Web object-based storage management in proxy caches

Abstract

Proxy caches are essential to improve the performance of the World Wide Web and to enhance user perceived latency. Appropriate cache management strategies are crucial to achieve these goals. In our previous work, we have introduced Web object-based caching policies. A Web object consists of the main HTML page and all of its constituent embedded files. Our studies have shown that these policies improve proxy cache performance substantially.

In this paper, we propose a new Web object-based policy to manage the storage system of a proxy cache. We propose two techniques to improve the storage system performance. The first technique is concerned with prefetching the related files belonging to a Web object, from the disk to main memory. This prefetching improves performance as most of the files can be provided from the main memory rather than from the proxy disk. The second technique stores the Web object members in contiguous disk blocks in order to reduce the disk access time. We used trace-driven simulations to study the performance improvements one can obtain with these two techniques. Our results show that the first technique by itself provides up to 50% reduction in hit latency, which is the delay involved in providing a hit document by the proxy. An additional 5% improvement can be obtained by incorporating the second technique.

Introduction

The growth of the World Wide Web (WWW) has motivated researchers to investigate techniques to improve its performance. Web caching is one of the common techniques used for performance improvement. Several strategies have been proposed for web caching including proxy caching [5], push caching [14] and Harvest caches [8]. In this paper we focus on proxy caching, in which one or more computers act as a cache for a set of WWW clients.

Proxy caching improves performance by reducing user latency in obtaining Web documents. The reason for this improvement is that proxy caches are typically closer to clients than to the web server. This proximity to clients also reduces network traffic. Furthermore, since the proxy serves most of the client requests, web server load decreases substantially, thereby improving its performance.

We introduced a Web object-based prefetching technique for proxy caches [3]. This technique takes the web document structure into account. A Web objectconsists of the main HTML page and all of its constituent embedded files. Our study of popular Web documents to characterize Web objects has shown that most Web objects are small in size and can therefore be cached effectively at proxy or client sites [4]. In the Web object-based cache policy, a reference to a member of a Web object is considered as a reference to all the members of the Web object. If the currently referenced file is not in the cache, it not only fetches the referenced file but also prefetches all the other members of the Web object into the proxy cache. Our studies have shown that this prefetching technique significantly improves the hit ratio without generating excessive network traffic [3]. In this research, the Web object concept is used to devise storage management policies for improving the performance of proxy caches.

In this paper we refer to the delay incurred in retrieving a hit document (i.e., a document present in the proxy) as hit latency. Most proxy caches use the main memory along with the disk as the cache. Hit latency is reduced if the proxy provides more documents from its main memory rather than from the disk. A recent study by Rousskov and Soloviev [28] reports that, for Squid [32], the disk delay contributes to about 30% of the total hit latency. Furthermore, recent research indicates that proxy servers spend more time on file-related activities [24]. For example, the results reported in [20] show that, on an average, several disk accesses are needed for each URL request. Thus, the disk subsystem may become a potential bottleneck in the proxies. Moreover, reducing the disk response time of a proxy reduces the latency of hit documents.

We have proposed two techniques to improve storage system performance of proxy caches. Both techniques are based on the concept of Web objects.

• In the first technique, we use Web object-based prefetching to bring the related files from the disk to the main memory.

• The second technique uses the Web object information to place the files in a Web object contiguously, so that the disk access time can be reduced.

To study the performance of the proposed technique, we have used trace-driven simulation using the DiskSim disk simulator [11] for three types of disk drives.

The remainder of the paper is organized as follows. Previous work on proxy cache storage management is briefly discussed in Section 2. The research methodology used in our experiments is given in Section 3. Section 4 gives an overview of the Web object-based caching policy. Section 5 presents details on the Web object-based storage management policy. Performance of Web object-based file placement is discussed in Section 6. Performance of Web object-based policy without prefectching is studied in Section 7. The impact of these policies on hit latency is discussed in Section 8. The last section presents our conclusions.