo(tHhfPJnHo(tHh,PJnHo(tHh,PJnHtHh1PJnHo(tHh&PJnHtHh&PJnHo(tHh7zPJnHo(tHh7zPJnHtHh+PJnHo(tHh7zh7wPJnHo(tHh7zh1/[PJnHo(tHh7zh!"PJnHo(tHh7zh6dPJnHo(tH] d f p     ! T ɼt`\N@hzh9YPJnHo(tHhzh PJnHo(tHh&h?hlL6CJ PJaJ nHo(tHh PJnHo(tHh'APJnHo(tHh_9/PJnHo(tHh7zh7x PJnHo(tHh7zhz7PJnHo(tHh}PJnHo(tHh7zhz7PJnHtHh7zh;PJnHo(tHh7zhYPJnHo(tHh7zh>APJnHo(tHh7zh>APJnHtHT y  @zwxyzƸ}pp\}R}D}pphzhr5PJnHo(tHhrPJnHtH'jhh!PJUnHtHhh2XjPJnHtH!jhh2XjPJUnHtHhzhM>6PJnHo(tHhh2XjPJnHo(tHhzhIPJnHo(tHhzhyPJnHo(tHhzhyPJnHtHhzhCxPJnHo(tHh!h#G6PJnHo(tHhzh#GPJnHo(tHLRqz{}͵}o_O_oAohzhPJnHo(tHh!h==6PJnHo(tHh!hw6PJnHo(tHhzhwPJnHo(tHhzhTPJnHo(tHhzhm+NPJnHo(tHhzhT{PJnHo(tHhzh:DPJnHo(tHhh2XjPJnHo(tHhrPJnHtH!jhh2XjPJUnHtHhhzPJnHtH'j}hh!PJUnHtH %.AGHNT0BǺǬvh[hM?hzhbPJnHo(tHhzh2XjPJnHo(tHhzhXPJnHtHhzhXPJnHo(tHhzhqPJnHo(tHhzhPJnHo(tHhzhRPJnHtHhzhzOPJnHtHhzhzOPJnHo(tHhh2XjPJnHtHhh2XjPJnHo(tHhzhPJnHo(tHhzhwPJnHo(tHhzh!;PJnHo(tHBFTakl},4պ㺬ugYgKgYhzhk\PJnHo(tHhzhbPJnHo(tHhzhnZPJnHo(tHhzhe PJnHo(tHhzhlPJnHtHhzhlPJnHo(tHhzho!PJnHo(tHhzh#{PJnHo(tHhzhPJnHo(tHhzhPJnHtHhzh4PJnHo(tHhzhlPJnHo(tHhzh|O)PJnHo(tH()*+BCDTUXY ŷţҏsh[P[E[:h:PJnHo(tHh$VPJnHo(tHhP PJnHo(tHhh:PJnHtHhXPJnHo(tHhzh4PJnHo(tHhzh+ZPJnHo(tHhrPJnHtHhzPJnHtH'jhzh|>2PJUnHtHhzh>O@PJnHo(tHhzh>O@PJnHtH!jhzh>O@PJUnHtHhzhbPJnHo(tHhhbPJnHo(tH"#-1V]` !ƻܰ{nnZ{P{Eh#PJnHo(tHhrPJnHtH'jwhh!PJUnHtHhh2XjPJnHtH!jhh2XjPJUnHtHhPJnHo(tHhh2XjPJnHo(tHh-2PJnHo(tHhB*UPJnHo(tHh.PJnHo(tHh#|PJnHo(tHhw4PJnHo(tHhh:PJnHtHh:PJnHo(tHhV-GPJnHo(tH!"#$;LMNQR]^_`w8Bִ֖}sh]hPhhcEPJnHtHhgKPJnHo(tHhcEPJnHo(tHhcEPJnHtHhRPJnHo(tHhhPJnHo(tH'jqhh!PJUnHtHhrPJnHtH'jhh!PJUnHtHhh2XjPJnHo(tHhh2XjPJnHtH!jhh2XjPJUnHtHhPJnHo(tHBGHIJabcstwxz~űҝsҝhhZhhe hcEPJnHo(tHhMWPJnHo(tH'jkhhcEPJUnHtHhcEPJnHo(tHhgKPJnHo(tHhrPJnHtHhPJnHtH'jhhcEPJUnHtHhhcEPJnHtH!jhhcEPJUnHtHhh;PJnHo(tHhhcEPJnHo(tH/0189:;RSTdehi67ƸƮ{n`RH=HhS$PJnHo(tHhS$PJnHtHjhS$PJUnHtHhmGHhmGHPJnHo(tHhmGHhmGHPJnHtHhf4PJnHo(tHhrPJnHtHhPJnHtH'jhshQU2PJUnHtHhsPJnHtHjhsPJUnHtHhsPJnHo(tHhf4PJnHtHhMWPJnHo(tHhe hcEPJnHo(tHhPJnHo(tH789=>LMfghlm"#$2@cݼƚƏwj\QCh6{h3PJnHo(tHhmPJnHo(tHh6{hW7PJnHo(tHh6{hW7PJnHtHhmGHh2XjPJnHo(tHhmGHPJnHtHhmGHPJnHo(tHhmGHhmGHPJnHo(tH'jhS$h<PJUnHtHhS$PJnHtHhmGHhmGHPJnHtHhrPJnHtHjhS$PJUnHtH'jehS$h<PJUnHtHcdijw|дvk]]P]B]h6{hMPJnHo(tHh6{hi}PJnHtHh6{hi}PJnHo(tHhKPJnHo(tHh0LPJnHo(tHh6{hNL`PJnHtHh6{hNL`PJnHo(tHh6{hl,PJnHtHh]_PJnHo(tHh6{hl,PJnHo(tHh6{h1PJnHo(tHh6{h3PJnHo(tHhD=PJnHo(tHh6CPJnHo(tHh{PJnHo(tH789:gq    $%~t~i[QFQh>-PJnHo(tHh>-PJnHtHjh>-PJUnHtHhN@PJnHo(tHh{PJnHtHh{PJnHo(tHh;DPJnHo(tHh}| PJnHo(tHhPJnHo(tHhENPJnHo(tHh1IPJnHo(tHh1IPJnHtHh6{hi}PJnHo(tHh6{hl,PJnHo(tHhRhPJnHo(tHh6{hMPJnHo(tH%&'+,.EbcefȽxncXMB5h7zhBPJnHtHhCEPJnHo(tHh(PJnHo(tHhPJnHo(tHh%dPJnHo(tHh%dPJnHtHhKPJnHo(tHhPJnHo(tHhrvPJnHo(tHhUJPJnHo(tHhUJhUJPJnHo(tHh~PJnHo(tHh!/#PJnHo(tHh{PJnHo(tHhrPJnHtHjh>-PJUnHtH'j_h>-h#o1PJUnHtHfo 3Qnvõ~s~h]R]G]hTWPJnHo(tHhPJnHo(tHh!APJnHo(tHhPJnHo(tHhG,PJnHo(tHhPJnHo(tHhg!PJnHo(tHhWPJnHo(tHh$wQPJnHo(tHh PJnHo(tHh@=h2XjPJnHo(tHhrPJnHo(tHh@=h2XjPJnHtHh.jPJnHo(tHh7zhBPJnHo(tHhBPJnHo(tH  V\kuv̩{ppe{ZOh!PJnHo(tHhwBPJnHo(tHhgBPJnHo(tHh& PJnHo(tHh =PJnHtHh%'PJnHo(tHh =PJnHo(tHh =h =PJnHo(tHh =h =PJnHtHh&PJnHo(tHh}PJnHo(tHhhEh/`PJnHo(tHhPJnHo(tHh!APJnHo(tHhYEPJnHo(tH!!##,%7%7&9&:&;&<&=&>&?&@&A&B&C&E&F& ,WD`,gd? & Fgd?`gd  & Fgd ,WD`,gdC & Fgd8% ,WD`,gdg ? !!!!!!!!""#"""""}###### $$h$i$$$$$$$$$$$,%-%7%8%o%t%%%7&8&9&hQEhhPJnHo(tH%jh3!PJUmHnHo(tHuh1PJnHo(tHh?PJnHo(tHh?PJnHtHh PJnHo(tHh PJnHtHh<PJnHo(tHh!PJnHo(tHh!PJnHtH49&C&D&F&L&W&X&c&d&e&p&q&x&y&z&&&!'('6'7'n'u''}kaVaVKVaVAVhPJnHtHhZqPJnHo(tHhS8PJnHo(tHhS8PJnHtH"jh(UPJUmHnHtHuhwBPJnHo(tHh1PJnHo(tHh1PJnHtH#hJ[hk~CJPJaJnHo(tH#hJ[hCJPJaJnHo(tH#hJ[h4CJPJaJnHo(tHh%'CJPJaJnHtH%jh3cPJUmHnHo(tHuhPJnHo(tHF&d&x&'''''''''))++,B-Z-$a$gdq_' ,WD`,gd; ,WD`,gd` gdq ,WD`,gd  gdR`gdR` & Fgd8%$a$gd4'''''''((/(5(8(9(j(o((((((4)5)6)^)g)t)ti^TITIh4bRPJnHo(tHh4bRPJnHtHhPJnHo(tHhvPJnHo(tHh/MPJnHo(tHhdPJnHo(tHh OPJnHo(tHhPJnHo(tHh%'PJnHo(tHh# PJnHo(tHh0PJnHo(tHh0PJnHtH%jh{PJUmHnHo(tHuh{PJnHo(tHhStPJnHo(tHh};PJnHo(tHt)w)))))))))))))9*L*N*O*S*W*a*k*u*w*ɻxmcmcmUHUh$2hmPJnHtHh$2hmPJnHo(tHhzPJnHtHhzPJnHo(tHhsdPJnHo(tHhW@htPJnHo(tHhW@htPJnHtHhrhtPJnHo(tHhtPJnHo(tHhth/htPJnHo(tHh$2PJnHo(tHhk<'PJnHo(tHh4bRPJnHtHh4bRPJnHo(tHh:PJnHo(tHw*******+++4+O+T+++++-,1,<,@,J,K,d,f,g,j,,´vkkkaVLVhxPJnHtHhjPJnHo(tHhjPJnHtHhvBPJnHo(tHh8PJnHo(tHhh`PJnHo(tHhh`PJnHtHh heCJaJo(h husCJaJo(h$2hPJnHo(tHh$2h:PJnHo(tHhPJnHtHhzPJnHo(tHh$2hRPJnHo(tHh$2hRPJnHtH,,,,,,,--B-C-D-U-V-W-X-Y-Z-`-a-ĶfUPB2hC[hq_'6PJnHo(tHhhq_'PJnHo(tH hq_'o(!jhbhq_'CJEHUaJ2j6DK hbhq_'CJPJUVaJnHo(tHhbhq_'CJaJo(hbhq_'CJaJjhbhq_'CJUaJhPih;6PJnHo(tHhh;PJnHo(tHhh;PJnHtHhh`PJnHo(tHhjPJnHo(tHhsPJnHtHhsPJnHo(tHa-b---------- . ........-...Z.[.\...´㤓㴃rer[Mh hq_'PJnHo(tHhq_'PJnHtHhhq_'PJnHtH!hphq_'6H*PJnHo(tHhphq_'6PJnHo(tH!hC[hJ6H*PJnHo(tHhC[hJ6PJnHo(tHhhq_'PJnHo(tH!hC[hC[6H*PJnHo(tHhC[hC[6PJnHo(tHhq_'PJnHo(tH!hC[hq_'6H*PJnHo(tHZ-.0111122347\9t99: :8:;`gd+NN$a$gdgdI $`a$gdH:G ,WD`,gdw#$ gdy ,WD`,gd$a$gd` ,WD`,gd]u ,WD`,gdogd0...../"/=/>/@/A/F/G/V/W////P0Q0e0f0g00000000庯xhhW!h\h:$6H*PJnHo(tHh\h:$6PJnHo(tH!htmh:$6H*PJnHo(tHhtmh:$6PJnHo(tHh2PJnHo(tHhXPJnHo(tHh:$PJnHo(tHh:$PJnHtH!hF&h:$6H*PJnHo(tHhF&h:$6PJnHo(tHhh:$PJnHo(tHhh:$PJnHtH00000 1 1 1Y1111111111111111111ֻ֟֔pcPCj hC5}h+IEHU$j?K h+IPJUVnHo(tHj@hn: h $EHU$j6=K h $PJUVnHo(tH h`o(h`jh`Uh`PJnHo(tHhZTPJnHo(tH!hxh`6H*PJnHo(tHhxh`6PJnHo(tHhA2PJnHo(tHhh`PJnHo(tHhh`PJnHtHhU-h:$PJnHo(tH11111111111 2222222222222ຩynbSKChyheo(hyhyo(hCJPJaJnHo(tHh hyCJaJo(hYPJnHo(tHhyRh;]6PJnHo(tHhPPJnHo(tHh;]PJnHo(tHh;]PJnHtH!j8 hT%h= CJEHUaJ2jŲK hT%h= CJPJUVaJnHo(tHhT%h`CJaJo(hT%h`CJaJjhT%h`CJUaJ h`o(22M3O3P3u3|33333333333333N4Z4_4c4d4444ɾ}rg\Q\g\gh1yPJnHo(tHhPJnHo(tHh FPJnHo(tHhhPJnHo(tHh?PJnHo(tHh&\PJnHo(tHh&\PJnHtHhk_PJnHo(tHho;hPJnHo(tHhRPJnHo(tHhN+PJnHo(tHheTPJnHo(tHheTPJnHtHh ;PJnHo(tHh-4PJnHo(tHh=PJnHo(tH445)5<5?5S5w5{5555550616z6777!7*7X7Y7777ҼǼǼǦ҇|q|g\QGhePJnHtHh-PJnHo(tHhJ PJnHo(tHhJ PJnHtHhUPJnHo(tHh{PJnHo(tHh{PJnHtHh@PJnHo(tHhPJnHtHhPJnHo(tHh>TPJnHo(tHhwPJnHo(tHh{PJnHo(tHhxPJnHo(tHhhPJnHo(tHhPJnHo(tHhPJnHtH77777788r8s888888889S9Z9[9\9]9^9o9p9q9r9s9t9~vrmrZMv~jhh*:EHU$jfK h*:PJUVnHo(tH h>o(h>jh>Uh+NNPJnHo(tHhtPJnHo(tHh#2PJnHo(tHhPJnHo(tHh? PJnHo(tHh4XcPJnHo(tHh$PJnHo(tHh$PJnHtHhRbPJnHo(tHhRbPJnHtHhePJnHtHhePJnHo(tHt9u9v999999999999999999999seWeGeh/Ah.y6PJnHo(tHhQh5 PJnHo(tHhQh.yPJnHo(tHh/AhW6PJnHo(tHhQhWPJnHo(tH!h/AhH:G6H*PJnHo(tHh/AhH:G6PJnHo(tHhQhH:GPJnHo(tHhThTPJnHo(tHjWhh9EHU$j-fK h9PJUVnHo(tH hTo(hTjhTU9: : :::::: :!:":3:4:5:6:7:8:>:?:@:a:b:c:e:g:鵭vfUfUHh6PJnHo(tH!h#h6H*PJnHo(tHh#h6PJnHo(tHhrhrPJnHo(tHjhhrEHU$jK hrPJUVnHo(tH hro(hrjhrUhPJnHo(tHjhhrEHU$jK hrPJUVnHo(tH ho(hjhUhQhPJnHo(tHg:h:i::::::::::::::::::::::::: ;ԿԿԢԄshXGXG!h&Fh6H*PJnHo(tHh&Fh6PJnHo(tHhfPJnHo(tH!hfhr6H*PJnHo(tHhfhr6PJnHo(tHhfhfPJnHo(tHhfhfPJnHtHhrhr6PJnHo(tHhrPJnHo(tHhPJnHtHhPJnHo(tH!hTh6H*PJnHo(tHhTh6PJnHo(tH ; ; ; ;.;/;0;5;6;7;8;9;?;D;;;;;ɹzodYNC8-hW>?PJnHo(tHh#gPJnHo(tHhQdPJnHo(tHh]iPJnHo(tHhPJnHo(tHhrnPJnHo(tHh1PJnHo(tH!hTh86H*PJnHo(tHhTh86PJnHo(tHh86PJnHo(tH!h#h86H*PJnHo(tHh#h86PJnHo(tHh8PJnHo(tHhPJnHo(tH!hTh6H*PJnHo(tHhTh6PJnHo(tH;;;;;f<g<<<<<<<<<==!=)=5=6=U=Z=l=ǼǩynycyYNCNhg~pPJnHo(tHh%PJnHo(tHh%PJnHtHh*[PJnHo(tHhPJnHo(tHht#PJnHo(tHht#PJnHtHh?A*AAC8EhEG~IDJ2KL-L PbU gdc $,WD`,a$gd4 ,WD`,gdk`gd](t ,WD`,gd' gdF ,WD`,gdon gdlx ,WD`,gdAl=o=v=w===== >>>>>>>>>-?2?7???]@^@@@AAAɾɨ~titi^iSHhklPJnHo(tHhPJnHo(tHh~{PJnHo(tHhfPJnHo(tHhfPJnHtHhvR4PJnHtHhBPJnHo(tHhPJnHo(tHhPJnHtHht#PJnHo(tHhl-PJnHo(tHhsPJnHo(tHh}{PJnHo(tHh PJnHo(tHh PJnHtHh%PJnHo(tHhL,PJnHo(tHAAAAAA A)A*A+A7A\AAAAAAaBbBBBBB@CACdCCCC󶫶seeseseZPha+PJnHtHhJPJnHo(tHhheH PJnHo(tHhheH PJnHtHheH PJnHo(tHheH PJnHtHhyPJnHo(tHhp?PJnHo(tHh}x?PJnHo(tHh}x?PJnHtHh hSdCJaJo(h hZCJaJo(hSdCJPJaJnHo(tHh hLXCJaJo(h hCCJaJo(CCCCCCCCCDDD&D(DJDLDDDDDDDD͠ͲueTG9hkh [PJnHo(tHhkh [PJnHtH!h7h!+6H*PJnHo(tHh7h!+6PJnHo(tHh[h&]6PJnHo(tHh7h&]6PJnHo(tHhkjPJnHo(tH"hG=h3v}PJmHnHo(tHuhG=h&]PJnHo(tHhG=h&]PJnHtH!jhG=h&]PJUnHtHh{JPJnHo(tHhf* PJnHo(tHha+PJnHo(tHDDDDDDDD EEEEE6E8E:Eohi6PJnHo(tHh>oh!+6PJnHo(tHhkh!+PJnHo(tHh:h!+6PJnHo(tHh:h2o 6PJnHo(tH!h7hC|6H*PJnHo(tHh7hC|6PJnHo(tHhG=h!+PJnHtH"hG=h3v}PJmHnHo(tHuhG=h:PJnHo(tHhG=h:PJnHtH!jhG=h:PJUnHtHhG=h!+PJnHo(tHEEEE2F4F\F^F`FtFvFxFzF|FFFFFFFFFFFĴңvhhh[hMhG=h|PJnHo(tHhG=hG@PJnHtHhG=hG@PJnHo(tH"hG=h3v}PJmHnHo(tHuhG=hPJnHo(tHhG=hPJnHtH!jhG=hPJUnHtHhLnhMn6PJnHo(tHhG=hMnPJnHo(tHhG=hzKiPJnHo(tHhG=h!+PJnHo(tH!h>oh!+6H*PJnHo(tHFFFFF,GBGDGFGGGGGGGGGGõzl^P?2hG=h0PJnHtH!jhG=h0PJUnHtHhG=h&]PJnHo(tHhG=hAUPJnHo(tHhG=hF*PJnHo(tHh9h2o 6PJnHo(tHhG=h2o PJnHo(tHh9h2\6PJnHo(tHhG=h<PJnHtHhG=h ohLn6H*PJnHo(tHh>ohLn6PJnHo(tHGHHHHHHBH^HlHrHH*I.IhIlI|I~III񳥗{m{m{_NAhG=h?PJnHtH!jhG=h?PJUnHtHhG=hG@PJnHo(tHhG=hnPJnHo(tHhG=hPJnHo(tHhG=h3 PJnHo(tHhG=hw,PJnHo(tHhG=hE%PJnHo(tHhG=hxPJnHo(tH"hG=h3v}PJmHnHo(tHu!jhG=h0PJUnHtHhG=h0PJnHtHhG=h0PJnHo(tHIIIIIIIJBJDJFJHJ\J^J`JbJdJfJJJ0K2K4K6KJKLKNKPKRKTKbK񳥗{j]O]jjO{hG=hPJnHo(tHhG=hPJnHtH!jhG=hPJUnHtHhG=hgNlPJnHo(tHhG=h+PJnHo(tHhG=hPJnHo(tHhG=hCQPJnHo(tHhG=h PJnHo(tH"hG=h3v}PJmHnHo(tHu!jhG=h?PJUnHtHhG=h?PJnHtHhG=h?PJnHo(tHbKdKfKzK|K~KKKKKKKKKKKKKKKKKKKKLLLL!L#L,L-L.LӢӇ{l{l`{Vh'XPJnHtHh h6&CJaJo(h6&CJPJaJnHo(tHh hzCJaJo(hG=h&]PJnHo(tHhG=h"}PJnHtH!jhG=h"}PJUnHtHhG=h"}PJnHo(tH"hG=h3v}PJmHnHo(tHuhG=hgNlPJnHo(tHhG=hgNlPJnHtH!jhG=hgNlPJUnHtH!.LYLLLLLMFMIMJMRM\M]M^MuMvMwMMMMMMMɾ|q|]SI?hD3MPJnHtHhrPJnHtHhkPJnHtH'jhwg>h#o1PJUnHtHhwg>PJnHo(tHhwg>PJnHtHjhwg>PJUnHtHhPJnHtHh-PJnHo(tHheuPJnHtHheuPJnHo(tHh/ PJnHo(tHhSPJnHo(tHhSPJnHtHhytPJnHo(tHh-PJnHo(tHh'XPJnHo(tHMMMMMN1N5N@NWNYNNNNO O=O>OPOOOO P PPϺϺ{peZOBh{(h{(PJnHtHhOCPJnHo(tHhzPJnHo(tHh'rPJnHo(tHhcPJnHo(tHhcPJnHtHhFPJnHo(tHhFPJnHtHh`PJnHo(tHhcrPJnHtHhcrPJnHo(tHhPJnHo(tHhgPJnHtHhgPJnHo(tHhLhD3M6PJnHo(tHhD3MPJnHtHhD3MPJnHo(tHPoPuPPPPPPPPPQQQQ'Q*Q+Q,Q.Q/QNQWQXQYQZQ\QaQbQcQRRRR RRRRRRξ|ٮ|lh>@h{6PJnHo(tH!hu(hu(6H*PJnHo(tHhu(hu(6PJnHo(tH!hu(h{6H*PJnHo(tHhu(h{6PJnHo(tHhRhR6PJnHo(tHhRPJnHo(tHhkh{PJnHtHh8PJnHo(tHhkh{PJnHo(tH'RRRR R"R*R,R.R0R2R:RR@RHRJRLRRRRೢscRBhbh{6PJnHo(tH!h>@h 'U6H*PJnHo(tHh>@h 'U6PJnHo(tH!h>@h5M26H*PJnHo(tHh>@h5M26PJnHo(tHh>@6H*PJnHo(tH!h>@h>@6H*PJnHo(tHh>@h>@6PJnHo(tHhkh{PJnHtHh>@h{6PJnHo(tHhkh{PJnHo(tH!h>@h{6H*PJnHo(tHRRRSS SLShSjSxSzSSSSSSSSSSSSTT T T T TѴљѴxgWFWFW!h4h46H*PJnHo(tHh4h46PJnHo(tH!hu(h46H*PJnHo(tHhu(h46PJnHo(tH!h4h{6H*PJnHo(tHh4h{6PJnHo(tHh)>PJnHo(tHhkh{PJnHtHh)>h{6PJnHo(tHhkh{PJnHo(tH!h>@hb6H*PJnHo(tHh>@hb6PJnHo(tH TTnTvTwTxTyTzT{T|T}T~TTTTT9U:UaUbUcUdUuUvUwUxUyUzUͿ各|s|dSHh](tPJnHo(tH!jrh"h](tEHUnHtHjv.zK h](tUVnHtHh](tnHo(tHh](tnHtHjh](tUnHtHhkh](tPJnHo(tH!h>@h%"x6H*PJnHo(tHh>@h%"x6PJnHo(tHhkh%"xPJnHo(tHh%"xPJnHo(tHhkh{PJnHtHhkh{PJnHo(tHh46PJnHo(tHbUzUUUUUUUzSkd@$$$Ifl0L t644 la$$Ifa$gdMl'$$Ifa$gdMl'gd](tzU{U|UUUUUUUUUUUUUUUUUUUUUUUŴ嗆zrirZIz>h](tPJnHo(tH!j$h"h](tEHUnHtHj.zK h](tUVnHtHh](tnHo(tHh](tnHtHjh](tUnHtH!j"hG&h~/EHUnHtH&jiK hG&h~/UVnHo(tHh~/nHo(tH!jhG&h~/EHUnHtH&jGiK hG&h~/UVnHo(tHhG&h~/nHo(tHhG&h~/nHtHjhG&h~/UnHtHUUUUUUUUUUUUUUUUUUUVV V V V V VVV V!VŴ嗆raM&j3iK hG&h>zUVnHo(tH!j,hG&h>zEHUnHtH&jiK hG&h>zUVnHo(tH!jX)hG&h>zEHUnHtH&j,iK hG&h>zUVnHo(tHh>znHo(tH!j&hG&h>zEHUnHtH&jiK hG&h>zUVnHo(tHhG&h>znHo(tHhG&h>znHtHjhG&h>zUnHtHUU V$V%V+Skd~0$$Ifl0L t644 la$$Ifa$gdMl'$$Ifa$gdMl'Skd+$$Ifl0L t644 la!V"V#V%V.V0V1V3V7VAVSVTVUV`VaVeVfVgVqVVVVV W W0W6W7W8W9WWW¹¹­­}of}Wh5 2hK%6H*nHo(tHhK%nHo(tHhS\hK%6nHo(tHh5 2hK%6nHo(tHh%PJnHo(tHhG&hK%nHo(tHhG&hK%nHtHhG&h](tnHo(tHh](tnHo(tHhVPJnHo(tHhVnHo(tHh>znHo(tHjhG&h>zUnHtH!jQ.hG&h>zEHUnHtH%VfVWW X%X&XZYZ8ZqfS$IfgdOl' ,WD`,gdD7Skd5$$Ifl0L t644 la$$Ifa$gdMl'$$Ifa$gdMl'`gd](tgd](t WWWWWWWWWWX X X X X XXX X!X"X#X$X&X'XVXXXʻvbQvFhG&hnnHtH!j3hG&h EHUnHtH&jiK hG&h UVnHo(tHh nHo(tH!j1hG&h EHUnHtH&jiK hG&h UVnHo(tHhG&h nHo(tHhG&h nHtHjhG&h UnHtHh](tPJnHo(tHh|5PJnHo(tHh](tnHo(tHhG&h](tnHo(tHhG&h](tnHtHXXYXZX[XrXsXtXuXyXzX{XYY7YCYKYXYZY[Y`YaYZZZZZܽѧ}pcWLh](tPJnHo(tHhBf6H*nHo(tHhz16PJnHo(tHhThBfQJnHtHhG&hBfnHtHhBf6nHo(tHh](tnHo(tHhG&h](tnHtHhG&h](tnHo(tHhrPJnHtH'jM6hIhnPJUnHtHhIPJnHo(tHhIPJnHtHjhIPJUnHtHhnPJnHo(tHZ Z Z.Z0Z2Z4Z6Z8Z:Zh gh!W~EHUnHtH$jK h!W~PJUVnHo(tHZZ]J]^6^f^m$$Ifa$gd'[l'$$Ifa$gd'[l'gd](t$a$gd=ZgdSkdB$$Ifl0F{7 t644 la[[[ [0[2[4[6[8[F[H[L[N[X[\[^[`[b[d[l[n[p[r[t[[[꥔~ra~U~ra~JhePJnHo(tHhG&h,xnHo(tH!h!W~h,x6H*PJnHo(tHh#Uh,x6nHtHh,xPJnHo(tHh'PJnHo(tH!h'h'6H*PJnHo(tHh'h'6PJnHo(tHhG&h+GfnHo(tHhS\h+Gf6nHo(tHh+GfOJQJnHo(tHh5 2h+Gf6nHo(tHhG&h](tnHo(tHh](tnHo(tH[[[[[[[[[[[\b\d\f\h\n\r\t\\\\\\\øΪߒßxd[LAhG&h>xnHtHhg=h>x6H*nHo(tHh>xnHo(tH'h5 2h>xCJ OJQJaJ nHo(tHh5 2h>xnHo(tHh!W~h!W~PJnHo(tHh!W~6PJnHo(tHh_|DPJnHo(tHh!W~6H*PJnHo(tHhjPJnHo(tHh!W~PJnHo(tH!h!W~h!W~6H*PJnHo(tHh#Uh!W~6nHtHh"PuPJnHo(tHh"PuPJnHtH\\\\\]]]]]@]B]D]F]H]J]L]N]p]r]t]󾳤yhYNN:&j*iK hG&h](tUVnHo(tHhG&h](tnHtHjhG&h](tUnHtH!jChG&h=ZEHUnHtH&jfiK hG&h=ZUVnHo(tHhG&h=ZnHo(tHhG&h=ZnHtHjhG&h=ZUnHtHh=ZPJnHo(tHh](tnHo(tHht PJnHo(tH!ht ht 6H*PJnHo(tHht ht 6PJnHo(tHhG&h](tnHo(tHt]v]x]]]]]]^^ ^,^.^0^2^4^6^8^:^\^^^`^b^d^h^n^x^z^|^~^ӿӿȬ߿Ȉw߿ӿl]RhG&h=ZnHtHjhG&h=ZUnHtHh=ZPJnHo(tH!jKhG&huEHUnHtH$jOK huPJUVnHo(tH!jHhG&hkEHUnHtH$jK hkPJUVnHo(tHh](tnHo(tHhG&h](tnHtHhG&h](tnHo(tHjhG&h](tUnHtH!jlFhG&h](tEHUnHtHf^h^z^^^^4_d_``aHacd|qq ,WD`,gd'gdK $`a$gdK`gdK$a$gdG4$a$gd=Zgd](tSkdM$$Ifl0L t644 la ~^^^^^^^^^^^^^^^^^^_ _2_4_6_ĵveWL=jhG&hG4UnHtHhuPJnHo(tHh=Zh=ZPJnHo(tH!jQhG&h=ZEHUnHtH&jz {{{޺ޣud!j=h!h!PJUnHtHhc6hL)HPJnHo(tHh0PJnHo(tHhL)HPJnHtHhL)HPJnHo(tHhvhL)HCJaJo(h; PJnHo(tHhyFPJnHo(tHh,aPJnHo(tHhbchc6PJnHo(tHhc6PJnHo(tHhc6PJnHtHh3v}hlCJaJo(&{{{{{-{.{/{0{3{Q{R{X{Y{Z{\{]{c{i{𮜍~p^L^:^:#hbchFrCJPJaJnHo(tH#hbchJoCJPJaJnHo(tH#hbchx;CJPJaJnHo(tHh%'CJPJaJnHtHhgZCJPJaJnHo(tHhx_CJPJaJnHo(tH#h/Z hUA<CJPJaJnHo(tH)j,Wh/Z h/Z CJPJUaJnHtHhFsCJPJaJnHo(tHhpCJPJaJnHo(tHh*sCJPJaJnHo(tHhUA<CJPJaJnHo(tHR{j{y{}7B gduiy`gduiy gd9$a$gdx<$a$gdJpggdT<$a$gdMgd$a$gd 6$a$gd> ,WD`,gd ,WD`,gdB. gdB.$a$gdi{j{k{m{y{z{4|5|`|c|||}}e}}}}}~6~7~~~~~{µªn_nh `fCJPJaJnHo(tH#h2|h`RCJPJaJnHo(tHh`RPJnHo(tH!jehW}h\ PJUnHtHhB.hB.PJnHo(tHhTPJnHo(tHh'yhB.PJnHtHh'yhB.PJnHo(tHhaPJnHo(tHhB.PJnHtHhB.hB.PJnHo(tHhLgPJnHo(tH!ο|m_M;M,hTCJPJaJnHo(tH#hDkh=cMCJPJaJnHo(tH#hDkhhr CJaJhS>hr CJaJo(#hr h CJOJQJ^JaJo(hQh CJaJh]$CJPJaJnHo(tHBpVLMM\N#OQPPrQRgRRRRRRSSSS!S"S*S+S$a$gd & F*$1$gds] & F*$1$gdEcJob Scheduling Strategies for Parallel Processing, Springer Verlag, pp.162-178, 1999. R. Buyya, D. Abramson, and J. Giddy, Nimrod/G: An Architecture for a Resource Management and Scheduling System in a Global Computational Grid, Proc. High Performance Computing ASIA, 2000. M. Litzkow, M. Livny, and M. Mutka, Condor A Hunter of Idle Workstations, Proc. 8th Int. Conf. on Distributed Computing Systems, pp. 104-111, 1988. L. Chen and G. Agrawal, A Static Resource Allocation Framework for Grid-based Streaming Applications, Concurrency and Computation: Practice and Experience, 18:653666, 2006. B. Agarwalla, N. Ahmed, D. Hilley, and U. Ramachandran, Streamline: a Scheduling Heuristic for Streaming Applications on the Grid, Proc.13th Annual Multimedia Computing and Networking Conf., 2006. A. Ramakrishnan, G. Singh, H. Zhao, E. Deelman, R. Sakellariou, K. Vahi, K. Blackburn, D. Meyers, and M. Samidi, Scheduling Data-intensive Workflow onto Storage-Constrained Distributed Resources, Proc. 7th IEEE Int. Symp. on Cluster Computing and the Grid, Rio de Janeiro, Brazil, pp. 401-409, 2007. L. Battestilli, et al., EnLIGHTened Computing: An Architecture for Co-allocating Network, Compute, and other Grid Resources for High-End Applications, Proc. HONET2007. A. Takefusa, et al., G-lambda: coordination of a grid scheduler and lambda path service over GMPLS, Proc. iGrid2005. I. Foster and C. Kesselman, Globus: A Metacomputing Infrastructure Toolkit, Int. J. Supercomputer Applications, vol. 11, No. 2, pp.115-128, 1997. J. H. Holland, Adaptation in Natural and Artificial Systems, University of Michigan Press, 1975. K. Kar, S. Sarkar, L. Tassiulas, A Simple Rate Control Algorithm for Maximizing Total User Utility, Proc. Infocom 2001. Grid Resources Bandwidth Storage Processor Queue Management Wrapper Wrapper Dispatcher Grid Directory Services Security Resource Discovery Other Middleware Services Scheduler Client Tool Management Engine Client Tool Bandwidth Remote Data Source Storage User process Wrapper GRAM server MDS server Dispatcher Scheduler Job submission Real-time data supply Figure 2. Overall mechanism L1LNLTL{L|LLLL MM/M0MCMDMbMiMkMlMyMzMMMMMMMMMM㳧ti]Uh02CJaJhs]hr]uCJaJo(hr]uhr]uCJaJhgwhCJaJhhi6CJaJo(h02hi6CJH*aJhhi6CJaJhgwhiCJaJo(hgwhiCJaJhw+hi6CJaJo(hs]hiCJaJhs]hk@CJaJo(hs]hiCJaJo(hk@hi6CJaJo(UMMMMMMMMMMMMNNN:N>NHNJNTNVNZN[N\NhNiNjNkNlNuN···«znbnWKWh-h` CJaJo(h-h` CJaJhs]hkuKCJaJo(hs]hiCJaJo(hSMnhr]uCJaJo(hSMnhSMnCJaJhs]hSMnCJaJo(h hSMn6CJaJo(h hSMn6CJaJhs]hICJaJhs]hICJaJo(hs]hSCJaJhs]hr]uCJaJo(hr]uhr]uCJaJh02CJPJaJnHo(tHuNvNNNNNNNNNNNNNNNNNNNNNO"O#O$O%O5O6O?O@OHOIOTOUOdOeOmOnO{O|OOOOƺƺƺƺƺѮznznznznznznznznznzhs]h;YCJaJo(hs]h;YCJaJhf3h;YCJaJo(hM!;hi6CJH*aJo(hM!;hi6CJaJo(hs]hiCJaJo(h-h/CJaJo(h-h/CJaJhs]hiCJaJhs]h` CJaJo(h-h` CJaJh-h` CJaJo(+OOOOOOOOOOP$P4P6P>PPPQPiPjPPPPPPPPPPP󥚌q]QhmGHhmGH6CJaJ&hmGHhmGH6CJPJaJnHo(tHhrCJPJaJnHo(tHhmGHhmGHCJaJo(hrCJPJaJnHtHhmGHhmGHCJaJhmGHh;YCJaJo(h] h;Y6CJH*aJo(h] h;Y6CJH*aJh] h;Y6CJaJh] h;Y6CJaJo(hs]h;YCJaJhs]h;YCJaJo(PPPQQ;Qhf3CJaJo(hs]h]::CJaJo(h]::hi6CJaJo(hs]hiCJaJhs]hiCJaJo(hhrCJaJo(hrhr6CJaJ&hrhr6CJPJaJnHo(tHhrCJPJaJnHo(tHhrhrCJaJo(hrCJPJaJnHtHhrhrCJaJhrhmGHCJaJo(hmGHCJPJaJnHo(tH R RRRARgRRRRRRRRRRRRRRSSSS!S"S*S4S5SĸĬĸnnnZZI h6Uh%'CJPJaJnHtH&h`h%'5CJ PJaJ nHo(tH&hDPh%'5CJ PJaJ nHo(tHh%'#h{h%'CJPJaJnHo(tH&hO;hO;6CJPJaJnHo(tHhO;h6CJaJhs]hCJaJo(hs]hCJaJhjhi6CJaJo(hjhi6CJaJhs]hiCJaJo(hs]hiCJaJ+S,S-S.S/S0S1S2S3S4S6S7S8S9S:S;SS?SGSHSSSTSUSVSnSoSpS$a$gdgd5S6S?SGSHSSSVSWSnSrS{S|SSSSSSSSSSSSSSSSSSSSSygy\PhLqh%'CJ aJ o(hLqh%'CJ aJ #h @h%'5CJ PJaJ nHtH&hhh%'5CJ PJaJ nHo(tH h6Uh%'CJPJaJnHtH h25CJ PJaJ nHo(tH#h`h%'5CJ PJaJ nHtH&h @h%'5CJ PJaJ nHo(tH&h`h%'5CJ PJaJ nHo(tHh%'#h6Uh%'CJPJaJnHo(tHpSqSrS{S|SSSSSSSSSSSSSSSSSSSSSSSS$a$gd gdQ{$a$gdSST T TTT#T$T,T-T2T9T:T>TETFTQTRT\T]TlTmTTTTTT$a$gd{gd2,$a$gd SSTTT T T TTTTT#T$T,T-T9T:TETFTQTRT\T]T^TlTmTnTTT÷|peYh#؞Jߊ DB:;( K`6VIpS |7?Kx3\ϥQh`ba[].p G#4n= 121)W2lԡ1 ~hDd b  c $A? ?3"`?2AʰSN| |Ӏ;`!AʰSN| |Ӏܲ @CTxڝR=KASs <QJ+EL  #x R*!XO5SٽKdafEX,#C)"!q:'.G93V'ab)}>F>>=iJn|:a|(k^#O->EWoXk=x'[-&ZxaRW3]$Z)F3.p:z{'KDDd Db  c $A? ?3"`?2R-?LJO9v ;`!R-?LJO9vb  |xcdd`` @c112BYL%bpuy \;p?gfbqjn?<4NPs0c[ k0vP{ M-VK-WM#| X0YM'A?p8W;Q 309s1@J0lf csAS#82mEJmLLJ% A2u(2tA4Aga0a Dd Db  c $A? ?3"`?2I5{ɯ`{| ;`!I5{ɯ`{b yxcdd`` @c112BYL%bpuobҶUsi#jl O+!)e((n@`sAmǝ 1j, #t+aL ԞA {B2sSRst;s2;Ȅ =L&ai%8+a|>0g`bD3BJ0~c%3]pD )l;- `6F&&\- {:  0z0eDd h  !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~Root Entry F:^@5Data |WordDocumentipObjectPool,'K:^_1269515318F'K'KOle CompObjiObjInfo  !$'*-./2569<=@EHKPUX[`cfknqvyz{|}~ FMicrosoft ʽ 3.0 İ DS Equation Equation.39qiPD R s =[p s b s st s ]Equation Native _1269447990 F'K'KOle CompObj i FMicrosoft ʽ 3.0 İ DS Equation Equation.39q\e4 p s n  >P s n   FMicrosoft ʽ 3.0 İ DS Equation Equation.39qObjInfo Equation Native  x_1269448610 F'K'KOle  CompObjiObjInfoEquation Native _1269543621"F'K'Koe4 b s n  >B"b ss"S R  " FMicrosoft ʽ 3.0 İ DS Equation Equation.39qOle CompObjiObjInfoEquation Native wЁD st s n  >S"st ss"S R  " FMicrosoft ʽ 3.0 İ DS Equation Equation.39qqHD w i t_1269523986F'K'KOle CompObjiObjInfoEquation Native _1269524013F'K'KOle "CompObj #i()=f i w io ,t() FMicrosoft ʽ 3.0 İ DS Equation Equation.39qa5$ f i w io ,0()=w ioObjInfo!%Equation Native &}_1269883082'$F'K'KOle ( FMicrosoft ʽ 3.0 İ DS Equation Equation.39q@~< w i t()=f i w io ,t()=w io +d i *ft,T i ()CompObj#%)iObjInfo&+Equation Native ,_1269883133)F'K'KOle 0CompObj(*1iObjInfo+3Equation Native 4 FMicrosoft ʽ 3.0 İ DS Equation Equation.39qp`1 ft,T i ()=floor(t/T i )_1269440534T.F'K'KOle 7CompObj-/8iObjInfo0: FMicrosoft ʽ 3.0 İ DS Equation Equation.39qk q i =r i r jj=1n " FMicrosoft ʽ 3.0 İ DSEquation Native ;_12662984863F'K'KOle >CompObj24?i Equation Equation.39qat P. FMicrosoft ʽ 3.0 İ DS Equation Equation.39qd@UD maxU ObjInfo5AEquation Native B-_1265215047^@8F'K'KOle CCompObj79DiObjInfo:FEquation Native G_1265205272EJ=F'K'Ks x s () s"S " FMicrosoft ʽ 3.0 İ DS Equation Equation.39q=}2 1() FMicrosoft ʽ 3.0 İ DSOle ICompObj<>JiObjInfo?LEquation Native M8_1266298515mBF'K'KOle NCompObjACOiObjInfoDQ Equation Equation.39qa s.t. FMicrosoft ʽ 3.0 İ DS Equation Equation.39q=?؄T x s d"Equation Native R5_1265205160OGF'K'KOle SCompObjFHTiObjInfoIVEquation Native W[_1265205292LF'K'KOle YI s"S " FMicrosoft ʽ 3.0 İ DS Equation Equation.39q=D} 2() FMicrosoft ʽ 3.0 İ DSCompObjKMZiObjInfoN\Equation Native ]8_1265205201QF'K'KOle ^CompObjPR_iObjInfoSaEquation Native bL Equation Equation.39q=0i x s "X s FMicrosoft ʽ 3.0 İ DS Equation Equation.39q_1265205299;6VF'K'KOle dCompObjUWeiObjInfoXgEquation Native h8_1265206007[F'K'KOle iCompObjZ\ji=pwd 3() FMicrosoft ʽ 3.0 İ DS Equation Equation.39q=MU x s d"I s"S A  "ObjInfo]lEquation Native mi_1265206031Yw`F'K'KOle oCompObj_apiObjInfobrEquation Native s8_1269520425eF'K'K FMicrosoft ʽ 3.0 İ DS Equation Equation.39q=dW 4() FMicrosoft ʽ 3.0 İ DS Equation Equation.39qOle tCompObjdfuiObjInfogwEquation Native xV:Bj x sk+1() =x sk() + k U ' x sk() ()[] X s  ifx sk() d"I s"S A  "  k x sk() [] X s  ifx sk() >I s"S A  " { FMicrosoft ʽ 3.0 İ DS Equation Equation.39q=y 5()_1265207988jF'K'KOle CompObjikiObjInfolEquation Native 8_1269439467roF'K'KOle CompObjnpi FMicrosoft ʽ 3.0 İ DS Equation Equation.39qdhl x sk+1() =0,"s "S A FMicrosoft ʽ 3.0 İ DS Equation Equation.39qObjInfoqEquation Native _1266395231tF'K'KOle CompObjsuiObjInfovEquation Native 8_1265209958h|yF'K'K 6() FMicrosoft ʽ 3.0 İ DS Equation Equation.39q8? y[] X s  =minB s ,mOle CompObjxziObjInfo{Equation Native axb s ,y()() FMicrosoft ʽ 3.0 İ DS Equation Equation.39q.SD U ' "()_1265210154~F'K'KOle CompObj}iObjInfoEquation Native J_1269439954F'K'KOle CompObji FMicrosoft ʽ 3.0 İ DS Equation Equation.39q•HKt" U"()=U s x sk() () s"S A  " FMicrosoft ʽ 3.0 İ DSObjInfoEquation Native _1269518286cF'K'KOle CompObjiObjInfoEquation Native 8_1269439964F'K'K Equation Equation.39q(D 7() FMicrosoft ʽ 3.0 İ DS Equation Equation.39q¡ U ' xOle CompObjiObjInfoEquation Native  sk() ()="U"()"x sk() FMicrosoft ʽ 3.0 İ DS Equation Equation.39qi8TD U s x s ()=0,"s "S A      !"#$%&'()*+,-./0123456789:;<=>?@ACDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~_1265224764F'K'KOle CompObjiObjInfoEquation Native _12652248921F'K'KOle CompObji FMicrosoft ʽ 3.0 İ DS Equation Equation.39qsQD U"()=U s x s () s"S "ObjInfoEquation Native _1269520176F'K'KOle  FMicrosoft ʽ 3.0 İ DS Equation Equation.39q›h~D U s x s ()= s ln1+x s (),s"SOh+'0CompObjiObjInfoEquation Native 1TableEb  c $A? ?3"`?2 MfD:|O ;`! MfD:|O @|~xcdd``fdb``baV d,FYzP1n:&B@?b = UXRY7`7LY \p[˭0ÊU7P4MB]j mU8 ^JB r16D@2wJ"ϝs`%ڽ,Wٽ DAd++& %0_`|6&T w)']}J0~8\|F=<( px`3=> #l󹠹 ΅v0omLLJ% s:  02 W|Dd xhb  c $A? ?3"`?2FɄt+ ob啢i'X u3q1fݫQI<]. 12H^c 5W?\(ܐbr<5!vPfWP| ȳ? @&/QM`{0/> ƙ #lKcw*f_' ڸ)l 4 0y{@Ĥ\Y\pdP"CD|f@3XcHDd hb   c $A ? ?3"`?20kxQrV ;`!kxQrVr@+ |xcdd`` @c112BYL%bpuobiY u9W,{5* '%5y Ә@FhV+ 1nV; pޫKh'Ž "U5$ 27$37X/\!(?71H>`\* 2!LJN™0;*??#qg?܉B Baop. ip)p7_7pcCjzzt2P %`k y,_Q^p{.- sAK'.p .l``H#RpeqIj.j @ ] @Y= 2GDd hb   c $A ? ?3"`?2#hň}y;`!#hň}y @|xcdd``> @c112BYL%bpu?JT Tը$d^,W:n@$s2Єa.ܐbr<5Qb.L  a`${dBf% B/`TrAlԔ_׳3%p~#XA, #ls) QM `P>hcdbR ,.Ie2C 0z0e뺋Dd b   c $A ? ?3"`?2ctצmu0;`!ctצmu0~`@ 8 xcdd``$d@9`,&FF(`TɁ-MRcgbR RnĒʂT@_L ĺE`A,6;3 Q9@&&ܤh./ +?Obbiv3@vqme`i}Z 1O?.ވs]mحf6jn? $X bbd12,(ㆫa:xQ( TA?dAm@ UXRY`7S?$LY ZZбcD V0\ϝ&T T5ի<ě2n *$ȇ\`n``¤I)$5A NE.B`;@n ITDd b   c $A ? ?3"`? 2`@A?r{~]];`!`@A?r{~]]`@ :}xcdd``> @c112BYL%bpu1tX> >L;T Tj@Z5* ]+y :pЄa.ܐbr<5^J.hpCl,4L=ncdbR ,.Ie8 2C D|b@3X?)b$$If!vh55L#v#vL:V'l t06,55LDd Tb  c $A? ?3"`? 2h='Xx'7D%;`!<='Xx'7 XJR xcdd`` $X bbd12,(ㆫaj`T`3H1g`Yl@Р6 憪aM,,He`0P&,e`abM-VK-WMcX|"+|.JSgF\ `U@FUl>Hx@|@GlO =ŕ\@u=0kcdbR ,.IePC  N v320lIbDd b  c $A? ?3"`? 2-̇C2:';`!-̇C2`:Nxcdd``Ned``baV d,FYzP1n:B@?b zjx|K2B* Rt/ @201W&00lZQc@`|HZ< P.P56hb\Us=m#7y>sPsq|5wf1 ZZG{&W^`C L |tf/L%TTrAc 2A``R#RpeqIj.C<E.Y`o+Dd TTb  c $A? ?3"`?2u /fhhT= Q);`!I /fhhT=   XJXJxcdd`` @c112BYL%bpu >I4AHq=`e,2@g>#ȄJlO= \*Q? 0vJ.hqCl:.v0o(.+KRs>ePdk1 ‹iP-Dd @Tb  c $A? ?3"`?2w&m'-#HS.;`!K&m'-#H  XJxcdd`` @c112BYL%bpu >d*F\ 3`sAmǝ 1jƳ #CM$ 27$37X/\!(?71H1^>f 08\|L>z8QL <+aP=\Nq` v 0y{@iI)$5dP"CDHg!t?3e,Dd TTb  c $A? ?3"`?2v8RoYVIH_R3;`!J8RoYVIH_  XJXJxcdd`` @c112BYL%bpu|T 9P>0BLڌmNA΋((% }*P]o1o_6P}^"jmyo#>Xnw{8t[IƼvc\,Q4l+Mw ~O|z:>~/ q#y4,NyHs^+<ߟMwvW׀4($ W1%T*x6nGx@&5 e\Wnj"8]/A^p |RulAv**g4]tַģI_M4;G.RH!*U1_*nb2,+)N['eA~Wq)&x,f2?_1q?;g'N"k'0ER_EzϿ,AJ ÿ}yXzq!mV>wixIJ؃ee߯aO{V}\|a ,{ (xs}>۵zKo=XW1o PmI[aJdG4 (7ȟ%#D"-Dd @Tb  c $A? ?3"`?2w^*NK4(S;;`!K^*NK4(  XJxcdd`` @c112BYL%bpu1tX> >JUUsi#ZFcVJBip6D@27 Ps1 eZZӝs60״ *\ v0o(+KRsA2u(2t5B ~`=]$$If!vh55#v#v:V'l t06,55LDd |b  c $A? ?3"`?2Emf|>\$߈W>;`!Emf|>\$߈ ` 0xcdd``db``baV !,&FF(`T)[RcgbR .qznĒʂTݿnkaR&&ܤ|i:V~, :X u9{5* Fn@$s73Єj.HdnHfnj_jBP~nbNwBe2fO4J/jo?EF_N 0;loXeわDwO=Jp~3oXn{1*8Ĵ[ 2B, 6S__ϰ {9 l,40y{@yI)$5d.P"CDHgat?1eqf=Dd TTb  c $A? ?3"`?2¶HfcfcA;`![¶Hfcf  XJXJ)xcdd`` $X bbd12,(ㆫaRB!?P  :P5< %! `31{ e`A 27)?P!'=i,@@@ڈјU|}Zm]U(1W}D%b-o7O[Hz^]KZ2-L|ީ6>#^nkw$B`m?g^aCMZ8=v\OTH3k$ZYwu0.=xqktoPb)-EoQ+sttrt1'<_[oQp-ODd hb  c $A? ?3"`?2#A۾.m~퉚!uF;`!m#A۾.m~퉚! @|;xcdd``ed``baV d,FYzP1n:&&f! KA?H1j ㆪaM,,He`c&,ebabMa`QAc@`𹼂/m3 P.P56Ǹ^JB]rm#dn!M psG5w Ps#27$37X/\!(?71H~aH:{I2ӽb B+isq(6F&&\ s:@Dg!t?0\Dd Db  c $A? ?3"`?2? $%CsUSRKVH;`! $%CsUSRKVr (xڥMHAj&(I":ԩKuȢKDl ADtc:u":EDnEt+E/̎lL- y  ]` > .G!!iQ#V[))#ב^dp-2<9UXy=Lh>[EAMŭgƤ{ 1F~m͹XsD.1ZR#kBAlͼEx7Ovye֥ǧ%b|y/iN%\/_tv8?Q fy5 ^S` H( Ž 7?g F{iI@c#W KR ;+Lgf4BocKRslOAHV]_H^zM'OC.+978XۧŦC8d@W2_=Pg?W7aDd 0hb   c $A ? ?3"`?2kA۵wI(FlQ;`!kA۵wI(Flr @H|~xcdd`` @c112BYL%bpuIs߀Sܿ  |.h.TF0y{@yI)$5A d/P"CX,Ā&`اCDd P hb  c $A? ?3"`?2\$Blقt. W;`!\$Blقt.@|xcdd``eb``baV d,FYzP1n:&! KA?H1Z@3aP5< %! `3{ vL,L I9 "\>3X|#+|.&T TݫQIȼb\7p? df%tNpsK9vMR+ȃކ PsA sC2sSRstoc?ȄJ߄/sp~<*_K?yBZndw<`eF1B%>=|&:|@Fw' 称 ցs |;vPVÒIj.p{= 121)W20ePdk{> c)Dd 0 " # A""~ҟ5NewF}<Y;@=y~ҟ5NewF}<_1dQ: =Gx |Mgǟj(ƠgLktTADB[$!4Zb+A#ZE әNgVޜSq%79G?{{~{=GlWH|!DdV]aA"J)#RuAQ(XפHʪE%LI;i_d 2\Dw4eϲl]_h7Ũ9997*(XWzKJϻxnoXTl5.t)ypbc9e:N^ݓ!FѲw3aW"FP^=WQ61 ư]$Go_D]"_D9cP`& qgcb2ILL?x9 VJe΄\Yk6o._,΅%׿LRP^YtҢ4G5P GP%2Q:.<4Oy ͓hHy8h^4/ ͅBs\h.4O1s\h.47Yޥ{:јk=zjTG%>c\3糈;gC{̿c|4|.n-sm;Q$N Y^SZqVޥcmsײyØUխn>GncH sO>ka zo#s[%΋kL┽1}/m̘1d.]n0,z͘6cf25ypys寿\FW džRxMk-,0 7}q/kX|c-|c!F_D|u>yԄW\K)6yXx8;& xw|f{ɻV!yqxqxg8;`^~mζvwM> Ӣf]+6X8|\~ Sg3r#3b1ed.8-9[^W6[;7&t _ck|Wao'mL _s]|t< d-`lqq8xx3b\@l20rqMQQ0 #b\@l20rqMAA0n 0xŸNdoac-`\A z`\+ }~8s6$ [\߹Ș>nk/5J_{QRڋ o^Gݯhe@Ơ/0e1<8 O^z_ wŊ{Q' dCɊ Ls1%bE1taDz7ysnǷu^_s}A~} ekf楈ʽF,7ʽE_smD׸FPUej!U S-d !D6`N"U{:@QN_u~DnVݥ!=UOzIw[>U.*B:uP( i",UhiJjb&jb)A*NiQH5FqR=Sj<&J=TW=#u$y\MjU3깐Jʺ J,wwUx9xUy%9JB^E^M_C%Je_K_G_O@߈l 6 q-V" ؆{lEv&;q]ˋnfN8%Fqp \nkĽqWpxdk8uό׌눴w_1^wwwwwŻKUVK[ҭ?Fyz/l t_b_2ˌ+FwQ~hAٌ*{-{{={~lYavDe2 f5}Ylc1{=P PzZ] mNx[-Df{ѱ(yhyHl7I뉨uD:"l-Jĥ}k5Djr5ѹ ^E+D "{H_N/#KɄdijd2d ^L,&{EȦdk E?y0xI&+瓝yd<;M"s9d2|6>EϢ̤̤"̠2LBLRLbLzLLLLLLLL㜿71~cg_}49'|'d.콑{#Y v֘;hi.zoszp.CaOF/EdV4?K=|nkS>ܿ1G~[8o:~(5w>ܟ#L;WGu {J|ާ)T=*)|[Ti*}1z^j'id~Cnǜ{1ه1vݧ:Oakx*8ɽw~uJx,`+DZ@3q8}1d4KxrǖCx1uPQ1y g3Z/Ǜ>n{T{˷{a=>c|>d*l {F?e\ܩ95>y]bCe$DÞbN~Č᧍) ƛxnG؍a_7޿z%d-Dd 0 % # A% "'8/~ Cb\\+pm;@='8/~ Cb\\+p1dQ: =x]lUy?ߋIшErX@7L[PR^bO6aAو`aGJ lHX׬l#k%c^hl*74-<4)J2POz{psk}9w~wbM%ā !~1!?b cc'*XsC6v|3߽\!jQ*x^X+ֈ h7Aml7+\ ٘ Zʵ&M}\ dzd,B߅5{ȵavhYh кBX0>ZifCZ3B- hz]!~n,L`z3L_ `l> Wq8]tL0]tLКbtL0]8 2n!-f&כ4!u7iۗhhsW5tUN%1ߚH.Ng yxF kLݭqx~m'Vt#EmܒIܳKb%O`8 ]s0uc,^n?4.O*F'υz {$A8`^x%:X(,e0'Z v| 0sM. &︸~—.(DLk R,MxmԤ4:Jax?߃6U^ch.H2pށۺz4^p5V?HGˆzSᄏk2ßrk8)F7WW)bUr3=Ƨ]I.Ro&p>.D6Jm!Nܴu0][~V;X4?1tfX[˖؜Nǣx=k% jfr7UΡ1Q0'Wboo}[,mQmq;ҷQb/i0E_ĩ_ԹPﬥYEzg~աޯJMUCwޝޝԻSѻӁ'}BjO`z[ݯ@~KC?V3 {sN \ozli|NoBMC7H ԻRZEZzZ]]@GtFG Y,ޠ,ޠMn~Gw_)PzP>@zP>@F Uݖzw+zw;лRnEnzPwCZ.^]6z(z8cGzXG]7E)z7[)zn/eb˸{bc}pc~Zڷߴ.fQ)6M0{Mc"&'r4Y9vהy{WnKs ٚcu*4 H/37e.5^!2H_{hFrύO@sbp]]=sKbfP̜oU=4L؞y"}~>x'g8kYkyl@og%&y6y5ONxޗx.![s18Hߋ `$=]+7O@W)uʲ&yl@o$&*9њ'a'<'bKcg /3b{׃' Y/)@c x=4L؞[=x-x'HY]QEk-^br11kA<4LcO%9Ʈ1T~_QlOO /:Rsyr<R J.![#C$_4KZ=yBW4j #i0_b>Xb~ʰ s>"19ϖ< ;i*}|%9ƞoߗ'" G=a ½' "b'X9H3J{Fbrgb>syr<R -J.![3C$_4ZJ oȺ0􄑴OX _b-5c؄z \$19E1>kKtOX5 _('`O@^Or';Zi >waԿ]+QE"f&%[(f<{b< ;ّbD%dk *HoQlODՃ' .,{Y 4VN}[0ֶY\}/Px"mv³#>%O5= *Hï=aJZOr'4Jke#-_|af#\brco[$۩. J.![shߗ'" F؞, q=E+}ާZ'>y<}yrs )J.![!x/OPE~K oʺK ozH'4I.(kQkyafg6yК'a'<b'4)l1va|_4 Ώb{B ж@h q֎[x=.191O{9<9w )K<ɗ'" l zAOpPi<ᠡ'}AZ VvZ^1llB1;%&g͓R1]x.![sd>4L|]kO% oOr'6:8XR)%O5؝ *Hï='O8''3􄑴O$?ƮE"f;&M[(1yy[$g{*}&%9dߗ'" MG=a|kO@w'X9HGG"v'r|$Nxb' 0{/;Vpٓ'\uxeCOIVVQE"fMk ϞfkܜbЪ5ح}y/k58? eh? 4VncD{kqQo9>C ͩs ٚca_S u 4VNRcZY/bȰ s+_Txyh͓/b'lUr ٚc쭆}y/jp~x (@czޥZ'.yv<}yrr 7x=5'|=4'txY5a #iMb[bb؄ z 엘gs5ONxJtOئ5 _('H> ɵ' JPOr'wkŒ߻_y"ǻ{!TL<c'" ' .=X'X9vHOk8:iQȰ syRbr'c< ;٘bC%dkw *Ho1Vymya÷l@_Tj'ߋEO858O *Hd z<}Yj<}COIvE;,j 12llB1s<Z$TLv%9n7T~_QlOH>'Pj~N+}R0^ K~B'r|)Nx.LtOqh1vn=AiJ~+ }#h {YvGYEZM{k٘g5ONx֦b'Qr ٚc=}y/cp~ ;4x{'z,>?uZG"Z'r}>?5a_-AO8<>x9COIRI.$ZC9 0Boo9N X$TLVњcu*'" ^lOOY=q~G+}~U~JLjg2O@W]xB>aKcz/OPE~&׌b{–' .ZOr'ׄ1.ZC̙ 0Aoo9 fZ$TLw5iT OPE~ |u3=o(@cO ݗ)|2y.y2 q OH'Qr ٚclSO *HQlO' yA+}QyM(͵5Ĝ° ((<`5ONxNItOqg1ZsN_w`WlO"7מx>tPjm|<b>3<9}\%96$_4Lxbh 5G4X\Og5= *Hï=3Oi<3CO}.~(qگC ΍”ܒ5z^!8 sXJϹk$%'\Lڑ/I b|oψB>*qq/]а+y/zmM➕]U+x9su]>zڈ:(1So%߸.&c'.^71$S=|fa|!86o:^ByE78OPϹ6]pԙsem)\s|/:^ p׊?R qy|Y~d[7ZǵBk_+E)z&^*Qar~J\SCyxʹ^Tsr~Cή_o3=>5}m,mζy2YR_K=U]CA}^/oΒ{c$Gߦ::DYg^?H>aZ/^C+Vm-֌жCB;Z7^hrH_0wC0= ӷ6 ӗ:>OU0}N0]tL0]tӁ?4Xd.ps L0]t ɤ[fvHfF;cDd 0 # # A#!"piM$ mL;@=DiM$ m1dQ: =x]~p (Z""6Rm\0g=[XAzQhiI8FCqҊ4ҭk,mi@5~0 u/y]GQ/ )/-ďnk/1菾!ň4=0(A䦔|ϖZ1]ԉH|G<)+b0b b(A ?L@T\.B;2/-^'b{=^}k!Z BdXUUZ7!:]Dq,ET & 3 &ElBt"YDqKރH!=.w"}[ބDz' C)] ] ] ] ] ] ] M9li֭ެmR닚ހۑ~\&?D4)>8MUSO~W^X,u3G/ZɭcCOy5D/>Tyݲf+Ř%,lהKp;\f6ٽrW03Q3CӸPҿX澞JG!y}MEIzi|A^+F.u w!kϜeaҮjr2hrj+F|w]Mh&-04ybi&eGTX⟇V/4ZrBs9ZNq>:2yuN'!^qʜ)Z<曺|>h4 grv!.a<- 'W9-w81 OZxaN|Դu* vM-s[ffA~fc -4oܦv򂰞 ׌+~)S^}6C^y)<͆6k6[зPfMf ~ے4ܗ Зp&Ba+:ԻPNMN zwݩiA,5zOwݚ6Ի[ӻۂ,Oj7ֈz1 z1 zkQwezR"}kW&1ӛU = 4Go_w޵޵5ԻVӻւ,=MjB=9+ ^҂+ ^҂zg|΄z4{N z4{N z?kQg/lHD;gwN;gA9M_Rӷ}~=wDg˴ YDＦwނY{"M;_osW>oVw,?;u,b0Ѱ9z|.P.f6`f"vQf"cƓ8v2< ļ1+WVRjNc)>!B|>1ǦL \F<bI!܅HxR}ᾋ˙-ϟxBF~2õ)b3fQ1'q,sdl@L[%ԜSRs}=abQ @E>y'{DuG9N1hGb<x19=%O 7I5*،9Q1_VI8Ycbun5ꔄ'\~C{Ҟ0CO'p_)yB(Fl5X͘hyTb<xj̓=bFm.)4vcï1'ֆ=nw_'p_)yB?l aW&k6/)1OhgEq]a+k~6 Q9ɘFcl Ĵ m\RJi춐 :?ׇ_[#iOk[e'p_)yBS\D_k9WfvF$~s5q3ט'c{<buF5ꔄ'\~!8iOsNxT['lW*b+* a*،Q1_Ɠ8V:<Yiq>9T$#O ii} Ruс aNVQ1d'q|ڙl̓=1mx:nAuJt~' =N {~n WJpP?XY5``l<֨<`˜'c{<+1mx:x5ꔄ'\~C{Ҟ@ow_R ~ 37zk?D| U^9Kͫ"D$@KHqvz+Q&|ʭSu&[׋%.ߺ>7YÅVzyQLs~H19?t sÜ4Yr >su s59~$q]?S|l:1F9)>}G,)yKxyy_%IO+GVn*,赽~N?տQR'BSal:F:|bV!*XQ_+R^+Ń U<%W90ׁ} lic}F5]X UR͡xL/*~3B'߅ɿ^US#F~n ` ƫՀ5kE\``z.]A0L`z/L?ӻa~[a:k`<^ ӫp .` Z 0]tL0=f@Ew\EQ|S Uulퟃ3PGFCz>C_s*2*KO E*^wp6m} \wGŵkW]/oM}n5mOF^U3zq~K*Ms?9)`!1 &s%KNsǙY 39ԹBJF?To;οOχ2i8M徂u\ 3ҊRjA򚴁U7&&'+}դ z]AC`w5Vd:VGup Ѷ+iZ\ZPZȲ%GՑ0bMVZs xasXswV>>'UMoIzSb={C={CXTM+zgY{Klǽ?.-y _ kz}_wwVzwwVz~cP5]e^Pbw8{Vñ;zPnQ5-|>X#Zxz}CXUMc^Xb={C={C_XWTMPW,뽨zxvfԻջ{F58MyZmet˸{bv91?gmoKu?7ަ f,bO=ʖ1_9glļ=1۳+X"sҮ a}M3 1{S \xo U+{9p+U_Ʈv/gzdV Ζj'& L}~~N+7M,Gҭצ"6aR1t'r|YNwI)陘4< _lgyk'z {^`i&`]k,bf-]yyv3 <}Y _lgykghJXe{UYfTտR8jt5aR8[*f<c.e ´a&MX稜Ci/6[jq} w4qU_ǚ&4aiսGMѿ= IxLC2 Z,y5hR(M~PiB?5hB&L}2U.ˡMl] Jx~̓Sdb',bX2K4A->քVM@v 4VN; տ'r\/uSdb'p s5Κ &&|H>Q}1&L} UUl\G1f؄ΖJaro;$ۙ> +X"s(M~+,5h^P4i@u?Mѿ&y 3& >4!cJ%b9?p Jt~ߪ O҄TAԄɴOhRf}Gg6a6R1&9<̓Sg31}X"sd?&b&#oM[Mh T&9ך96Iw!LL4rYt~Ԅ2& .r&X9Vwr5u9/b1l\ g]^xv%aKrW[ :o&l{3& k` 4Vnѽ_9.uS};X"s5Κ &&̕a4qc5iF 5pơaΖ)&)g;OO~}g)M~k,=oM {R 4VN{A뵂&߂< Ow)pMM'bXѷ&_lgf GeM@\r&X9՚A68&f8[*ynHx̓S/eb'4kD,(M~GޚةQWMh T:9^x":3|)>%M :oԄ7i›4MKMLUll6aR1O(LD3ONyOhbXwP Z,5@eMh T!׆4!D)Ls(3CLM'bX·ABi/6;15?&?kЄ~KMlVJ0&8[*&;D3ONyOD,;+Oy8fϗ&7ߚ/}˨/>ޯZ&~y' /i}&b !4qq5iЦ꿘E-v5|af-=D< ;t&}۩(;+Oyh/6k5"4!cX%b9G߶Jt~͚&%M :Ԅˁ4A.[jd'<꿏3pϡIMR1D< ;d&}SQwV_l Seh5q7٥i>ޣZ&y$C%b9GCi/6밸>քSoM@; Y+c<6ѽ_׬D ]g;ٜy,9gM~P~Ižc[+}Biptڡs%&N8[*&[DJg\c619S :-}<޳X,=޳=K>^z !09B]g Lc&dNkD,VBi/6Y3քs4w pR&>ZE/QÚ 0R1&ٝwI)LLZ,9>h?&bGޚ-@TU&X9oѽ_9!uS> /i}w&b eM@\c 4VNCl<G 0YWx&KrY :&|WoM@5MrHy9;!j o< :ﰅ wh ?_4CzZABoAar& ǂM}Krm5)}/6Y3CM8&ԄɴOQk)=ptʡs.&8[*&D\g򜛉c.Pcs<4AXք_ߚe_ 4VNSU߂< Ow)0M@O8o[M :f[ o`Є,5a2^UkBWʡ^Uq6a gKD~OX%yvʳ:>(;+Oyh/6{bϗ&.qMr'PݏjvABQyMxQ M'TiD,VBi/6Y3քeM@\r&И>Ml?ؚ.쯀c(#4GOMq+9>c{,r. >yOE:$@ 6"zL<*J:_b41tϼK+6'#uuSh%S&9C\c0r#o <9M(fے_oz| Ο9yi;~1̘G{PVCmO_sOi ^=.ĉ~ GkT5w:wc5*uVQ *DVIU*PF#O~az}^lz],+yv=Zg|DQ1o05|wQżs5|O}ߪpu/T:/4@q߮b^hk{Q,g]CJq~*bGߦz{E֬ۋSE+IogUUk[֊` ]4`%~^~w0u0y0WtL0]tL0Aa)L0]tӓ}/zU*vQcDd 0 & # A&#"KUa3>͖'^';@=KUa3>͖'^1dQ: =x]pUUz?KtQI-㠭3ev-,K]&tXc$𧅒qI Zqwg @BtNwtq {/s^^^7}~wO}y/Bl)/ Q;5 o!$h m >1V&Bxr_D51CZ47@Z:([X  Z1mN EI_6BķBtu b͘!-wNh zvZ%)f@Z#BAh]zv1`E~.~wm07:>πS`x%  .` .`8 & .`9_l19ߤ#R{r|CEƛ|0I>~`_mcعpćODwpl8ʖ5pݳ1cO*.k{wn)=Fqg~}g/9&`Ί/́aRY.<~M ]gn/l(׹@[mPeR+k9p.w-g|H?Eֲ֞r؄bf%&癍y>m͓ObKcg /3B{^Gzރ' S/Ux.!s18H - [$ˮS< ]=q?g WJRkg,yrR \B6细IxKl@Wj'<qNhWr ٜ'_^'OxCA'a #iHs\ta"kS I TLEJ.!s0/OPE~ B{jh4?'P_PZ;|PTt`3NtNไl1v0{/k/BO? {Z WJR\NX/bv1l\ {1̞4N'ź+ R\Y[K-1b؄Y{\*19ϥ1ϷyvTLz%9Ʈ7T~_QhOx?x=$ WJҽO> OQ}yLR K\B6}C$_4'^u]'n #iР9W6\4< grlI . J.!s`ߗ'" О 7ЖC; {Rk_*19֘ϴ2O@.<:KusE~K ^<q3{ྣx:t _qz`3lKtNqh1a_w='O,>ˆ0 R^vl^EM 0a:³7ɚ'a'<7bX'4+l1va|_4 B{~h;C M^Oไl1v0{/-BOFV+ufVv [[-13l {*19ϭ1< ;>:aKco6T~fc8[@#j]H}JObrb>yrR [\B6I^I TLJ.!s0/OPE~ B{V׵' C{J>g Rk'-(2􄑴N+b/lM5Ĝɰ s/Tx" Lk𜙊b>s6O_ .ҵ' t|*@}N Z; 19ρ 0O@.6y6gۼz>4M )/{~Oy)!0>-8oFwTkKĿ]pΚ1"Em=9BJhS̀6ZFą Z'к@;b 0 ]0 `x ou0| çJ^ 0\p 0\p & \p 02H2鑙b-rKqsDd 0 ' # A'$"rIykyR"A_@r;@=rIykyR"A_@1dQ: =rxxUHK PZ(.x@H .-g]=^.cwNZ&a&tbW$oZ,Z,Y+{8[,N5SZuXRZc}$dWR%zowOk@ ϓZ<,-w+XZ--QƖۯJk}>[_m4povooK_m\9kK]K׎zim?:ܶ6TI>͍Ԗ>Y|?vIkgF۝o|ZR[Yd޳k›XRS_qpj}d| >K{Y?mǷX} U}e*֓~Gzz~{Sw?Ngtok{'sr&֖w$֙M)ߟ39):nW8^߶?.dI2o㷂᏿[,[z*[_㭯o[_筯֏k#o}W\[_㭯zku6[oo[oo};Xo^zr[oj=}햷Se----|>nM;ج UfW8{&y>*`S5c;?v7+ٿX?h}屾\m?^[I3?jg_;%9_bcl;N}so9bS4%^|+۟ 7|W4$~El7j۝2s?}.>jJݦfN\}0Mfs7ٯvvtr?l_ӿ CYv_,S~$lj#_?o?_﫥}[ö3Zr0;:eomyO7<8& 8HWv|ߖ?o)`=Ox8Ox8(<3]nwynwyZy<_y?#3y?#3y ۙ&u}ϻ?| Z"7o."7o."!g<e WI3fBſپ#!_IS_/7֟h?I4*~oߋw~N'K>=;߿o+= -&&Op8U%SUSN 8ҩ8 N^%pr#MԄtjNMNHԜlr![N:ȩ1 N`3N>#p q| )tRt N'Oɨ.t8Sv:vW'?dNN:ɛt'oS#ҩ85850NA:Sɇt'SCҩ!858El$)PK:)tXS}t^S8 H'?p9ŐN1#pgU)XO:5 SSv:vW SdNM:ES)d;S)t2bcgp5&5Ӕ83#LN)t `S=ҩ88lW jS SN!"() NuMN5)t PS$ N0*AS-p%pF:Uj")"NuLӱJmp-pN:ӅN.pmUSS ɨ.T jlcw0p 8$H0p 84N*udNWtNuN5LӱJ88 jNFtIT[Td;DSdNWtNNLӱJ$8E H'vSMp)p 5N*Q%pK: jS Sv:vWNNᤓQ;]uҩ:8U8lcwT_A:U#S5Sv:vWhS$dNJ:S)d;4(ɨ.t Sv:vWS=ɨ.t `v:vWXS}ɨ.t  v:vWiN NѤQ;] N'_t NFtS8|LӱJ8 bH'v:N&]%NQ;N~'/t NFt/8 #p0N* p#I'op8lcwTO:yN^%pr3N*1dNI:yv:vWiNNMH'v:<N&]88%NFt;8 \LӱJKpj)pjJ:ӹNn&pbUZS+S3ɨ*T MӱJkpj-pjN:ӹN*plUڀSS ɨ΅tr'S%t ZNFtUH*TETd;vVQ;3 N &v~4%MT"*S%S9=;SSɨ]"T* ʒNU3#8u8%U*N NdԮRt*NNI]38u8'ʑNv:vWN]NH'vSYp*+p*iUSWSGɨ TJlcwnMԉt2j+M:&];8u8u&JN wSIp*)p*J:'G*= dNWt*N%NEH2]'88u%銓N0T^nQ;]1ҩ88"Jw[ԝt2j+J:_MӱJp#pA:!SSAt zNFtITXTd;~^Q;]!ҩ887N*7dN++8*pd; >Q;]Aҩ 883NMk)qk]%? N3x8 8#U~!~_NI_3x08 8'UN))T=ɨ]%/ ~& w4T4t2jC:<&]e8 8 "rN)'t2 NFtH\Kv:vWN#NCH'vIgSt2F NFt9ISd;;G7ٯ*?N?ӏlS.pr  NNI'vSp!pt=ǀɨ]%; N93x,88$U~ ~NߑN?wq4N4t2jF:eloMӱxp/pM:}O:}N 1N*ii dNt NYN_lcw4Q4t2jt8}eU&$8ɨ[[pVv:vOoo_3ME*_N_̤SVpr NSNH'vH+w< &NF*_N_ӗLӷ< &NF*YH,E9 8}!>ONiH ]e88">&>NIѻ|p/pM:ӥ#ҁS:Gӧ]e8-8!ҒNi)) ҐNi))% 89zWYN NsI'vԤSjpJ-pJA:} NU"<ɨ##pHtJNUb|ɨ. R Ni)g&5Ӕ[$5ᮒtJN)NIH NF*I\D:}NiidԮtJNN)89z/ŤQJR))8%8pN))) ~'RwBഄt2js"IQ;tEtJNUV*2ɨl;&wZജt2jt~8ot ~8$ѻpZ#pZA:ӽ!ހ NUւZJɨ5^ N&Uց:*ɨ ^ Nw^ചt2j{I:&]e8m8!^N/v:vWNNkI'vspz.pzlU6&:ɨ lcwYഞt2j{J:=&]e 8m8m 鞐NOv:vW N[NI'vǤcpz,poU6&ɨ lcw]ങt2j{H:=&]e88m!N w}p/pM:='G*;ii+dNwtNNHG]e88m#Nw&Q;88 w=Gഃt2jM:&nNtQ;M&88]%w}Oഋt2jA:+pr vNFtIt]tt NUɨt .lcwtPഗt2jJ:]&鮐NWtC}Q;e28]8'wtXടt2jD:]Ks5prrNFtItQtt NUQAɨt.ΘlcwctLtt2j;O:&]888&ΑNv:vW9N'NGH'vYp:+p:iUNIQɨ tNlcwStJtt2j;M:&|v~4}nfUNN'(t=π ɨ]tNNg3,88$UN0t=ρ9)ɨ]t Nѻyp:/p:M:%QAtr.ΐNFtGH#tDtd;\Q;a0887N*dNwt:NNL$A^(89zW NNI'vp: pC:'G*WdNtNNI]*8]8]$Niit N{ii't\kKQ;i88 wt]tt2jM:~prrnNFtH]K഍tNUnMUɨn'v lcw[tKtt2jA:-&鶓Nii3뤓Q;6i8m8m"vw;tGtt2jJ:mNprr nNFt[H-EഁtNU=-ɨn36 ֛lcwt_tt2jD:mMu&v~4%kMl 6i 89z?Qzi=88&6g#pz$pG:#ց:*i89z?QZi-88$6w'Dt2j[C:5&])8=8=$VNii wU*pZ%pZF:'G*dNtZ N+NKI]98=8=&VN+ii 'Q;ri98-8-&VwRt2j[F:-eE&]88=#閒NKiiv:vWy NNI'v%pZ"pZ`Uހ ɨn1 lcw7Kɨn lcww+ɨn! lcwVt2j[@:-9&]Nk dN7tNNMӱ3Mk)qe]-IdfgL`v$Q3x4fldNoo_3M` g,i88M#恓gKrpJ.pr"lI&5Ӕ83l̔Noo_3M`f LedfqM6[H#Sr)98%9M#4$͖$SjpJ-pJA:"Tp*phtJNiN)IR4tNSNL$#҂SZS*)89M&&dxt쮒 >"&Niiv:vW>8&&Niiv:vW>8!&Niiv:vWINNiI'vxp/pmU>ONH'vq8p'peU>NNFtcI4V4d;d OH'v1p#paU2SFSzɨn44F lcwsӧQ;(i88 3N*)3ɨn$4F lcw/ Sɨn4FlcwYt2jN: &]% 8e8}N: #0 t%8})pD: %P@t8}%pt2jB: !&]kpZt2jL: &]pFt2jD: A~&][pV%dN7tNN}Mӱw+ɨn4lcwU5dNןtNNMӱɨ zlcwlM-dNlۛLSi]k);8e8e%SVSҩ880ٮ r'ENI[dʷӏ) zNvH'ȩ'z lcwSt2jA:.&]gpYt2jN:u&]%88H:ӱ38~{iJLd'SSN)'89u%SWSG*9HW338,rB:u.&U~$S>S.)89u&:SgS{*?N/ܤSnp-rD:uNv&]%?88!:Nv:vW)NNyI'vSp pjcU SAS>Ɉ7ٯędJҩ88'S~S;ҩ882ٮt* NNHT@Ԗtj NmN-M#SSAҩ 89!ڀSS *NEWWpUt*NNHBTHԚtj NNM SqSaҩ088$JS Sҩ89"ZS+S3ҩ8Ʃ$88%SQSKҩ%885%ڂS[laJS)S1ҩ88&JSiSqҩ889 ZS Sf TJN%ȩ9 l+J:SIp*)p*F:rRS)p*%rjF:5fƤS+pj%-Syp*/p*M:"SSpj*p7NWtNNeH2TFTtNNeITV@:%S)tjN-qNNHrTNԄtjNMNLӕ&*SeSyҩ<88!YTtNDNIXԐtjNqNUNITQTtr'S%ҩ8U9œN/p%SSla\UTt NENqS8 bHpJƩ*8U89N,pH: TNf N'\DN IPm2N'W\NΤ'8y NU)tXS})q'/Nn")bNLӹN-pr']J:Ƀt'Sҩ858EN`0N+p$p'pM:~p/pC:#p: pK:Ap:(pG:cC!p:$r*B:"S p*!xf2N0t N'tN'(t Nt N8t NgtNNg$tN NT 9p:'p:E:Syp:/p:M:p p:C:3Ep(p:K:S%p$p:G:sӤep,p:O:3p"p@:] Up*pH:]"BS!p*$pO:bg&t .N N^NO!^ NO ^ N1^ N ^N/ȩT N8N ^N/;8.pzE:W[pz+pzM:*Wf}5w~4MUɅlr&wA;M."ƤS3pj&pjL:5'ɕl&A M"xҩ)858œNMmJ6yBɓl MUENqS8%HpJ` M^d4NMf0 Mނ&oEN IXԐtj N3i&AM&O|Md'4ybIxp8ŒN/-L4 &/h9ŐNq'p!)N0[&h4M-rj@:5FS#pj$LS4&h9EN )tjN 3i @AS M"S,8 N+LS4 &?h9#b)FTt f)XL6C)qDM!no_3M!dS4ɦPh 4M(r$)ZI:ES4UjjdS4"HT_A:4UꂦdS04IzTON:z4ՀdS4NQ%pK:ES`3M5& M"0)"NaS$8E f0T j jMՠȩN:S8Ef0Tj jMաȩ6NڤS88 f0TM5ȩT jNu`3Ma&h #jBSMSM) N5I0p `.44%jAS-S ҩ88 S f¡)\N6Ն"Smp-pN:ڂ4E@S)lMuDNHZTKTtN3iHAS$Ma"Pҩ&88N5`3MQ%h"BS]STjBHTC0zTOTl pS0T IT]0T_TlST HjTM0hh4EM)r $B)TH:S`3M MQ")BNS8f0M149N,p')X0Xh4ŒMɏt  NA45dS44E|I@p 8N(LS#hj$hjD65"')N>S8f0Mq8)bDNޤ?8 I'p`)MdS,4ŊH'?p8yN~'LSchj,hjL65"'O|N/8 f0MɃt'N>4%@S)l8; N'w3ij MMMMɦxh9N^%pr#K0fLԌlj MENUI'Op8U%>dSh!r*H:"Sp*"LS_h+hK6"Sap*,p*@:‚4~~dS/h%rO:BS!p*$MmdfM)}# SAS>ҩ 8`i4 4 @SS^ҩ88% S fB@A@/49!S~S)?8`i4 4 "AS?SnpEt~`i04 4 &CSS.)88"S> f@Ai4 9L:IW04T4lMEN9I8#>O 4M3i)- ҒN)`WaB\A\i&49!>ONiHO4̓yyd,h%rJM:} N RNHtN vi>44'fClS*)-88"҂SZ4-dh#rJI:4SpJ#UдPдl MsEN)HZtJ N389Ѵ6Mki4 G2)88%#RS* fCbAbi>49%%RSJSR)%8`i 4-4-!@S)88%!RS fBRARi!4-99N)ɉtJNN)8%8YHdL0MˠiiٴZ8T4-EdrhZ.hZN6-"I$D;vi44 @o89 ~#I0дRдlZ MKENoH' 8YNoH' 8YNI[ tziiٴ V VMˠi;8.pzE:Nlri544&CrK7pMt ئdhZ#hZC6"pz#pzA:74dJhZ)rzN:kpz-L:hZ'hZG6U"g+pz%pzF:W4djhZ-rzJ:Kpz)Lh h@65"' pz!pzB:4mdZhZ+rzL:=Ǥspz.L&h$hD6u"G3pz&pzD:=g4m͂dzhZ/rzH:=Spz*Lh"hB6m "pz"pz@:='4mdFh(rO:=cpz,L6h&hF6mM"{#pz$pG:=GCpz(pK:=]iMMɦдYttzNNwH@0M;iiٴnN6t vi'44$BV-88"= fvA.A.i4m9$]M.88  8*Lnh-hM6m"mp-pN:ۂ]iM{M{ȦдCttNNH[tK0M{ii/ٴvN7*tn f0Ӵ Mi tnN7`3Mii?ٴv.N2t .Nt v44 @E*8]8]$U4d^h+r@:]+ p"UCtHtlMDNItYtt N3i: MMɦд_ttNNHKtI0#tDtl:MDNgItQttN3i: MGMGɦtPttNNgH tA0ctLtl:MDNIt^tt:N3i:MMɦtXtt:NNHstN0tBtl:MGDN'ItVtt: Ng3t6Mk$t ڞip:-p:N:ӂ4SSd1h:&r:F:Sc)p:%Lih:-h:M6"Ip:)p:J:433d h:!r:B:# p:!LYh:+h:K6"äqp:.p:L:4ssd)h:%r:D:cC1p:&Lyh:/h:O6"Qp:*p:H:4] dh:#r:@:#p:"pO:ap:,UtQtl: MgENHCtHഏt:N t . .Mi/t N`Wa.CeAe<49!88&~ni?8*Lh"hB6] "]>p'pE:}]i MWMWɦtQ䴓t N{N;IW0Mנ隠t .vN{iif0t Mi;v Ni`3M7醠tNii v N;ii+v v&44$BUi88m!v4݂[[d5h&rL:mͤvp.Ut[tlMENHmMഉtN3iMwMwȦtC䴑t N[NIU0tWtl M7ENH-EഁtN[3x=t6Mkt n :i8m8#6& fC}A}649%6FZi#8m`4=4= @i8m8!6 fBCAC.49&փzji=8`4=4="A=*i88"ց: fCcAc>49$ւZJi-88 ր i 8*Lhz"hzB6="jpZ-pZN:Ղ]iz MOMOɦP䴌tZNNHUJ0MϠ陠N+ii)V f0 Mi VN+i`3M/酠Nii1 f0^ ^MOi Ni`3M镠 NKii! f0^ ^MϡiNKi`3Mo鍠 ^Nii> f04&hlz M/ENHEH4tZN3i~4N6W"BpZ(pK:-fe.̖p#pA:94%[pz+pJ-Ӊdg~4%#,d4%>3Mm\{iJN69A)Y0hJa{ڛLS ) 4%4%%RBSJASJ))4%9M!f i8<3TДJДlJMDNI4]4tNi>4}D6%"I4p&pD:Mi4ԂdS hJ!rH:MTp*LShJ#hJC6" p"p@:M)4dS*hJ%rO:Mdp,LS:hJ'hJG6}MƑNii4 & f014},hlJ MENcI4Q4tN3i>4}B64"1p pC:M 4dSZhJ+rM:Ѥxp/LӧS)49"Ɓ8(i8`3hL14},rI:Xp+LSh h@6}MFNcii4f0Ӕ2 2M)i84F Ni`3MC炦ɦOS0i88 #F( f2AS&AS&3hL4t N#NCI4R0/ AdSh rB:!p!LSfh,hL6e"pp.pL: 4e,,di4 Ni`3M_Bӗ/ɦLДI4t NCNI4T0+AWdi4NCi`3M_CׂɦДYԟt NNI4X0oA7dSh"rG: A~ p$Lӷ[KhRԗtNN}I4P0;AwdWȩ4Ni`3MY))+54}-hJ-m3df'oMߒM٠))-4}+rE:~^S?p'xf2M?@Ȧ;SOҩ/88$S_laCSvASv)+4e9 SSҩ8`)44 ENI[ԝt N3i~4H6el"nS/p%pF:^4M? ~"~DN]ISԕt N=3i M9M9ɦД]Д8[M?f~4L6倦ɦ\ДKДl~9u&SwSgҩ;8u<3Д[Дl ~9u"S7S'ҩ8u)44!rBSNSGҩ+8u8u$SW fBS^AS^ghYԁtN]NH.E0|ДOДlMDNIYԞt N3i~4B6冦"vS'p$pjG:uN4処dSh#hJ-mno_3MȦДWДl*MMɦ|ДOԆtNNmHAd~_MM@/"֤S{pj/pjM:4BBdS~h/rjE:vVS;pj'LSah*,h*L6dSh*"h*B6"S[pj+pjI:4dӯT MȩZNm`0Mš8T N9Z f TJJME(TJ JMEȩ Z N`0MT N-)Z f TJ JMšTʐM%)tjN-N S pj!-LSYh*+h*K6"&Sspj.pjB:54rrdS)h*%h*M6dSih*-rjL:5fƤS3pj&-LSh h@62"xҩ)858œNM`0M"T ʑMTʉHpJ8őN -LSeh,hL6 d349 ɦ TAԈtjNMNH&D0[*TETlMEN IXԐtj Nir&A T * *M*hr%*CSeS,NX)i MUMU&ghr9ŐNq'p!)N0[&7hr4MUɅlr&wA;M."S#pj$pj@:5F4y@Ƀlr&WS4 IP0[&Oh4yMUɍl&/AMn"S,8 N+-L74y &whr9#b)FTtla|GC6y@ɓl&_A/ M"(ҩ858EN `0M~'h#KI:ES)thla_O6yCɇl ASM>"ҩ>88EN`0M(h $|WN:zpҩ8)MAd4 ɦ`h 4M/rK:ES.NQ4@S)l AS  MP)ENaS$8E HHp4U4U#)HTtS)"iMMɦ`h 4M5M!"ڤS88 jN.-LSMh)hI6BS)YheuMkT M6TjNa&pI:S4Ձ::dS h!hI6AS)l M5EN5H:TGTtNui MuMuɦZTKTt NNIT[0[ph 4MM49U#jS-S5ҩ8)"MdS4BITSJ:4EAS)l MuMdS=h'hG6CS)tN5N!S p!-LS}h/hO6E@S)tNNSup.-LS44E ɦHh4EM MQ" ҩ8U8N`0M1#h!AS=S  N@)BiXAS,T ɦPԐlhSN!)BijMMȦ@O:Sɟt `la)NG6@S)lxAS< M"'?)N~S8 f  M ɗt @/N45&&dS#hj$h#)AД@6ASɇt N45dS<4 M͠I'p8yN/-LSshj.hjN65&44 )AE:ɋt'?laZBSKASK)459yN+p$|W0[VJԊljMMɦZԚljMEN8GԇlM=M=ɦWԗl M=ENH'pr8U"\E0[~OԏlMMɦ_ԟl MM}Ȧ4@4lM}DNI*TETtNULi 4 4 $BS_AS?i4 4 "AS?ASi04 4 &CSSNH'gpr<3!4D4lMMɦ4T4lMMȦa4L4lMDNITYTt NLi84 4 'C`Ai44 @APi$44$BP< FA(A(>3ˑNT * Lhh-hM6 Ⴆdh#hC6dXh+hK6&xqq&}f%*SESYҩ"8U<34^4l MMcȦ 4A4lMcMcɦ4Q4l McEς2D$@Mki4 ƓMii24 &MSii 4&&MSii*4&JN4T L4h&hF6MIdth.hN6Mɂ)d h!hA6M)&x̄&}f"ʁS9S)ҩ8<3Y4K4lMMɦ4[4lMM3Ȧ94G4lM3L, MsMsM,I:SYp*+xf2Mii4 f fMii>4f M ii4Y Yt*NeN%H2TFdA"A"i44'CbAbi>44- @Ai4-0`)4-4-53},ee&}f'JSiSqҩ48<3д\дlZ MMKȦдBдlZMKMKɦдRдlZ MKM,XMMLdiiIHRTJTt*NLi 44!V@ AJi-44%VBJA*i44#VA*< CzAz>3gh h`g&,MMM,J:SIp*)xf2Miiٴ ֓Mii3ٴ 6M[ii ٴ6Y > A6A6>3N%TJLvh.hN6m͂-dh!hA6m-dNh)hI6m&x삦]]&|fςд[дۄLYYt*NNIT\dB^A^i'44"A>A>i44&C~A~i7464403}&|fςCtHtȄLYpO{]UoAG&Έ3Dl (z B $.E^T HQF]D!@(I PR)^&s1 ܇zHn{{s퀦hLS/8i:䀦C~TY[:U֩i鰦j:i:h8鈡h6tTtMG 5Ek= b4M1h@4XMSb=3M N瀦8LS/8i:怦c꙾TI[J:U֩i鸦j48)PS)Mb5Mh3tBtM' 5i< 4M hJ@4DMS=3M IӔ䀦$LS/H4%;)= NjN:餇zfuSE֩:U֩i锦NjJ4%:)PiMi46ԔiJr@S33h:c)YӔ^iJq@SzjRДꁞii44ygzA)M虖h٭[R4M)hJ5Ԕip@STMS 54s@9CMi44j4e:)PS) KӔ倦,LS/4e;)= r4M9h@4h:cN~:9N>䧭i)WӔ뀦\CMY,4eji 5Mh1tQtM 5hr< .i.9z\4]v@eLS/i…+虦^pUtMW=3Myg7\'_m|X䫭i麦ji쀦+nhn8醡++hj馦njiZMuqioP61kY{_4x1%'4@x1 8|]}Um Wu$LՕ>V%T5^_zJwHU5T@鬚I'\U ZIhZK;FڪF֪R 4W$Du*DPuC@}S^RWPK*LR}TW 5@U@y] U5D^Q%K%5B*|(ƈWc'/ջ75^W$򬊐gTy4! O?jA͑2;5WT'jJ-Gj< bZ"%!TJeR ZNJ) T* Wd\/"\/%WZɂLY'dd9+HlTHrF"lSM!IK)'dn9qjc߃z.6ǿX%FrX!h9g̿ek=j=zFo9.ζ`+J(eQINFo_,25ƌәYV 8'3UdZ9yV/WVJYerI>f喲KXɏ,fU YtV~>;0@%{\vix,vn&;8CQiTyLMa#$O &o'A*e,U3C?Ro*i5RL>RzsTsYxDb?OʋPASaTznHT*n%/+tիtkPEuk=袮R ޠC M-pF }@kz&\KӥR t|!NGh MpS41C hS1qf8I3%g \#@3nBpdTC* ߵ, 7 p/o ᪖)`4-y0^)`"LɪD Tu3`&RolsU5yއ`̇>TUe,`z]ǰLUV'5>UZ*%|^a ueYب^M"Ue[alWd]`qZ/p/j=HT`|?^Q(?9p+G!b! r$@"$%N)8 g RA: 8٪yȅ p.eʑ帎/z٭ NNQ}n{W4Ac ftLshAZAk: Ֆkt]tNЙBw@W:zн=^ts(aNaN0G3 !Pb 9F #qa2wcqwpq8ϻ0MI8d*'Sqitflo.8?y|r>Aᤋ#\u q7W VЫq \K+\kXz\~n 6IHB*l%aE| Iٽ?*wv?#uP5$^iDNIpĊ&`ivTX.;F')@&ILB$)O BiN%a3iIdQِC'sI pDZ_+$Un]Ug|M6ޭu^s[lMk۾\8?MzGƟ`j[cMgՓj<+e %yxn G ^{un{}^ aSwY_{Lv͞s[~9}nLf9[S뺩^Ӻߣ}Oeek~Xׁ➛,Y@B4=T&^όr^-q_z/ޚwݷzXsim{ jjm} MNqkr흧Ϭ|=P)ϵugീr-0%\=ɿr ɿe]= _qًg>(^w-e?ǫ <ӫcWM!'ߋw|NƾwN ϻ{.>ܷM~Z |_&mMA𼁦{nM&:j 4AiOzNVwػ/ǵݒgf!m=]{4?~`X@\O>8RR1MUX;NwQj:tmRӵPMjZtݪj:x.OY?_HɹZgzz?24xTh1!*(H< bxãU _`x(1ܟUpo Åpa0\. ;@=9xyJRX1dQ: =cx\g F1gK 1&1&Ēd-`EEE:\X( {Ŭ v P5;Vh27LP}{y99ϙr+ !F!=]_m#DR!l[ !.BZ Z]6?RZ3X {pL ;'|"}o@9*xVPAVR|@SP*%QU _HUb[QUhW!rr3\Ml dF΀ASz 9&H |pq/ {:ɸG{opo{mmqCv w].䚂].p ܍(+e*UG+6 E1U@S=Se'gKOt,uKd;QX,ә{:Z3a1̜p/1/rtꍻ|z+c_. mhl1~UfxŸW? tV9=rW68HΏzaE6VRʑ/_VfBf5hr*iTM"5Aɿ&\'zib3yghVb[p$c{d/4T6Tr544ot4(GwUjNeK$ūYf5ȌC;nSkzAWVySl-o{Kjop9 q2_O1q2iGkW+`u'pYuZ"9}LUXQy~*+XyiM6L6\{HpP6QP:QP:뭓J,YØ5٬ ˟66ꝮQt:蝮Qt:蝪ީJ;BR\zA\zAlV4=Zy).ԨwJB.ԨwJBޗM}B(4>6Mo\oZ J~z˱dMk"Ro{z۫Ao{z۫A&:DԵXwRQp:Qp:q%o JwF3TzgwF3Tzgh'XwRꝯQ|:蝯Q|:YG7+,Ae)`k:~E*U6(dM_+{M4yڋ߾mZC(T6^k ^ScCT3`F}T:蠯F}TjtA_%VkE߮*}M2/зٵm55j\SqM4Q*kq-5h\Y/FU[렱FU[:jDk\k\%5>-i|ZVBZ8JdZkOÊ5>l _R,g4A,g4ʩ Vb-U4QilYY;ΗJ:Mi*t:Mi*tzZ'*ZO.AD[* 66蠵AQkOE%hi򳰼Nv*tNv*tкZ7V/Ai-?;eըJk[ըJk[PG(Z\U,ҺaZ_{O '¤߯nw^F{ߝgnumEj#z- oܯ$xpEhh<`p1 E:ځg]|7q`_'ǎ\s4FY@$0X@< gL^Bo{-ow @_{_gG9O?` g@ H s%ȟ'Oua tD<@쇁q-> qXm bCcKbdуa9 pt$y"H08k `TR+uH=H"K0H;%`2=哨?G zD1xj8zHv&sƂX)3@ " ;@ ]ID <&g">Z FIĚDI8Mq2'3?B)h7ߩT,S9?8ӈ>vt̀ r`O:Lr4LQ E|KpWLi ċY}g;È flg s?Kҋq.8=i>Os9pM- XH ٿP}fce,Y^˹tq-YG㻞YZOM9zzMឍo6FH\o]zЇsȳ m6q~7}37Sfjs VzYz+zom|F;r} o'v=vvy'~;ṋvQ.|vs.ɝ \\z,!fvs/^egD"pu6ِy3--ѧ}G}Թm?9@ A&{# DbVw-azp<98;GQ8Fcy:8qzwk8;'5'$N}:)S>͹[N4g8 u~9;Lyb<@}w`D BE_v.R%8_K/Guug[WP <+Orޫ=>[&s]]׉{ ܠ f&}&9og<6%=}! q["mܦ}ww0wsw{GXߣ}4gn Ѳb@!n#|?B'f'tc)|O]&pgT?f,"Ep*?B<@:P;'$W0D kRA7]"E$ )$Sÿʑ%^\@"BEyΕǯ<óq*_|A2 ?I@ {KQ"*RcE07`xSlT7#E &)9xoIѠ+`@[QVF{`oϷEez[ \71* T*VPU #] w%oR_ Pgj3qL߃{'uǖ}r~48~?`]ձW'4 ؁~#x}Dc,%r}B>!'RQk¡&ud҃O3tL AgMUTu-xԢ0?QuЧԁw]bօi6j3HPfCſ.s.}!;#W=T}^.C}z0 iCoHdd%3%"Wh_kfzI+Ո\Ոk1:5|c6Ư 1 u7!VSۦjL4%7!7ofRs0@4{s4Kszڂ^-&]dOKԒ$Վ9CV7VkMZ~m{O[jjKmK۱y{4kmڞ闽#:D#{;\{8@,f}N;#{ӑkݑ>tltb&R^VfCwfgL<'wbD 5u!~x8s;csKW4>]J s 35ҍwCnՍkĕsEW߃<=_7rawZ>'Iz{b_o|osw%}ه}ۇ=]=+O֞…1vofC >1wǗ^c?|FI?hLmF=4/RG_/9 $^ x?~ǹ b"l09B->q(9Bf#h$[ِp(1\> A" na\ka c~é/>3Ap@Hf u$=ADAӓ!fgB1CJ:|Q\g7afcO ˞X!yðwØpéq8O8,m#n$F܌W :{s:{"$$IG{qGwhbУ1a dx$'u :H%f*qR)i 4⍅Xlc8⌣qaz9h9xZ p<8c=:37=x kkg!qgZm/"b,w19| /|&ǙԛI]ؗѣe[Fˈs93˱+ȷkwEעZž:b:_O貁7 /{6d#=܈Fo$v>9^YΈxo>v;nvkNfn' ]]Բػ9d.9s'2x}C=ܷi/=K޽/{qM;8}q݇m?1x?SbwA?Y!BCuar&avGrf*9dQ8(R1>FOp<89NNI|NRI=)Ν~4ivipY,=<9r/)<:O{ |\{.E/rǗs%n B\yer^_af \۫*|kF_'uױ߀ ~1nҋo&Gx` ~ .m 쿍~wuw{n#=4Oľ>(RBnCb?$#d%=¯lTUL?} Da&,y?Cie#K~}fKR~o:q_dӚl/_''' /=J#QڃZ7^yf7^y{ޫ'?jy<{~ꙓ_uJPj*~̅s}/zYWd/Dd 0 ) # A)&"*Χ9CL EpW;@= Χ9CL E@E1dQ: = x |!/RPԈPRUE((Up HTh"K@(R KdhD- A\EE`X4ˌy f=9߹w̠"2C >Jj"OYD%H0G#PCl[_.KNI҃%e q2V[***8Kug;6Q]-71_n%k~Y^Idd_}Efn QڋPjQOF6,QB{<91ߋyc,̓0y=1ƼQW0悹`. 悹`.K٧ш@b. 悹`dY{n$+vY kLò*a:/ӮcV~cձ?ԉF?$'PA`2l;(?&FZ>p%)9wרψms'L"Rj!QRFlDَ/olW!eLOb]~LC^Πe͇&â.>wCF/!3C/]z-Pнfod>,w98 E`rzvBv]&f'LdRT&1)mi:/9Nz$<<~o_e1ak͑je[Y} Gp}5;} Y?_uگρ1 !rm[_v5eTX7K(2[(>@ƫ;6';gl7v3Na!]ѳA OB#f\X28U+]?l:+_5h}MS7a2c>̳>=>".;{7%xxAV[cAeBvC8G;%x9c1M {NwKޙޙt;;ޛ=b;!nZL9.y8xx;%x>i{_wM>&5oc^wSei0m]Vwwc]uw[ywy(&D=wwy{trBݲS]NuNwKީީ!ӵAxwbr;;Y.yg9xgy{XL? {K[j8A޹޹>V[i '~K0Ǹ+g߇/rs<7~#zأ{QՑ#˼jqɷoqɷ[jnk17\3]kK>k}.Y}3:2X>#E7S͔f{n|ѳ}ݎXj%RXn\̱uu]Nvvc Ļ,u\JY&Ii}LJ$ ISTpN%{uN˓rJfI%'dpv rDI6:!O){Q#-io-Ŵ_gd K2=RӲDjZ^DkKeNkY~crÔGF"e gc_Ueۿ-RN׈  I3vb_5YIݝp}ޓ>|r7пa#1ڋ>APϥ~$"|>[|fyQ=3X$'Sii.byDeQU-aJjYSGP.IZA%ny-T+R"u+Qu t(ED[t'bG"-5J|JKi_M?5QOܸ| Wӏ?7Lym{ZkN=G?>l@dD!>5$p#ݘhBy`7n35kz- ҭ!pq=q=rB;Yv,kK%hڎ,jGþwľ.w>2+mtndx7N;= =XFMt3zNzKEyf4ܪJfYer.E)̪Ϡ?1>vImf,Cff4cT88{q >ĿAA7X7 KkQܧ/4kj vn Us %ֻ|. -!oRbtG6+*O*Zʑh Jx7?Y2 :Od4Kҙh9z- MG('2N&Iy8vc7 4Œ>}*3T}\1OH>ԟA_3tf&gdb9j$c3\fO,B,",b,f|.KȽ%0YJ,rdN͝{L2ɽ;ݹoA嵣`RTM?8och):A Qş^9pB?K'x^EیhhG6iz}KUF롶9hq>.GJMO>Q>"OJ%zȺb]==Pѳh9tZ:V8~h,Z2,Z:R_G<ϢetϡYtϢ0E~tAGVrtttttttGhhLtttttwۼĭs.skYη`7`C Zwnsha>GxCQ9'q6U$ ҃BuW;z*;>u[XFhûC8v=sm`k8=, |z=KLL?zM5'38΍_i=p,G.aD ?X95bJJT?wA,ګ {1$sQB:*'k[8ɧpBT9y Qck?s8~-N"at=Ϡm{,/"W\­- QFg|/uS%*Qp4~^x`pu*;69R0dV{5ګ@}vsOO%H[F}oA0iTa1L0> |ݧgn|w3hH !ccsCIC'5'-=i/Z"s/$yCy | k|-}"Wӯ<!1V{BsAb&D%v9DSŹL㔱81W3W+m*9S=1|Gº;;@>aP(??YVk.g7cEy_Jl3GƜ*yg)%M8o_L[%Ns<Fkx϶&Pܘ(l4A҄M 7Y~qXKJl3G$|4dS.Ӗ&m*9(MP9{5_h@q?E X &o"MH3SPiܚ2iIk1x4l)rq㔱8}cЄ֖s;zip}4Z6nY8i3hok ZjhBbM .(}#[ [ۑbfhؘ83nS.ƴ -Q%Tw&`/h. .P)MdYm]T0_6[WEx]vLʘ}*78 U7)cq}c^ 2T낎7uSG(.=ZdYm]T0͟i6xff=1 >tZ"0v״>i6Aߘ6tAW>GȜSks_\Q䒂Q * >gCr)uArzgVkrvwwvfuޤgoTkE剂CO?E1 mC#S"1׳i =GG{z~@ߧYzNy}6 lv\loN|o Xq%>kp}ke1k'T|<rk(YB]U%5>V݅ՁX +`aӌ [`wa5&:½%_%_{p?)G؝Np' w;æa& w;=ls]ZS]r+oh뚨<`rpS;O9/9CLYVK u>n_у>#[wFXԭٷϿô-:C4МPvߠf B`vz[c'Ї͞?ۚ7gQք#޵ӨY+*o=stBלt퇡Øq䟛jlů=DaK2SbMz0M_O44o/5A?yܗ&){vʇk|?C߲eK߲}ˎ-} b ߌWe:80 C1^t{{уދz/Zz/zGE}%W^^zMz?PݎcA7.ݭfSoi7c_Rznz'ԻwһA%Kxhz2B#1>PyG-==+лS ^v{{كˎz/[z/{GG}=z˙(f; ߁=BN`I cVNerD儲^7Z[u1)[e[Lg=899S e-#\w9旍Wނ9)䄒rt@[C^s^֪ÜBUfU0M!US?0ՔV[2F̹jl~~-8JD)yaBC gn;<7 Zpߖ5('sg3#/9x旍؂T:1>^e |O(ks]riz̆ý s_&OXp< &s+/m1=/[p1^`\< {^r_兜7rоrrY91HM0M3Ϛ1a!rV[2F̹://:)嵆T:.Ӂ yyWex?(Ԃ{Ly-̼dA0M !1a )ǔڒ1buUoce#{SIp]tY0l/SAoFwANIQw;2V)r(8]C8  G_ 㕱6-c2:I/M\I)cqLq?H\]cߒ4!cבG8/^x #M Dd 0 * # A*)". v)BՌeU ~ {;@= v)BՌeU ~J1dQ: =x\]h\E>Vm5DRSbm,dk(XCF nPRDcCa_|䡂BC>ԂEA5߹;N&wޙ6ꆳ{=|ssνYADRVCtZK(#DJtuK`{ai>0CsMhzۣcmmxB)z^U:Hqwnn-픶7N ̃d>?R0sDxZ(ML7:;)1>vv 6 ,/`mNX/l'lqaac3 %t }~p?1{p ^w½ )/ew;Np' !p' wӲ7dr32J5RǕEM3-lըx:wǨW; ?cQxУHqW+z00_{_>*+4x՝<;왽S~;_Dql$h>M],:t-B`zU)Cfz2oNRb8qx"$ynꍰW,Qz\w{8jwRVkr&h=4=I]#H,ed+MjKD|OVz$jXzՃmXʑL 7ȶ6h0<#G?-s xjZi2[[25_+8pv)QJqҗ[Nsv [-Nj(Njߦ͌SmĹkoP7 fa$ f#f͖c fH5m>,+Z`(\kZ[>OnQu\ȇy>AQK}G }G;j﨡}_w.~K8PRdDsz sY39CKM |DW{wV6]!f]ooéQӛX|U]9;boww-7wz#kbꝵ;kuwRﬡwցާ}Zj:zkAf ;CHMk}%bm{k ܽ,[}}]/\>Һ{yZԬ/BD`v"L4.&2x2C˚. ,:o=:Sz=ZoWGsaq7&(h KSQ~򍬏}{!ZE]b's+ZcJN`gh!( 3Ҹ3S9l橰 <C1]/ -v:b :?/_^!U '0n3py} \NX`9h̋s0(yQb[HyA>mn4@> 5 "Ob<|)k 3"/4}Ih1"_`5D yq'>7yAՕ ku-Mz^["cN+h.fbpR^<:9Y00LWz_Z̹M^w4yr6QuGguGUW>/[PyaVއbkyYfi.=!9Fuq")CZaa.ޮGbm{&?/-xOW /0&߇$ߣTuBv?"ez}Sv [ #< v`+ w܋TuqQŸrNżeM2VqˉqY[ŸP\d\3.K怸Ō˝T{C~kNSummaryInformation(DocumentSummaryInformation8CompObjm  , L X d p|Author Guidelines for 8Thomas BaldwinNormalܾ3528Microsoft Office Word@$Vd@&ʗ@( "y@&ʗ!Xg՜.+,0 X`t|  ΢й>;y   FMicrosoft Office Word ĵ MSWordDocWord.Document.89q0000000000000000000000000000000000000,\@\ cke$*$1$A$a$+B*CJOJPJQJ^J_HaJmH sH tH>@> h 1$ & F@& CJ5KHB@B h 2$ & F @&CJ56@6 h 3$ & F@&5$A@$ &؞k=W[SOFi@F nfh@6 h$a$CJ56J@1B6 oRh$a$6CJ]aJ6O6 Abstract Text @O@ Affiliation!$a$CJ6FO"F Abstract Title"$a$CJ5^O2^ Figure and Caption Captions# OJQJ58OB8 Callouts$ OJQJCJ8OAR8 Frame contents%VObV vS Char&d*$!B*CJOJPJQJaJphtH f@sf *fQ*phO r]ua..  ZyblFhe,g*CJaJ8Y@8 3ech~gV+-D M $/AJKLMNOPQRSVWXYZ[\]^gstu     ,5CLYeq|A{     ,  +  *# -./}|{wvrqponl=)< >!A#?"@$$/AJKLMNOPQRSVWXYZ[\]^gstu     ,5CLYeq|      !"#$%&'()*+,-./0123456789:;<=>?@AA{p1p]^M  #,779:;<=>?@ABCEFdx!!##$B%Z%&())))**+,/\1t112 282344679*99;{<<=>?x?? @CGGGGGHH6H7HOHgHhHHI8JPJhJiJKKKKKKKLL2MJMMMMMMMN N?NWNsOOOOSR{SSUBWYWXZZZ\\)^+^F^H^f^^`````%a\a#c7cefgiiiiijj*l@nBnNnPn\n^njnnnooqq1rrsDttuvw xxxyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyzzz z zzz'z(zBzCzDzNzOzQzRz^z_zqzrz~zzzzzzzzzzzzzzzzzzzzzzzzzzzzzz{{{{8{9{:{;{<{={>{?{B{000!0!0!0!000"0 0 0 0 0 0 00 0 000 00000000000 0! 0##0#0###0#0#0## 0****0*0*0*0*0*0** 0444 09099999999 0?8?0?0? 0? 0? 0?0? 0? 0? 0? 0? 0? 0?0?0?0? 0? 0? ?0?0? 0? 0? 0?0? 0? 0? 0?0?0?0? 0? 0? 0?0?0?0?0?0?0?0?0?0??? 0{S{S 0 BWBWBW0BWBW 0 \0\0\0\0\0\\\0\0\0\0\0\\ 0 #c#c 0 f0f0f0f0f0f 0 jj0j0j0j0j0j0j0j 00n 0 0o 0o 0o 0o 0o 0o 0o 0o 0o 0 o 0 o 0 o 0 o 0 o00000000000000000000000000000000000000000000000000000000000000000|X ]M  #,779:;<=>?@ABCEFdx!!##$B%Z%&())))**+,/\1t112 282344679*99;{<<=>?x?? @CGGGGGHH6H7HOHgHhHHI8JPJhJiJKKKKKKKLL2MJMMMMMMMN N?NWNsOOOOSR{SSUBWYWXZZZ\\)^+^F^H^f^^`````%a\a#c7cefgiiiiijj*l@nBnNnPn\n^njnnnooqq1rrsDttuvw xxxB{@0000!00!00!000@00"0{0 0 7 0 0000 00 00 00 00 0000000000000000 00000000000 00 00000000000 000000000000 0000 0000000000 00000 0 000 0 00 0 00000 0 0000 0 000 0 00000 0 0000000000000 00 00000 00000000000000 00 000000 000000000 00 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0HM ] T B!B7c%f9&'t)w*,a-.01247t99g: ;;l=ACDfEEFGIbK.LMPRR TzUU!VWXXZZ[[\t]~^6__GacdeEgiMloorrt{i{hMuNOP R5SSTTBEFGHIJKLMNOPQRSTUVXY[\]^_abcdefghijkmnopqrstuvwxyz{|~F&Z-;bUU%V8ZZf^dpR{B+SpSSTTCWZ`l}TDx) T X "MQ^Hsw9dh8=Lgl &+B%V%X%)))))))))\1p1r1t111222 24262;;;{<<<<<< ======>>>???x????????????;AfAkAGGGGGGGGGGHHHHHH2H4H7HKHMHOHcHeH8JLJNJPJdJfJJJJKKKKKKKKKKLL2MFMHMJM^M`MMMMMMMMMMNNN?NSNUNOOOA{::::::::":"""""""""::::::::::::::::::::::  (*,R$wƐ!/.X48y 0e0e     A@ A5% 8c8c     ?1 d0u0@Ty2 NP'p<'pA)BCD|E||s " 0e@        @ABC DEEFGHIJK5%LMNOPQRSTUWYZ[ \]^_ `abN E5%  N E5%  N F   5%    !"?N@ABC DEFFGHIJK5%LMNOPQRSTUWYZ[ \]^_ `ab`www3f@*B)|  )(  @ |*D    H) ? "6@ NNN?N)  )  # l( ?8 <P"6@ NNN?NM !  (  H' ? "6@ NNN?N("  '  T% ?8 "6@ NNN?NLO !v  %B  HD?"0@NNN?N!Z #Z   T ?8 "6@ NNN?N3' !b B B BD?"0@NNN?N !   H ? "6@ NNN?NN!0   <  ?"6@`NNN?N/   H ? "6@ NNN?N40 B B BD?"0@NNN?N|bB  BD?"0@NNN?N(i \t "^*C # #" "d |*""  0?"6@ NNN?N"^*C  H# ? "6@ NNN?N# * #  H" ? "6@ NNN?N#I )T  "  H! ? "6@ NNN?N# )  !  H ? "6@ NNN?Np#x )    H ? "6@ NNN?N#q)  "  0      !"#$%&'()*+,-./0123456789:;<=>?@A?"6@ NNN?N(DB B BD?"0@NNN?N ] B  BD?"0@NNN?Nr&B B BD?"0@NNN?N 0( B B BD?"0@NNN?N$'.&B  BD?"0@NNN?Ns ] B B BD?"0@NNN?N3   <&  ?"6@`NNN?NX &t ?*-, # #" E'2  H ?8 "6@ NNN?N?* -,   Z ?8 "6@ NNN?N *b-,   H ?8 "6@ NNN?NM*-,   <  ?"6@`NNN?N   6<  ?"6@`NNN?N <xT j;7+" )#" t l S ;l"`;&U ;Z2 q 3 7q"! " 7t n S :n"`jr!! :t o S 9o"`  9t p S 8p"`zx"  8Z2 r 3 6r;#x@&S! 6`B s c $D|e e fB t s *Dg Q e `B u c $D"\ E#e  v s *5v"`%\!(d" 5t" w S 4w"`(m7+ ! 4`B z c $D %"ZB { 3 3{&2`) 3Z2 | 3 2| 2Z" } 3 1}V#\ 1`B ~ c $D`B  c $D.# &ZB  S D`!`B  c $D!R!ZB  S DE'E`B  c $D'q';`B  c $D))gB  <D?"0@NNN?NWWB  BD?"0@NNN?N&\ &L!B  B <D?"0@NNN?NL&S &S B   <D?"0@NNN?N'H 'V!B  BD?"0@NNN?N'H (H B B <D?"0@NNN?NU))B B <D?"0@NNN?N("("B & <D?"0@NNN?NO< C 7CxA{ - t t)8U t5t _Ref178070316 _Ref181866729 _Ref182636441 _Ref182728811 _Ref196322353 _Ref181867177 _Ref181867214 _Ref181867223 _Ref196407260 _Ref182131562 _Ref190156686 _Ref202169507 _Ref202169520 _Ref181868613 _Ref196707636 _Ref178653073 _Ref182923659 _Ref192854071oooqq1rrsDttuvw x xxxxB{ qq.r/r0rrsCttuvw xxxxxxyB{v%D7v%Z!v%|"v% 6v%8v%4E7v%v%8v%l8v%8v%8v%L 7v%,8v%8v%\v%Ýv%ĝv%\ŝv%T Ɲv%D7ǝv%|7ȝv%l6ɝv%6ʝv%T}6˝v%7̝v%6͝v%L{6Νv%6ϝv%$@6Нv%d6ѝv%DZ! "+7GGSSSSSnnrrsvvvxxxB{     *5>LLSSSSSnn)r)rsvvvxxxB{ 9*urn:schemas-microsoft-com:office:smarttagsplace8*urn:schemas-microsoft-com:office:smarttagsCityB*urn:schemas-microsoft-com:office:smarttagscountry-region=*urn:schemas-microsoft-com:office:smarttags PlaceType=*urn:schemas-microsoft-com:office:smarttags PlaceName $^az,2`%b%%%%% &&&& ) )**++++f,j,@.D...y/}/11>2@2a2c22222.303443969><@<J<K<k<m<<<<<<<N=Q===7E9EAECENESEYE[EdEeEgEiEFF@FBFMFOFFFLL=L@LHLKLSLUL\L^LLLLLSSBTHTYl\loooopp5p;p@pIp_pfpipppppqqqqBrLrQrZrermrrrrrssssssStZttuuuu*uuuuuuuuv$v*vvvvvw wwwx$x'x-x1x>xyy yyyykyryyyB{uwx{FOZ%_% ) )1182=222;Q<S<< ====>>??w????????? EgEiELLSLULLLbMdMMM N'NOOZZf^o^``&a/aijjnsnppqqvvyyzzzz{%{B{33333333333333333333333333333 ,7Fx!!##**449*9? @|GGhHHKKKK2MIMMMMMMN N>N{SSBWYW\\+^F^H^f^````#c6cfgjjjjBnNnPn\nnn tDtkyxyyyyyyyyyyyyyyyyyyyyyz zzz&z(zAzDzMzOzPzRz]z_zpzrz}zzzzzzzzzzzzzzzzzzzz{{{{7{B{zx) Y "R^Hx9i>Lm ,))))\1s1t1122 272;AlAJJKKKKMMMMSSUU jjkkl(l2n>n\n]nnnooooooooooooyyB{2\ FNHSt3Xz@^`^`^`^`^`^`^`^`^`\^`\B*CJaJo(phhH[]. H\H^H`\hH) \^`\hH. \^`\hH. 4\4^4`\hH)  \ ^ `\hH. | \| ^| `\hH.  \ ^ `\hH) \^`\hH. \^`\o(hH H\H^H`\hH) \^`\hH. \^`\hH. 4\4^4`\hH)  \ ^ `\hH. | \| ^| `\hH.  \ ^ `\hH) \^`\hH.\^`\OJ QJ o(hHlt\t^t`\OJ QJ o(hHn\^`\OJ QJ o(hHu\^`\OJ QJ o(hHl` \` ^` `\OJ QJ o(hHn \ ^ `\OJ QJ o(hHu \ ^ `\OJ QJ o(hHlL\L^L`\OJ QJ o(hHn\^`\OJ QJ o(hHu2St3X FNOutlineOutline2h@        o%                   x9>-8>N@MdN\jDk5wy{X$T%?)*6?[CZg{7@DMYf_t{ l-8;^LPioxsv-wy{}{n/  7*?@$Vl*vv{W} r')_-19EBR^uek7AG   ]$%'556<>G@dHVV _d#|~ '#i$Q3D`fhk{.:=UJKMVl+~ f* / / 1 h l 2o t KPXioZq$})w,;*E&]dz"A(&4? A#D]FgTm'np|E o!u)/?'K\bjwA##(2@KP[TY_aisy~o$ ucO%.*.Jg]npRqrs4{!}<  !m05<&AlBE[IJMQ>^_f}r1!&4;PQb&gIjiu>|,*-f3DP Y[g2ktf0$02EESToy[#$@'(-15A=RVilo~!KY#,, ?IN^`vak{yt-(*0>>BFHIR]bdg%orv~{e Ks-6ZT\3}   y( `, - t; ? E UO zi k |~ . ^ !5!!!g6!/=!C!H!e!q!y!K"="!">"Z"a"c"j"It"#\,#!/#I/#D#(e#0h#i#o#t#$$&"$w#$w%$:$H$T$Y$_$d$n$,q$`q$u$v$%%%u.%;%E%f%/h%dr%s%f&&&j9&Q&U&[&m&Mt&u&z&{&'''` 'F''W'$'k<'N'BO'\'q_'n'q'u' {'H(q((K()(<*(0(F(rM( S(|S(Z(+k(p({({()!)V) )@)),)-)N1)rH)|O)Q)`)r)v)>*#*$*D*F*\*s*+}.+9+:+F+N+O+1U+0Z+|b+w+w+L, ,,,,2,_6,":,X=,G,G, J,V,l,m,t,P%-;->-N-|Y-]-e-j-l-u-Fx-+y-.* . ..!.h#.3.F.$G.[.f./N// /!/_9/F/]/_/W_/_/b/j/y/ 0000}0*0207070k001V)101B31>1E1rK1+Y1a1zk1#o1t1~12$2&2-23.2|>2A2E25M2:M2QU2W2f2g2"l2q2t2z2n{2)}23[ 3P 33*%353y=3A3Ek3444-4s/494:4;4L4vR4]4_k4p4u4w4r5h555U!5)5A5JV5X5`5d5d5i5 66'66q66b6(6M>6,E6J6r`6a6c6/p6t6774:7D7qE7W7n7z7{788.8F8J8L8S8S8r8ft8999!999O9:9a/969E9P9kf9r9 ::]::W:IY:s]:v::N;; ; ; ;M!;!;$;X=;A;D;G;jM;O;wf;\x;x;};<,<s<#<$<%<*<N-<UA<P<d< s<<<R== ==k="=4=H:=====>D=,F=dM=N=o={=.>6>>>)>*>*>*>CG>O[>wg>2{>W?I?p?\'?0?4?68?W>?pN?vN?N?]?w?}x?x?@@k@!!@&@6@N@>O@ T@X@A AJAAAcA!A'AE+A/A98AXA]aAlApA7sAi}AvBwBB6BD;B=B*KBQhBjBwB5 CC!C2C4C6C7C*=CLCNClDADD3D:D;DDDgD_|D|D_ExE E E EE8 E2EMKGMyJM+ZM}ZM`M=cM!pMkN NNNm+N9NFN>KN+NN&TN\N%`NhNiNBqN}N~NOO*O,O0OgOQnOzOO$P)P1P5DPEPMPTPPBQQk Q)Q*Qs1QCQCQ]Q^Q_Q$wQQ5R%R&ORUR[R`R4bRyR"SSS)S[3SFSHSdS/kSkS?pStSxSeTTT T T T>Tj>T>TMTUTZT]TjfTiTW{T(UQU3U 'UB*U6U 9UWU}UVVVV)V<,VAVFnVVyVzVzVVTWW5W!W~"Wr.W<4W7W!CW$KWMWsW'XX XX'X-X@X HXLX~SXgXYYY5YY2Y8\YJjY8oYxYZZZ+Zr,Zv8Z:Z a a aaPHaSa[a)_a,a%b&bp*ba8bDbIbRbhb1jbwbc*c+c&c0c3c9c CcCcEcFcIcESc4XcYch\cbc0icicxc\dd%ddd6dQdsde e+e:eZe`eeeheueWweS*f:f?f+Gf.MfkMf,Tf `fafQrfufyfg$gLg[g}egkgJpgrgIxg hRh"h"hw-h8h:ho;hFhKh=Sh7Vh]h`hvthvh,whyh]i1 ii=ivii#i(izKiWi_i`ijNjjj[j?jFjJj2Xj$]jkjqj~j^kkkk5k=k@kEk\gkmkpkvk>lldll8ll.lu1lG6lJlgNlWlelhlklkl`ylmm0 m(m/m4m>mEmUPmam/fmVimnmtmumxm nnnnonrn2nN;nFnLnSMn9NnNnt{nooo>oCoCoJoGtoy|o p"p4p5pPpPpppBrpg~peqq+qP'q:4qS4qaqhqbqqxqXr'rrr1rJrTrYr?[rlrqrusC sFssI)s*sM2sgsus}st/ t$t](t:t=t MtdYtZtgtNjtxqtyt uu,ux?xCxDxQxZxZxex9vxoyhy1yy4yyg%y'y.y4y|C| D|@UHQQU.eo8{9#U-9M]e|["OP Z|\hv,5|OVzV M 2->?.?X?^ACEab gzF?&c.LX/fluI37%,4E=J[bm7F1IV)g7oM0 k2q?k@QRWXn[`.js| }x'5X\}0!"4'?? @sWZ`fS8%%f1n;F?VA}AiTP7#)h0<@vSXMbbknuI 5 #$U*0f:GOBU8c?i,.C7J^Y]gm=/w {"+,r.KFZFI ODPXcbch=Y.338CLnyxr!2#2>JgkEpr!s!$&+1>4AO`u-FJW)\hl 0g!=#CLaS\_fAgj7pwS{ D#'*-AD HHK.RW~`MhpppStNvvG t e 3)+5OC-TQVVs-m&2Z4wBZ\`VpXuv&$C//4{?GpX\ '}67:X?^aLgw" "r4^:h:x<6@BEP[[yjXne\A_?nwyyT 19CJBf//4B9>RIu $2 EIKdLqcrt&z{4cR( H &28aGIOS#ZZ`awiuuCx#y#{|9I'(9jEY]_dhWv~J'/DNPV`Nru$d!Z").:W@F\ackkx ,.03::=A0EFdf-rszv)/@1NNi@nFrzU>SY[F\^az !"-f4uA Z[_b fM A"D=JVP.`f7rmr6&''~/M4/8CkKMmoS#(+#9I`JJlL?O{"% -669{CGPT~mmnnxZ-;=JKLWabKray& +#!k"#:DDtN tlxO?x!!=yAe foiqy#8RhUYco]u}u,{R11Va\cxZj18dH^{Iy W 5B|NP*fkCxz0DFtN`6d~ 2>[H\e|p~f2<a>>AGSU]^ dHop" (CDM0PSde d(B2!5@=@FHaeos )F&B..';hBPQE^_aqW3!l#A.|5FMOTYj\$]x_Yap1+<GnSTuvw{$O%".;?SWXw%z>gBEASYffr}P67DY`kMn| 1x2;HK;L>\`epJuv}wLQZQ +p/nBgnv!vG ?>AFCM/T3^klm~ &4--8~Hcmcin`!@$w'(-@DEGIICou#'*!+%,:<vL^cipo u^*58;GLlu;zk~Yp&^-/d@LUa7bsf#K%45*6lCwOT_!aI3 k<CSS~WWX Z^_kb|fw/YT`eg!isvr{{q}X-_/+M4X[rbj-l'{S#&J4n-vv+xQ b&,4Wn3S$$%*:]~_9j}~l"(-g0(2378?&FPP VakA81+9<CKS``qJn L  1%2~8/;BEQ|Xhm2a+r,7X?I]$^ikvwGGGGGHH6H7HOHgHhH8JPJhJiJKKKKKKKLLJMMMMMSPUUGVVAWf^oyyyyyyyyzzzB{ttK۸311@Adobe PDFNe00:Adobe PDF ConverterAdobe PDF ConverterAdobe PDFS 4dA4PRIVB ''''0P4(Dk0EBDAStandardAdobe PDFS 4dA4PRIVB ''''0P4(Dk0EBDAStandard;<=>?CECFKLMNrA{`@`D`F`H`J`@`R`@`Z`\`^`@`@Unknown Gz Times New Roman5Symbol3& z Arial;[SOSimSuncTimesNewRomanTimes New RomanQ& Luxi SansMS PGothic3z Times;" Helvetica7&  Verdana;WingdingsBAhdyff -!Xg>!Xg>I424d;y;y 2qHX ?vSAuthor Guidelines for 8Thomas BaldwinfQZ