Efficient Remote Data Access in a Mobile Computing Environment

Size: px
Start display at page:

Download "Efficient Remote Data Access in a Mobile Computing Environment"

Transcription

1 This paper appears in the ICPP 2000 Workshop on Pervasive Computing Efficient Remote Data Access in a Mobile Computing Environment Laura Bright Louiqa Raschid University of Maryland College Park, MD bright@cs.umd.edu, louiqa@umiacs.umd.edu Abstract We consider the problem of remote data access for multiple clients in mobile computing environments. In this environment, a base station provides a wireless communication link between mobile clients and remote servers that reside on a fixed network. Clients send requests to the base station for objects which reside on the remote servers. The base station downloads these objects from the remote servers and transmits them to the clients. A drawback in this environment is that remote data access across a fixed network is slow. When locally cached copies of data exist at the base station, accessing them reduces latency. However, the cached data becomes old as more recent data is made available at the remote servers. In this paper we consider a framework that allows mobile clients to specify their preferences about recency. The objective is to maximize the recency of the data for all clients while delivering data to clients as quickly as possible. We develop an on-demand strategy which determines when the base station should access data from remote servers and when it should return data copies from its cache. We present analytical results that show the benefits of the on-demand approach. Introduction The growing presence of mobile computing and wireless communication has increased the number of clients that access data remotely. In mobile computing environments, there is typically a set of servers on a fixed network that provide data to a set of mobile clients. The geographic area is divided into cells. Each cell has a base station, a fixed server with a wireless interface to communicate with mobile clients. The base station is connected to a fixed network and provides a wireless communication link between mobile clients in its cell and remote servers on the fixed network. This research has been partially supported by the Defense Advanced Research Project Agency undergrant , and the National Science Foundation under grant IRI Mobile clients send requests to the base station for objects which reside on the remote servers. The base station then downloads the objects from the remote servers and transmits them to the clients. The architecture is illustrated in Figure. Details of this architecture are described in [2]. A drawback in this environment is that remote data access is slow. This is because the base station must download objects from remote servers, and there may be delays due to network traffic and server workloads. Also, as the base station downloads more data over the fixed network, the overall latency may increase due to bandwidth contention. This may be unacceptable if clients require fast answers to their requests. In addition, a client may be connected to the base station in its cell for a short period of time, and then disconnect or move to a different cell, so the base station must serve client requests in a timely manner. Finally, in a mobile computing environment the wireless downlink from the base station to the clients typically has limited bandwidth. To deliver data to as many clients as possible, it is important to maximize utilization of this bandwidth. If there is too much delay in downloading data from remote sources, some of the available downlink bandwidth may be idle. Maintaining a cache at the base station can improve the above situation. The cache at the base station can store data obtained from remote servers. If the base station has a copy of an object in its cache, the cached copy can be used to answer client requests and therefore reduce the latency of downloading data across the fixed network. However, the cached data becomes stale as more recent data is made available at remote servers. Clients may have different preferences about the recency of the data. Some clients may prefer the most recent data even if it takes longer to access, while others may accept less recent data if it can be accessed quickly. In this environment, a challenge is determining how to maintain the appropriate level of recency in the cache while delivering data to clients as quickly as possible. Given a request for an object and a client s recency preference, the challenge is choosing which objects the base station should access remotely and which to access from the cache. One possible strategy is to refresh the cache in the background as in []. When an object is updated at a remote server, the base station can download a new copy of the ob-

2 ject without waiting for a client to request it. We refer to this as an asynchronous approach. While this strategy speeds up data delivery to clients by bringing the requested objects into the cache, it has two major drawbacks. First, asynchronous updates may waste bandwidth, since some objects may not be requested by clients between consecutive updates. This is especially true when there is a skew in the popularity of objects. Second, no matter how frequently the base station downloads objects from remote servers, the cache will never be completely up to date. This is because objects may be updated at any point in time. Even if the base station could request every object as soon as it was updated, there would still be some delay in downloading the object from the remote server, and during this time the cached object would be stale. This is studied in detail in []. In Section 3 we present analysis that explores these two drawbacks in greater detail, and motivates our proposed solution. In this paper we present a solution that provides more recent data and consumes less bandwidth than the asynchronous approach. We propose an on-demand approach, where we read some objects from the cache and access others remotely. Reading some data from the cache reduces the amount of time that the wireless downlink bandwidth may be idle while the base station waits for data to arrive from remote servers. It also consumes less bandwidth on the fixed network than accessing all requested objects remotely. We examine the problem of how to choose which objects the base station should access remotely when it receives requests from a large number of clients, and when copies of objects cached at the base station have varying degrees of recency. We map the problem to the knapsack problem [3] and use dynamic programming to solve it. We analyze how the relationships between object size, number of requests per object, and recency of cached objects affect the solution space. This analysis provides useful insights into the problem and presents areas for future work. We note that while we focus on a mobile computing environment in this paper, our results are applicable to any enviroment where time or bandwidth constraints make it impractical to access all requested data remotely. For example, our work could be applied to web proxy caching. This paper is organized as follows. Section 2 formally describes the problem. In Section 3 we present analysis that motivates an on-demand approach. In Section 4 we explore the solution space of the on-demand approach. Section 5 surveys related work, and Section 6 concludes. 2 Problem Definition We consider one or more mobile clients in a single cell that send requests to the same base station. The base station has cached copies of some objects which can be used to answer requests, but the cached copies may be less recent base station mobile clients wireless cell remote servers fixed network base station mobile clients wireless cell Figure. Architecture of a Mobile Computing Environment than the copies on the remote servers. The model is pullbased. Remote servers do not know the interests of the mobile clients, and only provide data when it is requested by the base station. All queries from the mobile clients are read-only, but data at the remote servers may be updated. Any object that the base station does not download will be read from the base station cache. Note that each object may have a different size. When a client requests an object, he can specify a target recency value C for the object. If the recency score x of the cached copy of the object meets or exceeds C, the object gets a score of.0. If the cached copy of the object is less recent than the target value, its score is computed using a scoring function. The score approaches 0 as x gets further from C. Example scoring functions are f C (x) = =(+abs(x=c? )) and f C (x) = exp(?abs(x=c? )). The score of any object accessed remotely is set to.0. We focus on two problems. The first problem is as follows. Given a set of requested objects (of varying sizes), a set of client recency requirements for each object, a cache containing a possibly less recent copy of each object, and a limit on the amount of data that can be downloaded, we need to decide which objects to access remotely, and which to read from the cache. The goal is to maximize the average client recency score over all objects. The second problem is to determine an upper bound on the amount of data that should be downloaded to answer a set of requests. Downloading too little data will result in delivering stale data to a large number of clients. However, downloading too much data will increase latency for all clients and may result in idle downlink bandwidth. In this paper we present analytical results that provide some insights into these two problems. We make a number of simplifying assumptions. We assume that each client requests only one object, but the same object may be requested by multiple clients. We do

3 not consider the workload on servers from clients in other cells. We assume that the base station can cache a copy of every object that is requested. We consider only objects that are stored in the cache. Since any object that is not in the cache must be downloaded, this assumption allows us to focus on studying the tradeoffs between downloading data and using cached copies. Given a set of client requests and an upper bound on the amount of data that can be downloaded, the problem maps to the knapsack problem [3]. The knapsack problem considers a set of objects Ū, where each object u has size s(u) and profit p(u). Given a knapsack with capacity C, the problem is to choose a set of objects in Ū to maximize the total value P u2ū p(u) subject to P u2ū s(u) C. The mapping of our problem to the knapsack problem is as follows. Let each data object u have size s(u). The profit p(u) of each object u is computed as follows. Suppose that the set of clients who request an object u is requests(u). For each client i, we compute the client s recency score of the cached object using the client s target recency (C i ) and the recency (x) of the cached copy. Recall that the remote copy has a score of.0. We compute benefit(i), the difference between the score of the remote object (.0) and the score of the cached object for client i. This value is the benefit to client i of downloading object u. Note that the value of benefit(i) increases as C i is more recent and when the cached object is older. The profit p(u) of an object u is the sum of the benefit values of all clients that request u. This takes into account both the benefit to each client of downloading the object and popularity of an object, i.e. the number of clients that request the object. This mapping gives higher profit (i.e. a greater benefit of downloading) to remote objects that are requested by many clients or have older cached copies. Note that this mapping considers only the limit on the amount of bandwidth that can be used to answer a set of queries, and does not model network latency. The knapsack problem is known to be NP-complete but can be solved in pseudo-polynomial time using dynamic programming [3]. There are also polynomial time approximation algorithms. In this paper we use dynamic programming to explore the solution space. 3 Motivation In this section we present analysis that provides motivation for accessing some objects on demand, rather than updating the cache asynchronously in the background. Recall that there are two potential drawbacks to the asynchronous approach. First, asynchronously updating the cache may waste bandwidth if the base station downloads objects that are not requested by clients. Second, regardless of how often the base station downloads updated objects, the data in the cache may not be completely up-to-date. Below we present analysis that illustrates these two drawbacks. 3. Amount of Data Downloaded In our first analysis we measure how much data must be downloaded to deliver the most recent data to all users using an asynchronous approach and an on-demand approach. We compare the asynchronous approach, where every object in the cache is downloaded every time it is updated at the remote server, to an on-demand approach where an object is downloaded only if it is requested and its cached copy is stale with respect to the remote server. Note that in this first analysis, we do not use any stale cached objects to answer requests. We do not assume any limitations on the amount of data we can download. We illustrate the potential savings of the on-demand approach with some simple analysis. For simplicity, we consider a scenario where there are 500 objects of uniform size. All objects are updated simultaneously, and this occurs once every 5 time units. Thus, updates occur at time 0, 5, 0, etc. We note that our results were similar for varying object sizes, but we omit these results due to space considerations. In our analysis, we consider the effects of both the number of client requests and the skew in these requests on the amount of data that must be downloaded. We ran multiple simulations, each with a fixed number of requests per time unit. We varied the number of requests per time unit from 0 to 500. For each simulation, we started with an empty cache, and warmed up the cache for 00 time units. We then ran the simulation for an additional 500 time units and measured the total number of objects downloaded during this time. Recall that in the on-demand approach, an object is downloaded only when it is requested by at least one client and its cached copy is stale with respect to the server. In the asynchronous case, the amount of data that must be downloaded depends only on the rate that objects are updated and is independent of both the number of client requests and the skew in these requests. In our analysis, each object was updated once every 5 time units, and each simulation ran for 500 time units in the steady state. Therefore each object was updated exactly 00 times. Since there were 500 objects, a total 50,000 units of data needed to be downloaded to keep the cache completely up to date. This is an upper bound on the amount of data that must be downloaded to deliver the most recent data to all clients. The asynchronous approach and the on-demand approach are compared in Figure 2. The dotted line across the top of the graph indicates the amount of data that is downloaded in the asynchronous case, and this is compared with the on-demand approach for three different client access patterns. The solid curve in Figure 2 shows the case where all objects were requested with equal probability. For low re-

4 quest rates, the on-demand approach provides considerable savings. This is because many objects that are stale in the cache are not requested before they are updated again at the remote server, so there is no need to download them. When the number of request per second is greater than 300, the on-demand approach downloads nearly as much data as the asynchronous approach. These results suggest that when all objects are requested equally, the on-demand approach provides the greatest benefit when there are few client requests. As the skew in client requests increases, the benefit to the on-demand approach increases. The dashed curve in Figure 2 shows the case when the requests were skewed uniformly (i.e. the ith most popular object was requested with probability proportional to i). The dotdash curve shows a higher degree of skew, where the requests had a zipf distribution (i.e. the ith most popular object was requested with probability proportional to =i). For higher degrees of skew in requests, the on-demand approach provides greater savings in terms of the amount of data that must be downloaded. This is because many more objects are not requested between updates. Number of Objects Downloaded x Objects updated every 5 time units asynchronous on demand uniform access on demand skewed(uniform) on demand skewed(zipf) Number of Requests per time unit Figure 2. Amount of data downloaded to provide the most recent data to all clients, for varying skew in requests 3.2 Recency of Data In the previous analysis, we showed that the on-demand approach requires downloading less data than the asynchronous approach to deliver the most recent data to clients. Next, we assume that there is a limit on how much data can be downloaded. We compare the recency of data delivered to clients using these two approaches under this constraint. For this analysis, we considered 500 objects of unit size and uniform access probability. 00 objects were requested per time unit. We ran multiple simulations and varied the Average recency low update frequency 0.2 on demand asynch Average recency high update frequency 0.2 Data Downloaded per time unit on demand asynch Figure 3. Average recency of data as amount of data downloaded increases number of objects downloaded per time unit from to 00. We began each simulation with an empty cache and warmed up the cache for 50 time units. We then ran each simulation for 00 time units and measured the average recency of all objects delivered to clients. We note that our results were similar for varying object sizes and skew in popularity. In the asynchronous approach, objects were downloaded in a fixed order in a round robin manner. At each time interval, if k was the upper bound on the number of objects to download, the next k objects in the fixed order were downloaded and updated in the cache. In the on-demand approach, at each time interval, the k requested objects with the lowest recency in the cache were selected to be downloaded. The remaining requested objects were read from the cache. Both simulations used the same set of randomly generated client requests. We measured the recency of cached objects as follows. Any object in the cache that was up-to-date had a value of. Each time an object in the cache was updated at the remote server, its recency score x decayed using the function x0 = C=(=x + ), where C is a constant. The results of these simulations are shown in Figure 3. The graphs show how the average recency changes when objects are updated with low frequency (once every 0 time units) and high frequency (once every time unit). Recall that 00 objects were requested in each time unit. In the ondemand case, as the amount of data downloaded per time unit approaches 00, most requested objects can be downloaded, so the recency approaches. In contrast, using the asynchronous approach, 00 arbitrary objects are downloaded at each time unit, and the remaining 400 objects become more stale. Since the 00 client requests are independent of which objects are downloaded, many requested objects are stale in the cache. When objects are updated with high frequency, the asynchronous approach performs poorly.

5 Parameter range distribution Object Size [-20] uniform N um Requests [-20] uniform or constant Cache Recency Score [0.-.0] uniform Table. Parameter values for each object and their distributions These graphs show that when objects are updated more frequently and there is a limit on how much data can be downloaded, the on-demand approach provides a greater benefit. This is because the asynchronous approach may download objects at a lower rate than they are updated. In contrast, the on-demand approach always accesses the most recent copies of some objects, and therefore provides more recent data to clients. 4 Analysis of Solution Space In this section we present analytical results using the dynamic programming solution to the knapsack problem. Recall our problem: given a set of requested objects, a set of client recency requirements for each object, a cache containing a possibly less recent copy of each object, and a limit on how much data can be downloaded, we must determine which objects to access remotely, and which to read from the cache. The goal is to maximize the average client recency score over all objects. Our results show how the average recency score for all clients improves as we increase the upper bound on the amount of data downloaded. We note that as more data is downloaded, the average recency score for a set of requests will improve. However, downloading large amounts of data may increase the delay in delivering data to clients. The dynamic programming algorithm works by finding the optimal solution for downloading a single unit of data, and then finds the solution for larger amounts of data until the upper bound is reached. The algorithm is well-suited to our analysis because it allows us to observe how the quality of the solution changes as the upper bound increases. We first describe the parameter values we used in the analysis. We then discuss our results. 4. Setup The name, distribution, and range of values for the three parameters that were varied in this study are summarized in Table. The first parameter is the sizes of the objects, Object Size. We used a uniform distribution in the range from [-20]. The second parameter is the number of clients who request each object, N um Requests. In the first analysis, N um Requests was set to a constant of. In the remaining analysis, N um Requests was uniformly distributed in the range from [-20]. The third parameter we consider is Cache Recency Score. This is the recency score of a cached object averaged over the clients who request the object. A uniform distribution in the range from 0. to.0 was used. The number of clients in our study was 5000, and the number of distinct objects requested by these clients was 500. The sum of the sizes of these objects was 5000 units. Therefore, the maximum number of units of data downloaded in this analysis is All analysis was performed using synthetic data that was generated to conform to Table. In each set of results, we measure the value of Average Score as a function of the upper bound on the number of data units downloaded. Average Score is the average recency score of a solution for all clients. It is computed as follows. Recall that a solution is a set of objects to access remotely, and the remaining objects are read from the cache. If an object is downloaded, the score for every client that requested it is.0. If an object is read from the cache, the score for each client that requested it is computed as a function of the client s target recency C i as described in Section 2. Average Score is the average of the scores of all of the clients. Since our objective is to provide the most recent data to the largest number of clients, Average Score is a good way to measure the quality of the solution. We note that Average Score can be understood in the context of the knapsack problem as follows. In our mapping to the knapsack problem, the profit of an object is equal to the number of clients requesting the object, N um Requests, times the average benefit to these clients of downloading the object, avg benefit. Recall that benefit(i) is the amount by which the score will increase for client i if an object is downloaded. In this mapping, choosing a set of objects with the maximum profit will maximize the average value of benefit over all clients. It is straightforward to verify that maximizing the benefit will also maximize the average client score, i.e. Average Score. 4.2 Analysis We study the effects of the parameters Object Size, N um Requests, and Cache Recency Score on the Average Score. We consider the effects of both positive and negative correlations between each of these three parameters. We show when there is a benefit to downloading a large amount of data, and when downloadinga small amount of data will provide an answer that is almost as good. Our analysis indicates that the distribution of these parameters and the way they are correlated play a significant role in the rate at which Average Score increases as more data is

6 downloaded. These results identify workloads and cache states for which downloading a relatively small amount of data provides nearly as good an answer as downloading a larger amount of data. This information is useful in determining how to use the available bandwidth most effectively Uniform Access We first consider the case where all objects were requested by the same number of clients. Figure 4 plots the Average Score versus the upper bound on the number of data units downloaded. The solid line shows the case where larger objects had higher Cache Recency Score values in the cache, i.e. there is a positive correlation between Object Size and Cache Recency Score. The dashed line shows the case where larger objects had lower values of Cache Recency Score, i.e. a negative correlation. The dashdot line shows a case where Cache Recency Score and Object Size were uncorrelated. Figure 4 shows that Average Score increases rapidly and then levels off when the large objects have the highest Cache Recency Score values (solid line). In contrast, when the largest objects have the lowest Cache Recency Score (dashed line) values, Average Score increases gradually. The uncorrelated case (dashdot line) lies in between these two. When the upper bound on the number of data units downloaded is low, only smaller objects will be downloaded. As the upper bound increases, it is possible to download an increasing number of large objects. Thus, when the large objects are the ones with the highest Cache Recency Score values, the Average Score will increase dramatically when small objects are downloaded, and it will level off as larger objects are downloaded. In contrast, when the larger objects have lower Cache Recency Score values (dashed line), there is less benefit to downloading small objects. However, there is an increasing benefit to downloading the larger objects with the lower Cache Recency Score values. Thus, the Average Score continues to increase as the number of units of data downloaded is increased Skewed Access Next, we study the effect of skew in the number of users requesting each object. A hot object is one that is requested by many users, i.e. its value for N um Requests is high. We control skew by varying the correlation between N um Requests and Object Size. Figure 5 plots the Average Score versus number of units of data downloaded when there is a correlation between Object Size and N um Requests. In both Figure 5(a) and Figure 5(b), when no data is downloaded, the scores vary due the differences in correlations between Cache Recency Score and Object Size. All objects accessed equally large objs high scores large objs low scores no correlation Figure 4. Effect of correlations between Object Size and Cache Recency Score when all objects are requested equally Figure 5(a) shows the case when the smaller objects are the hottest, i.e. there is a negative correlation between N um Requests and Object Size. For the solid line, large objects have the highest Cache Recency Score values. Since the small objects are the hottest, when no data was downloaded the Average Score was lowest. The Average Score values converge very quickly. The corner of the dotted rectangle indicates the point where the scores of all three cases are greater than. There is not a significant increase in the score once 2000 units of data are downloaded. This analysis suggests that when small objects are the hottest, there may not be a significant benefit to downloading a large amount of data. Figure 5(b) shows the case when the large objects were hottest. As the number of units of data downloaded increases, the Average Score increases steadily in all cases. The scores do not approach until about 3500 units of data are downloaded, as shown by the corner of the dotted rectangle in the figure. This suggests that when large objects are hottest, there is an increasing benefit to downloading a large amount of data. The next set of analytical results studies the effect of a positive or negative correlation between Object Size and Cache Recency Score. Figure 6(a) plots Average Score versus the upper bound on number of units of data downloaded when small objects have the highest Cache Recency Score values, and large objects have the lowest. In this case, Average Score increases steadily independent of the positive or negative correlation between Object Size and N um Requests. Thus, there is a significant benefit to downloading a larger number of units of data, especially when the large objects are hotter (solid line). The scores of all three do not reach until a

7 (a) Small objects hot (a) Small objects have highest recency scores large objs high scores large objs low scores no correlation large objects hot small objects hot uniform access (b) Large objects hot (b) Large objects have highest recency scores large objs high scores large objs low scores no correlation Figure 5. Effect of correlations between Object Size and Num Requests large objects hot small objects hot uniform access Figure 6. Effect of correlations between Object Size and Cache Recency Score large amount of data is downloaded, as shown by the dotted rectangle. There is a significant benefit to downloading as much as 4000 units of data. Figure 6(b) shows the case where large objects have a higher Cache Recency Score. The Average Score values for all three cases (correlations between Object Size and N um Requests) converge very quickly. Note that all three curves are similar to those shown in Figure 5(a). The corner of the dotted rectangle indicates the point where the scores for all three exceed, which occurs when about 2000 units of data are downloaded. This suggests that when large objects have the highest Cache Recency Score values, there may not be a significant benefit to downloading large amounts of data. 5 Related Work There has been much work on data dissemination that addresses the issue of which pages to broadcast to a set of clients in a limited bandwidth environment [4, 5, 6, 7, 8]. This work is similar to ours in that multiple clients share limited downlink bandwidth. However, this work assumes that data cached at the base station is up-to-date. In contrast, we consider the problem of providing the most recent data to clients when data cached at the base station may be stale. The work in Broadcast Disks [4, 5, 6] that is most similar to ours is [6]. This introduces a pull-based backchannel which allows clients to explicitly request data that will not be broadcast for a long period of time. As in our work, a client must choose whether or not to access data remotely. In our work, if data is not accessed remotely, the cached copy will be used instead. Similarly, in [6], if data is not requested via the backchannel, clients must wait for it to be broadcast. Another similarity is that there is contention among all clients for the available pull bandwidth. When a client sends a request via the backchannel, it is queued with other requests. In our work, a similar problem arises due to contention among clients for the fixed network bandwidth. Another focus of work in data dissemination is cache consistency. Work in [5, 7, 8] addresses the issue of how to maintain consistent data in client caches when there are updates at the server. All of these works assume that servers can push information to clients, while our environment is pull-based. Work in [5] extends broadcast disks to handle

8 updates to pages. In [8] a server periodically broadcasts invalidation reports to clients to tell them which data has been updated. Work in [7] describes a weaker type of consistency which is similar to what we propose. It introduces the term quasi-copy, a cached value that is allowed to deviate from a server value in a controlled way. For example, a client querying stock prices may be satisfied with cached stock prices that are within 5 percent of actual prices. This is similar to our work which allows users to specify the desired degree of recency of cached data. Work in [] describes strategies to maintain the recency of cached data in the presence of updates at the remote servers. This work assumes that all requested objects will be read from the cache, and the goal is to maximize the overall recency of the data in the cache at any point in time. This is similar to the asynchronous strategy we describe. The authors provide techniques to keep the cached data as up-todate as possible. In our analysis of the on-demand approach, we have shown that it is possible to provide more recent data by downloading some requested objects. The benefit is greater with higher update frequency and bandwidth limitations. Further, [] does not consider the popularity of objects, and therefore may waste bandwidth by downloading objects that will not be requested by any clients. There is relevant work on Web caching and replication that aims to reduce latency and bandwidth usage in Web environments. Research reported in Harvest [0] and in [9], as well as several commercial products, for example Akamai s FreeFlow [], Inktomi s TrafficServer [3], and CacheFlow [2] aim to reduce latency and bandwidth consumption on the internet by replicating web content in multiple locations. These products offer a solution when the Internet itself is a bottleneck, i.e. the cause of latency is network congestion. 6 Conclusions and Future Work We have identified a challenge that arises in mobile computing environments when multiple clients in a cell connect to a single base server. To access objects from remote servers, clients send requests to the base station and the base station downloads the objects from the remote servers. In such an environment, it may be inefficient for a base station to download all data from remote servers. Mobile clients must rely on less recent data cached at the base station to answer some requests. We have shown that asynchronously updating a cache may waste bandwidth and may not provide the most recent data to clients. We have proposed an ondemand approach to determine which objects to download and which to read from a cache to answer a given set of requests. We have shown that our approach maps to the knapsack problem, and we have studied how the quality of the solution changes for varying client access patterns and recency of cached data. In future work, we will develop techniques to determine how much data the base station should download to satisfy a set of requests. The techniques will use knowledge of the current workload and recency of cached data to determine an upper bound on the amount of data to download. Our analysis shows that under some circumstances there is not a great benefit to downloading large amounts of data. In these cases the techniques will choose a smaller upper bound. Another area of future work is developing caching policies when cache space at the base station is limited. In this case, the problem is not only what data to download, but also what data to cache. We will consider cache replacement policies based on client requests and knowledge of server updates. Our analysis has shown the effect of the cache state on the recency score of a set of requests, and may provide some insights into effective caching policies. 7 Acknowledgements We thank Bobby Bhattacharjee, Selcuk Candan, Mike Franklin, and Samir Khuller for their feedback. References [] [2] [3] [4] S. Acharya, R. Alonso, M. Franklin, and S. Zdonik. Broadcast disks: Data management for asymmetric communication environments. Proc. ACM SIGMOD Conf., 995. [5] S. Acharya, M. Franklin, and S. Zdonik. Disseminating updates on broadcast disks. Proc. VLDB Conf., 996. [6] S. Acharya, M. Franklin, and S. Zdonik. Balancing push and pull for data broadcast. Proc. ACM SIGMOD Conf., 997. [7] R. Alonso, D. Barbara, and H. Garcia-Molina. Data caching issues in an information retrieval system. ACM. TODS Vol. 5, no. 3, 990. [8] D. Barbara and T. Imielinski. Sleepers and workaholics: Caching strategies in mobile environments (extended version). VLDB Journal Special Issue of the best of SIGMOD System-Oriented Papers, 995. [9] A. Bestavros, R. Carter, M. Crovella, C. Cunha, A. Heddaya, and S. Mirdad. Application-level document caching in the internet. Proc. 2nd Workshop on Services in Distributed and Networked Environments, 995. [0] A. Chankhunthod, P. Danzig, C. Neerdaels, M. F. Schwarz, and K. Worrell. A hierarchical internet object cache. Proc. USENIX Technical Conference, 996. [] J. Cho and H. Garcia-Molina. Synchronizing a database to improve freshness. Proc. ACM SIGMOD Conf., [2] T. Imielinski and B. Badrinath. Mobile wireless computing. Communications of the ACM, October 994. [3] S. Martello and P. Toth. Knapsack Problems: Algorithms and Computer Implementations. John Wiley and Sons, Chichester England, 990.

Stretch-Optimal Scheduling for On-Demand Data Broadcasts

Stretch-Optimal Scheduling for On-Demand Data Broadcasts Stretch-Optimal Scheduling for On-Demand Data roadcasts Yiqiong Wu and Guohong Cao Department of Computer Science & Engineering The Pennsylvania State University, University Park, PA 6 E-mail: fywu,gcaog@cse.psu.edu

More information

Project Report, CS 862 Quasi-Consistency and Caching with Broadcast Disks

Project Report, CS 862 Quasi-Consistency and Caching with Broadcast Disks Project Report, CS 862 Quasi-Consistency and Caching with Broadcast Disks Rashmi Srinivasa Dec 7, 1999 Abstract Among the concurrency control techniques proposed for transactional clients in broadcast

More information

Web-based Energy-efficient Cache Invalidation in Wireless Mobile Environment

Web-based Energy-efficient Cache Invalidation in Wireless Mobile Environment Web-based Energy-efficient Cache Invalidation in Wireless Mobile Environment Y.-K. Chang, M.-H. Hong, and Y.-W. Ting Dept. of Computer Science & Information Engineering, National Cheng Kung University

More information

Evaluation of a Broadcast Scheduling Algorithm

Evaluation of a Broadcast Scheduling Algorithm Evaluation of a Broadcast Scheduling Algorithm Murat Karakaya 1 and Özgür Ulusoy2 1 Department of Technical Sciences Turkish Land Forces Academy, Ankara 06100, Turkey 2 Department of Computer Engineering

More information

Quasi-consistency and Caching with Broadcast Disks

Quasi-consistency and Caching with Broadcast Disks Quasi-consistency and Caching with Broadcast Disks Rashmi Srinivasa and Sang H. Son Department of Computer Science University of Virginia Charlottesville, VA {rashmi, son}@cs.virginia.edu Abstract. The

More information

Dynamic Broadcast Scheduling in DDBMS

Dynamic Broadcast Scheduling in DDBMS Dynamic Broadcast Scheduling in DDBMS Babu Santhalingam #1, C.Gunasekar #2, K.Jayakumar #3 #1 Asst. Professor, Computer Science and Applications Department, SCSVMV University, Kanchipuram, India, #2 Research

More information

Pull vs. Hybrid: Comparing Scheduling Algorithms for Asymmetric Time-Constrained Environments

Pull vs. Hybrid: Comparing Scheduling Algorithms for Asymmetric Time-Constrained Environments Pull vs. Hybrid: Comparing Scheduling Algorithms for Asymmetric Time-Constrained Environments Jesus Fernandez-Conde and Daniel Mozos Department of Computer Architecture Universidad Complutense de Madrid,

More information

Multiversion Data Broadcast

Multiversion Data Broadcast 1224 IEEE TRANSACTIONS ON COMPUTERS, VOL. 51, NO. 10, OCTOBER 2002 Multiversion Data Broadcast Evaggelia Pitoura, Member, IEEE Computer Society, and Panos K. Chrysanthis, Member, IEEE Abstract Recently,

More information

Energy-Efficient Mobile Cache Invalidation

Energy-Efficient Mobile Cache Invalidation Distributed and Parallel Databases 6, 351 372 (1998) c 1998 Kluwer Academic Publishers. Manufactured in The Netherlands. Energy-Efficient Mobile Cache Invalidation KUN-LUNG WU, PHILIP S. YU AND MING-SYAN

More information

CPET 565/CPET 499 Mobile Computing Systems. Lecture 8. Data Dissemination and Management. 2 of 3

CPET 565/CPET 499 Mobile Computing Systems. Lecture 8. Data Dissemination and Management. 2 of 3 CPET 565/CPET 499 Mobile Computing Systems Lecture 8 and Management 2 of 3 Based on the Text used in the course: Fundamentals of Mobile & Pervasive Computing, 2005, by Frank Adelstein, et. al, from McGraw-Hill

More information

Volume 3, Issue 9, September 2013 International Journal of Advanced Research in Computer Science and Software Engineering

Volume 3, Issue 9, September 2013 International Journal of Advanced Research in Computer Science and Software Engineering Volume 3, Issue 9, September 2013 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Optimal Round

More information

Signature caching techniques for information filtering in mobile environments

Signature caching techniques for information filtering in mobile environments Wireless Networks 5 1999) 57 67 57 Signature caching techniques for information filtering in mobile environments Wang-Chien Lee a, and Dik Lun Lee b a GTE Laboratories Incorporated, 40 Sylvan Road, Waltham,

More information

An Adaptive Query Processing Method according to System Environments in Database Broadcasting Systems

An Adaptive Query Processing Method according to System Environments in Database Broadcasting Systems An Query Processing Method according to System Environments in Database Broadcasting Systems M. KASHITA T. TERADA T. HARA Graduate School of Engineering, Cybermedia Center, Graduate School of Information

More information

Pull vs Push: A Quantitative Comparison for Data Broadcast

Pull vs Push: A Quantitative Comparison for Data Broadcast Pull vs Push: A Quantitative Comparison for Data Broadcast Demet Aksoy Mason Sin-Fai Leung Computer Science Department University of California, Davis (aksoy,leungm)@cs.ucdavis.edu Abstract Advances in

More information

A Generalized Target-Driven Cache Replacement Policy for Mobile Environments

A Generalized Target-Driven Cache Replacement Policy for Mobile Environments A Generalized Target-Driven Cache Replacement Policy for Mobile Environments Liangzhong Yin, Guohong Cao Department of Computer Science & Engineering The Pennsylvania State University University Park,

More information

QUERY PLANNING FOR CONTINUOUS AGGREGATION QUERIES USING DATA AGGREGATORS

QUERY PLANNING FOR CONTINUOUS AGGREGATION QUERIES USING DATA AGGREGATORS QUERY PLANNING FOR CONTINUOUS AGGREGATION QUERIES USING DATA AGGREGATORS A. SATEESH 1, D. ANIL 2, M. KIRANKUMAR 3 ABSTRACT: Continuous aggregation queries are used to monitor the changes in data with time

More information

Load Sharing in Peer-to-Peer Networks using Dynamic Replication

Load Sharing in Peer-to-Peer Networks using Dynamic Replication Load Sharing in Peer-to-Peer Networks using Dynamic Replication S Rajasekhar, B Rong, K Y Lai, I Khalil and Z Tari School of Computer Science and Information Technology RMIT University, Melbourne 3, Australia

More information

On a Cooperation of Broadcast Scheduling and Base Station Caching in the Hybrid Wireless Broadcast Environment

On a Cooperation of Broadcast Scheduling and Base Station Caching in the Hybrid Wireless Broadcast Environment On a Cooperation of Broadcast Scheduling and Base Station Caching in the Hybrid Wireless Broadcast Environment Jing Cai, Tsutomu Terada, Takahiro Hara, and Shojiro Nishio Outline Background Hybrid Wireless

More information

A Proxy Caching Scheme for Continuous Media Streams on the Internet

A Proxy Caching Scheme for Continuous Media Streams on the Internet A Proxy Caching Scheme for Continuous Media Streams on the Internet Eun-Ji Lim, Seong-Ho park, Hyeon-Ok Hong, Ki-Dong Chung Department of Computer Science, Pusan National University Jang Jun Dong, San

More information

Competitive Analysis of On-line Algorithms for On-demand Data Broadcast Scheduling

Competitive Analysis of On-line Algorithms for On-demand Data Broadcast Scheduling Competitive Analysis of On-line Algorithms for On-demand Data Broadcast Scheduling Weizhen Mao Department of Computer Science The College of William and Mary Williamsburg, VA 23187-8795 USA wm@cs.wm.edu

More information

Maintaining Mutual Consistency for Cached Web Objects

Maintaining Mutual Consistency for Cached Web Objects Maintaining Mutual Consistency for Cached Web Objects Bhuvan Urgaonkar, Anoop George Ninan, Mohammad Salimullah Raunak Prashant Shenoy and Krithi Ramamritham Department of Computer Science, University

More information

Hybrid Cooperative Caching in a Mobile Environment

Hybrid Cooperative Caching in a Mobile Environment Hybrid Cooperative Caching in a Mobile Environment Hong Va Leong 1, Chi-yin Chow 2 1 Department of Computing, The Hong Kong Polytechnic University Hong Kong cshleong@comp.polyu.edu.hk 2 Department of Computing

More information

Improving VoD System Efficiency with Multicast and Caching

Improving VoD System Efficiency with Multicast and Caching Improving VoD System Efficiency with Multicast and Caching Jack Yiu-bun Lee Department of Information Engineering The Chinese University of Hong Kong Contents 1. Introduction 2. Previous Works 3. UVoD

More information

CERIAS Tech Report Autonomous Transaction Processing Using Data Dependency in Mobile Environments by I Chung, B Bhargava, M Mahoui, L Lilien

CERIAS Tech Report Autonomous Transaction Processing Using Data Dependency in Mobile Environments by I Chung, B Bhargava, M Mahoui, L Lilien CERIAS Tech Report 2003-56 Autonomous Transaction Processing Using Data Dependency in Mobile Environments by I Chung, B Bhargava, M Mahoui, L Lilien Center for Education and Research Information Assurance

More information

Objective-Optimal Algorithms for Long-term Web Prefetching

Objective-Optimal Algorithms for Long-term Web Prefetching Objective-Optimal Algorithms for Long-term Web Prefetching Bin Wu and Ajay D Kshemkalyani Dept of Computer Science University of Illinois at Chicago Chicago IL 667 bwu ajayk @csuicedu Abstract Web prefetching

More information

Reduction of Periodic Broadcast Resource Requirements with Proxy Caching

Reduction of Periodic Broadcast Resource Requirements with Proxy Caching Reduction of Periodic Broadcast Resource Requirements with Proxy Caching Ewa Kusmierek and David H.C. Du Digital Technology Center and Department of Computer Science and Engineering University of Minnesota

More information

Intelligent Data Receiver Mechanism for Wireless Broadcasting System

Intelligent Data Receiver Mechanism for Wireless Broadcasting System Intelligent Data Receiver Mechanism for Wireless Broadcasting System Lien-Fa Lin 1 and Yuan-Ko Huang 2 Department of Information Communication of Kao Yuan University 1,2 lienfa@cc.kyu.edu.tw 1, huangyk@cc.kyu.edu.tw

More information

Scaling Distributed Machine Learning with the Parameter Server

Scaling Distributed Machine Learning with the Parameter Server Scaling Distributed Machine Learning with the Parameter Server Mu Li, David G. Andersen, Jun Woo Park, Alexander J. Smola, Amr Ahmed, Vanja Josifovski, James Long, Eugene J. Shekita, and Bor-Yiing Su Presented

More information

THE falling cost of both communication and mobile

THE falling cost of both communication and mobile IEEE TANSACTIONS ON KNOWLEDGE AND DATA ENGINEEING, VOL. 15, NO. 5, SEPTEMBE/OCTOBE 2003 1251 A Scalable Low-Latency Cache Invalidation Strategy for Mobile Environments Guohong Cao, Member, IEEE Abstract

More information

Cost Models for Query Processing Strategies in the Active Data Repository

Cost Models for Query Processing Strategies in the Active Data Repository Cost Models for Query rocessing Strategies in the Active Data Repository Chialin Chang Institute for Advanced Computer Studies and Department of Computer Science University of Maryland, College ark 272

More information

Transaction Processing in Mobile Database Systems

Transaction Processing in Mobile Database Systems Ashish Jain* 1 http://dx.doi.org/10.18090/samriddhi.v7i2.8631 ABSTRACT In a mobile computing environment, a potentially large number of mobile and fixed users may simultaneously access shared data; therefore,

More information

Data Warehousing Alternatives for Mobile Environments

Data Warehousing Alternatives for Mobile Environments Data Warehousing Alternatives for Mobile Environments I. Stanoi D. Agrawal A. El Abbadi Department of Computer Science University of California Santa Barbara, CA 93106 S. H. Phatak B. R. Badrinath Department

More information

Cache-Miss-Initiated Prefetch in Mobile Environments

Cache-Miss-Initiated Prefetch in Mobile Environments Cache-Miss-Initiated Prefetch in Mobile Environments Hui Song and Guohong Cao Department of Computer Science & Engineering The Pennsylvania State University University Park, PA 16802 {hsong, gcao}@cse.psu.edu

More information

Broadcast Disks: Scalable solution for an asymmetric environment

Broadcast Disks: Scalable solution for an asymmetric environment Broadcast Disks: Scalable solution for an asymmetric environment Byungho Lee Computer Science & Engineering, University of Texas at Arlington blee@cse.ute.edu Abstract As mobile computing gains more popularity,

More information

Assignment 7: TCP and Congestion Control Due the week of October 29/30, 2015

Assignment 7: TCP and Congestion Control Due the week of October 29/30, 2015 Assignment 7: TCP and Congestion Control Due the week of October 29/30, 2015 I d like to complete our exploration of TCP by taking a close look at the topic of congestion control in TCP. To prepare for

More information

On Improving the Performance of Cache Invalidation in Mobile Environments

On Improving the Performance of Cache Invalidation in Mobile Environments Mobile Networks and Applications 7, 291 303, 2002 2002 Kluwer Academic Publishers. Manufactured in The Netherlands. On Improving the Performance of Cache Invalidation in Mobile Environments GUOHONG CAO

More information

A Hybrid Data Delivery Method of Data Broadcasting and On-demand Wireless Communication

A Hybrid Data Delivery Method of Data Broadcasting and On-demand Wireless Communication Vol. 47 No. SIG 19(TOD 32) IPSJ Transactions on Databases Dec. 2006 Regular Paper A Hybrid Data Delivery Method of Data Broadcasting and On-demand Wireless Communication Jing Cai, Tsutomu Terada, Takahiro

More information

Inital Starting Point Analysis for K-Means Clustering: A Case Study

Inital Starting Point Analysis for K-Means Clustering: A Case Study lemson University TigerPrints Publications School of omputing 3-26 Inital Starting Point Analysis for K-Means lustering: A ase Study Amy Apon lemson University, aapon@clemson.edu Frank Robinson Vanderbilt

More information

Record Placement Based on Data Skew Using Solid State Drives

Record Placement Based on Data Skew Using Solid State Drives Record Placement Based on Data Skew Using Solid State Drives Jun Suzuki 1, Shivaram Venkataraman 2, Sameer Agarwal 2, Michael Franklin 2, and Ion Stoica 2 1 Green Platform Research Laboratories, NEC j-suzuki@ax.jp.nec.com

More information

An Optimal Cache Replacement Policy for Wireless Data Dissemination under Cache Consistency

An Optimal Cache Replacement Policy for Wireless Data Dissemination under Cache Consistency An Optimal Cache Replacement Policy for Wireless Data Dissemination under Cache Consistency Jianliang Xu HK University of Science and Technology Clear Water Bay, HK xujl@cs.ust.hk Wang-Chien Lee Verizon

More information

Improvement of Buffer Scheme for Delay Tolerant Networks

Improvement of Buffer Scheme for Delay Tolerant Networks Improvement of Buffer Scheme for Delay Tolerant Networks Jian Shen 1,2, Jin Wang 1,2, Li Ma 1,2, Ilyong Chung 3 1 Jiangsu Engineering Center of Network Monitoring, Nanjing University of Information Science

More information

On Sample-Path Staleness in Lazy Data Replication

On Sample-Path Staleness in Lazy Data Replication On Sample-Path Staleness in Lazy Data Replication Xiaoyong Li, Daren B.H. Cline and Dmitri Loguinov Internet Research Lab Department of Computer Science and Engineering Texas A&M University April 29, 2015

More information

Leveraging Transitive Relations for Crowdsourced Joins*

Leveraging Transitive Relations for Crowdsourced Joins* Leveraging Transitive Relations for Crowdsourced Joins* Jiannan Wang #, Guoliang Li #, Tim Kraska, Michael J. Franklin, Jianhua Feng # # Department of Computer Science, Tsinghua University, Brown University,

More information

Autonomous Transaction Processing Using Data Dependency in Mobile Environments Λ

Autonomous Transaction Processing Using Data Dependency in Mobile Environments Λ Autonomous Transaction Processing Using Data Dependency in Mobile Environments Λ IlYoung Chung, 1y Bharat Bhargava, 1 Malika Mahoui, 2 and Leszek Lilien 1 1 Department of Computer Sciences and Center for

More information

Optimized Paging Cache Mappings for efficient location management Hyun Jun Lee, Myoung Chul Jung, and Jai Yong Lee

Optimized Paging Cache Mappings for efficient location management Hyun Jun Lee, Myoung Chul Jung, and Jai Yong Lee Optimized Paging Cache Mappings for efficient location management Hyun Jun Lee, Myoung Chul Jung, and Jai Yong Lee Abstract Cellular IP maintains distributed cache for location management and routing purposes.

More information

Periodic Scheduling in On-Demand Broadcast System

Periodic Scheduling in On-Demand Broadcast System Periodic Scheduling in On-Demand Broadcast System Nitin Prabhu, Vijay Kumar SCE, Computer Networking University of Missouri-Kansas City Kansas City, MO 64110 npp21c (kumarv)@umkc.edu Indrakshi Ray Computer

More information

6 Distributed data management I Hashing

6 Distributed data management I Hashing 6 Distributed data management I Hashing There are two major approaches for the management of data in distributed systems: hashing and caching. The hashing approach tries to minimize the use of communication

More information

Performance Evaluation of a Wireless Hierarchical Data Dissemination System

Performance Evaluation of a Wireless Hierarchical Data Dissemination System Performance Evaluation of a Wireless Hierarchical Data Dissemination System Qinglong Hu*& Dik Lun Lee Department of Computer Science University of Science and Technology Clear Water Bay, Hong Kong qinglong@cs.ust.hk

More information

Data Indexing for Heterogeneous Multiple Broadcast Channel

Data Indexing for Heterogeneous Multiple Broadcast Channel Data Indexing for Heterogeneous Multiple Broadcast Channel Andrew Y. Ho and Dik Lun Lee Department of Computer Science The Hong Kong University of Science and Technology Clear Water Bay, Hong Kong Email:

More information

The Google File System

The Google File System October 13, 2010 Based on: S. Ghemawat, H. Gobioff, and S.-T. Leung: The Google file system, in Proceedings ACM SOSP 2003, Lake George, NY, USA, October 2003. 1 Assumptions Interface Architecture Single

More information

Lecture (08, 09) Routing in Switched Networks

Lecture (08, 09) Routing in Switched Networks Agenda Lecture (08, 09) Routing in Switched Networks Dr. Ahmed ElShafee Routing protocols Fixed Flooding Random Adaptive ARPANET Routing Strategies ١ Dr. Ahmed ElShafee, ACU Fall 2011, Networks I ٢ Dr.

More information

Replication architecture

Replication architecture Replication INF 5040 autumn 2008 lecturer: Roman Vitenberg INF5040, Roman Vitenberg 1 Replication architecture Client Front end Replica Client Front end Server Replica INF5040, Roman Vitenberg 2 INF 5040

More information

Investigation on OLSR Routing Protocol Efficiency

Investigation on OLSR Routing Protocol Efficiency Investigation on OLSR Routing Protocol Efficiency JIRI HOSEK 1, KAROL MOLNAR 2 Department of Telecommunications Faculty of Electrical Engineering and Communication, Brno University of Technology Purkynova

More information

A Top Catching Scheme Consistency Controlling in Hybrid P2P Network

A Top Catching Scheme Consistency Controlling in Hybrid P2P Network A Top Catching Scheme Consistency Controlling in Hybrid P2P Network V. Asha*1, P Ramesh Babu*2 M.Tech (CSE) Student Department of CSE, Priyadarshini Institute of Technology & Science, Chintalapudi, Guntur(Dist),

More information

FLEXIBLE DATA DISSEMINATION STRATEGY FOR EFFECTIVE CACHE CONSISTENCY IN MOBILE WIRELESS COMMUNICATION NETWORKS

FLEXIBLE DATA DISSEMINATION STRATEGY FOR EFFECTIVE CACHE CONSISTENCY IN MOBILE WIRELESS COMMUNICATION NETWORKS FLEXIBLE DATA DISSEMINATION STRATEGY FOR EFFECTIVE CACHE CONSISTENCY IN MOBILE WIRELESS COMMUNICATION NETWORKS Kahkashan Tabassum 1 Asia Sultana 2 and Dr. A. Damodaram 3 1 Department of CS & IT, Maulana

More information

Data Access on Wireless Broadcast Channels using Keywords

Data Access on Wireless Broadcast Channels using Keywords Data Access on Wireless Broadcast Channels using Keywords Mr. Vijaykumar Mantri 1, Mr. Nagaraju A 2 Dept of IT, Padmasri Dr. B. V. Raju Institute of Technology, Narsapur, Dist.Medak, Andhra Pradesh, India.

More information

The Design and Implementation of a High-Performance Storage Server. Final Report. Stanley B. Zdonik Brown University

The Design and Implementation of a High-Performance Storage Server. Final Report. Stanley B. Zdonik Brown University The Design and Implementation of a High-Performance Storage Server Final Report Stanley B. Zdonik Brown University 1. ARPA ORDER NUMBER: 8220 2. BAA NUMBER: BAA 90-21 3. CONTRACT/GRANT NUMBER: N0014-91-J-4085

More information

Peer-to-Peer Cooperative Caching in Mobile Environments

Peer-to-Peer Cooperative Caching in Mobile Environments Peer-to-Peer Cooperative Caching in Mobile Environments Chi-Yin Chow Hong Va Leong Alvin Chan Department of Computing, Hong Kong Polytechnic University, Hong Kong E-mail: {cscychow, cshleong, cstschan}@comp.polyu.edu.hk

More information

Coding and Scheduling for Efficient Loss-Resilient Data Broadcasting

Coding and Scheduling for Efficient Loss-Resilient Data Broadcasting Coding and Scheduling for Efficient Loss-Resilient Data Broadcasting Kevin Foltz Lihao Xu Jehoshua Bruck California Institute of Technology Department of Computer Science Department of Electrical Engineering

More information

A Comparative study of On-Demand Data Delivery with Tables Driven and On-Demand Protocols for Mobile Ad-Hoc Network

A Comparative study of On-Demand Data Delivery with Tables Driven and On-Demand Protocols for Mobile Ad-Hoc Network A Comparative study of On-Demand Data Delivery with Tables Driven and On-Demand Protocols for Mobile Ad-Hoc Network Humayun Bakht Research Fellow, London School of Commerce, United Kingdom humayunbakht@yahoo.co.uk

More information

A Novel Replication Strategy for Efficient XML Data Broadcast in Wireless Mobile Networks

A Novel Replication Strategy for Efficient XML Data Broadcast in Wireless Mobile Networks JOURNAL OF INFORMATION SCIENCE AND ENGINEERING 32, 309-327 (2016) A Novel Replication Strategy for Efficient XML Data Broadcast in Wireless Mobile Networks ALI BORJIAN BOROUJENI 1 AND MEGHDAD MIRABI 2

More information

Answers to Sample Questions on Transport Layer

Answers to Sample Questions on Transport Layer Answers to Sample Questions on Transport Layer 1) Which protocol Go-Back-N or Selective-Repeat - makes more efficient use of network bandwidth? Why? Answer: Selective repeat makes more efficient use of

More information

Adaptive Data Dissemination in Mobile ad-hoc Networks

Adaptive Data Dissemination in Mobile ad-hoc Networks Adaptive Data Dissemination in Mobile ad-hoc Networks Joos-Hendrik Böse, Frank Bregulla, Katharina Hahn, Manuel Scholz Freie Universität Berlin, Institute of Computer Science, Takustr. 9, 14195 Berlin

More information

Increase-Decrease Congestion Control for Real-time Streaming: Scalability

Increase-Decrease Congestion Control for Real-time Streaming: Scalability Increase-Decrease Congestion Control for Real-time Streaming: Scalability Dmitri Loguinov City University of New York Hayder Radha Michigan State University 1 Motivation Current Internet video streaming

More information

Adaptive Aggregation Scheduling Using. Aggregation-degree Control in Sensor Network

Adaptive Aggregation Scheduling Using. Aggregation-degree Control in Sensor Network Contemporary Engineering Sciences, Vol. 7, 2014, no. 14, 725-730 HIKARI Ltd, www.m-hikari.com http://dx.doi.org/10.12988/ces.2014.4681 Adaptive Aggregation Scheduling Using Aggregation-degree Control in

More information

Effects of Sensor Nodes Mobility on Routing Energy Consumption Level and Performance of Wireless Sensor Networks

Effects of Sensor Nodes Mobility on Routing Energy Consumption Level and Performance of Wireless Sensor Networks Effects of Sensor Nodes Mobility on Routing Energy Consumption Level and Performance of Wireless Sensor Networks Mina Malekzadeh Golestan University Zohre Fereidooni Golestan University M.H. Shahrokh Abadi

More information

Last Class: Consistency Models. Today: Implementation Issues

Last Class: Consistency Models. Today: Implementation Issues Last Class: Consistency Models Need for replication Data-centric consistency Strict, linearizable, sequential, causal, FIFO Lecture 15, page 1 Today: Implementation Issues Replica placement Use web caching

More information

Stochastic Models of Pull-Based Data Replication in P2P Systems

Stochastic Models of Pull-Based Data Replication in P2P Systems Stochastic Models of Pull-Based Data Replication in P2P Systems Xiaoyong Li and Dmitri Loguinov Presented by Zhongmei Yao Internet Research Lab Department of Computer Science and Engineering Texas A&M

More information

ACMS: The Akamai Configuration Management System. A. Sherman, P. H. Lisiecki, A. Berkheimer, and J. Wein

ACMS: The Akamai Configuration Management System. A. Sherman, P. H. Lisiecki, A. Berkheimer, and J. Wein ACMS: The Akamai Configuration Management System A. Sherman, P. H. Lisiecki, A. Berkheimer, and J. Wein Instructor: Fabian Bustamante Presented by: Mario Sanchez The Akamai Platform Over 15,000 servers

More information

A Simulation-Based Analysis of Scheduling Policies for Multimedia Servers

A Simulation-Based Analysis of Scheduling Policies for Multimedia Servers A Simulation-Based Analysis of Scheduling Policies for Multimedia Servers Nabil J. Sarhan Chita R. Das Department of Computer Science and Engineering The Pennsylvania State University University Park,

More information

Disseminating Updates on Broadcast Disks

Disseminating Updates on Broadcast Disks Disseminating Updates on Broadcast Disks Swarup Acharya Brown University sa@cs.brown.edu Michael Franklin University of Maryland franklin@cs.umd.edu Stanley Zdonik Brown University sbz@cs.brown.edu Abstract

More information

An Enhanced Dynamic Framed Slotted ALOHA Algorithm for RFID Tag Identification

An Enhanced Dynamic Framed Slotted ALOHA Algorithm for RFID Tag Identification An Enhanced Dynamic Framed Slotted ALOHA Algorithm for RFID Tag Identification Su-Ryun Lee Ajou University Department of Electrical and Computer Engineering Suwon, Korea srlee@ajouackr Sung-Don Joo LG

More information

Assignment 5. Georgia Koloniari

Assignment 5. Georgia Koloniari Assignment 5 Georgia Koloniari 2. "Peer-to-Peer Computing" 1. What is the definition of a p2p system given by the authors in sec 1? Compare it with at least one of the definitions surveyed in the last

More information

Report on Cache-Oblivious Priority Queue and Graph Algorithm Applications[1]

Report on Cache-Oblivious Priority Queue and Graph Algorithm Applications[1] Report on Cache-Oblivious Priority Queue and Graph Algorithm Applications[1] Marc André Tanner May 30, 2014 Abstract This report contains two main sections: In section 1 the cache-oblivious computational

More information

Analytic Performance Models for Bounded Queueing Systems

Analytic Performance Models for Bounded Queueing Systems Analytic Performance Models for Bounded Queueing Systems Praveen Krishnamurthy Roger D. Chamberlain Praveen Krishnamurthy and Roger D. Chamberlain, Analytic Performance Models for Bounded Queueing Systems,

More information

An AIAD-Based Adaptive Routing Protocol in Ad-Hoc Wireless Networks

An AIAD-Based Adaptive Routing Protocol in Ad-Hoc Wireless Networks An AIAD-Based Adaptive Routing Protocol in Ad-Hoc Wireless Networks Youn-Sik Hong 1 and Ki-Young Lee 2 1 Department of Computer Science and Eng. 2 Department of Information and Telecommunication Eng.,

More information

THE CACHE REPLACEMENT POLICY AND ITS SIMULATION RESULTS

THE CACHE REPLACEMENT POLICY AND ITS SIMULATION RESULTS THE CACHE REPLACEMENT POLICY AND ITS SIMULATION RESULTS 1 ZHU QIANG, 2 SUN YUQIANG 1 Zhejiang University of Media and Communications, Hangzhou 310018, P.R. China 2 Changzhou University, Changzhou 213022,

More information

Network Load Balancing Methods: Experimental Comparisons and Improvement

Network Load Balancing Methods: Experimental Comparisons and Improvement Network Load Balancing Methods: Experimental Comparisons and Improvement Abstract Load balancing algorithms play critical roles in systems where the workload has to be distributed across multiple resources,

More information

Max-Count Aggregation Estimation for Moving Points

Max-Count Aggregation Estimation for Moving Points Max-Count Aggregation Estimation for Moving Points Yi Chen Peter Revesz Dept. of Computer Science and Engineering, University of Nebraska-Lincoln, Lincoln, NE 68588, USA Abstract Many interesting problems

More information

Data Dissemination Techniques in Mobile Computing Environment

Data Dissemination Techniques in Mobile Computing Environment ABSTRACT 2016 IJSRST Volume 2 Issue 6 Print ISSN: 2395-6011 Online ISSN: 2395-602X Themed Section: Science and Technology Data Dissemination Techniques in Mobile Computing Environment R. Kavitha *1, S.

More information

Performance Modeling of Proxy Cache Servers

Performance Modeling of Proxy Cache Servers Journal of Universal Computer Science, vol. 2, no. 9 (2006), 39-53 submitted: 3/2/05, accepted: 2/5/06, appeared: 28/9/06 J.UCS Performance Modeling of Proxy Cache Servers Tamás Bérczes, János Sztrik (Department

More information

Today: World Wide Web! Traditional Web-Based Systems!

Today: World Wide Web! Traditional Web-Based Systems! Today: World Wide Web! WWW principles Case Study: web caching as an illustrative example Invalidate versus updates Push versus Pull Cooperation between replicas Lecture 22, page 1 Traditional Web-Based

More information

Performance of Multihop Communications Using Logical Topologies on Optical Torus Networks

Performance of Multihop Communications Using Logical Topologies on Optical Torus Networks Performance of Multihop Communications Using Logical Topologies on Optical Torus Networks X. Yuan, R. Melhem and R. Gupta Department of Computer Science University of Pittsburgh Pittsburgh, PA 156 fxyuan,

More information

CS 425 / ECE 428 Distributed Systems Fall 2015

CS 425 / ECE 428 Distributed Systems Fall 2015 CS 425 / ECE 428 Distributed Systems Fall 2015 Indranil Gupta (Indy) Measurement Studies Lecture 23 Nov 10, 2015 Reading: See links on website All Slides IG 1 Motivation We design algorithms, implement

More information

Surveying Formal and Practical Approaches for Optimal Placement of Replicas on the Web

Surveying Formal and Practical Approaches for Optimal Placement of Replicas on the Web Surveying Formal and Practical Approaches for Optimal Placement of Replicas on the Web TR020701 April 2002 Erbil Yilmaz Department of Computer Science The Florida State University Tallahassee, FL 32306

More information

STATISTICAL TECHNIQUES. Interpreting Basic Statistical Values

STATISTICAL TECHNIQUES. Interpreting Basic Statistical Values STATISTICAL TECHNIQUES Interpreting Basic Statistical Values INTERPRETING BASIC STATISTICAL VALUES Sample representative How would one represent the average or typical piece of information from a given

More information

Plot SIZE. How will execution time grow with SIZE? Actual Data. int array[size]; int A = 0;

Plot SIZE. How will execution time grow with SIZE? Actual Data. int array[size]; int A = 0; How will execution time grow with SIZE? int array[size]; int A = ; for (int i = ; i < ; i++) { for (int j = ; j < SIZE ; j++) { A += array[j]; } TIME } Plot SIZE Actual Data 45 4 5 5 Series 5 5 4 6 8 Memory

More information

SLA-Aware Adaptive Data Broadcasting in Wireless Environments. Adrian Daniel Popescu

SLA-Aware Adaptive Data Broadcasting in Wireless Environments. Adrian Daniel Popescu SLA-Aware Adaptive Data Broadcasting in Wireless Environments by Adrian Daniel Popescu A thesis submitted in conformity with the requirements for the degree of Masters of Applied Science Graduate Department

More information

Cost Based Data Dissemination in Satellite Networks

Cost Based Data Dissemination in Satellite Networks Cost Based Data Dissemination in Satellite Networks Bo Xu and Ouri Wolfson Department of Electrical Engineering and Computer Science University of Illinois at Chicago Sam Chamberlain Army Research Laboratory

More information

COMPARISON OF ENERGY EFFICIENT DATA TRANSMISSION APPROACHES FOR FLAT WIRELESS SENSOR NETWORKS

COMPARISON OF ENERGY EFFICIENT DATA TRANSMISSION APPROACHES FOR FLAT WIRELESS SENSOR NETWORKS COMPARISON OF ENERGY EFFICIENT DATA TRANSMISSION APPROACHES FOR FLAT WIRELESS SENSOR NETWORKS Saraswati Mishra 1 and Prabhjot Kaur 2 Department of Electrical, Electronics and Communication Engineering,

More information

Nowadays data-intensive applications play a

Nowadays data-intensive applications play a Journal of Advances in Computer Engineering and Technology, 3(2) 2017 Data Replication-Based Scheduling in Cloud Computing Environment Bahareh Rahmati 1, Amir Masoud Rahmani 2 Received (2016-02-02) Accepted

More information

FOUR EDGE-INDEPENDENT SPANNING TREES 1

FOUR EDGE-INDEPENDENT SPANNING TREES 1 FOUR EDGE-INDEPENDENT SPANNING TREES 1 Alexander Hoyer and Robin Thomas School of Mathematics Georgia Institute of Technology Atlanta, Georgia 30332-0160, USA ABSTRACT We prove an ear-decomposition theorem

More information

DELL EMC CX4 EXCHANGE PERFORMANCE THE ADVANTAGES OF DEPLOYING DELL/EMC CX4 STORAGE IN MICROSOFT EXCHANGE ENVIRONMENTS. Dell Inc.

DELL EMC CX4 EXCHANGE PERFORMANCE THE ADVANTAGES OF DEPLOYING DELL/EMC CX4 STORAGE IN MICROSOFT EXCHANGE ENVIRONMENTS. Dell Inc. DELL EMC CX4 EXCHANGE PERFORMANCE THE ADVANTAGES OF DEPLOYING DELL/EMC CX4 STORAGE IN MICROSOFT EXCHANGE ENVIRONMENTS Dell Inc. October 2008 Visit www.dell.com/emc for more information on Dell/EMC Storage.

More information

Research Article A Two-Level Cache for Distributed Information Retrieval in Search Engines

Research Article A Two-Level Cache for Distributed Information Retrieval in Search Engines The Scientific World Journal Volume 2013, Article ID 596724, 6 pages http://dx.doi.org/10.1155/2013/596724 Research Article A Two-Level Cache for Distributed Information Retrieval in Search Engines Weizhe

More information

Data Migration on Parallel Disks

Data Migration on Parallel Disks Data Migration on Parallel Disks Leana Golubchik 1, Samir Khuller 2, Yoo-Ah Kim 2, Svetlana Shargorodskaya, and Yung-Chun (Justin) Wan 2 1 CS and EE-Systems Departments, IMSC, and ISI, University of Southern

More information

Evaluating the Impact of Different Document Types on the Performance of Web Cache Replacement Schemes *

Evaluating the Impact of Different Document Types on the Performance of Web Cache Replacement Schemes * Evaluating the Impact of Different Document Types on the Performance of Web Cache Replacement Schemes * Christoph Lindemann and Oliver P. Waldhorst University of Dortmund Department of Computer Science

More information

Adaptive Local Route Optimization in Hierarchical Mobile IPv6 Networks

Adaptive Local Route Optimization in Hierarchical Mobile IPv6 Networks Adaptive Local Route Optimization in Hierarchical Mobile IPv6 Networks Sangheon Pack, Taekyoung Kwon, and Yanghee Choi School of Computer Science and Engineering Seoul National University, Seoul, Korea

More information

Multiprocessor and Real-Time Scheduling. Chapter 10

Multiprocessor and Real-Time Scheduling. Chapter 10 Multiprocessor and Real-Time Scheduling Chapter 10 1 Roadmap Multiprocessor Scheduling Real-Time Scheduling Linux Scheduling Unix SVR4 Scheduling Windows Scheduling Classifications of Multiprocessor Systems

More information

Using Statistics for Computing Joins with MapReduce

Using Statistics for Computing Joins with MapReduce Using Statistics for Computing Joins with MapReduce Theresa Csar 1, Reinhard Pichler 1, Emanuel Sallinger 1, and Vadim Savenkov 2 1 Vienna University of Technology {csar, pichler, sallinger}@dbaituwienacat

More information

Introduction to Operating Systems Prof. Chester Rebeiro Department of Computer Science and Engineering Indian Institute of Technology, Madras

Introduction to Operating Systems Prof. Chester Rebeiro Department of Computer Science and Engineering Indian Institute of Technology, Madras Introduction to Operating Systems Prof. Chester Rebeiro Department of Computer Science and Engineering Indian Institute of Technology, Madras Week 05 Lecture 18 CPU Scheduling Hello. In this lecture, we

More information