Abstract. Search engines and large scale IR systems need to cache query results for efficiency and scalability purposes. In this study, we propose to explicitly incorporate the query costs in the static caching policy. To this end, a query’s cost is represented by its execution time, which involves CPU time to decompress the postings and compute the query-document similarities to obtain the final top-N answers. Simulation results using a large Web crawl data and a real query log reveal that the proposed strategy improves overall system performance in terms of the total query execution time. 1

Low-latency, high-throughput web services are typically a-chieved through partitioning, replication, and caching. Although these strategies and the general design of large-scale distributed search systems are well known, the academic literature provides surprisingly few details on deployment and operational considerations in production environments. In this paper, we address this gap by sharing the distributed search architecture that underlies Twitter user search, a service for discovering relevant accounts on the popular microblogging service. Our design makes use of the principle that eliminates the distinction between failure and other anticipated service disruptions: as a result, most operational scenarios share exactly the same code path. This simplicity leads to greater robustness and fault-tolerance. Another salient feature of our architecture is its exclusive reliance on open-source software components, which makes it easier for the community to learn from our experiences and replicate our findings.

Query result caching is an important mechanism for search engine efficiency. In this study, we first review several query features that are used to determine the contents of a static result cache. Next, we introduce a new feature that more accurately represents the popularity of a query by measuring the stability of query frequency over a set of time intervals. Experimental results show that this new feature achieves hit ratios better than those of the previously proposed features.

Semantic Web search is a new application of recent advances in information retrieval (IR), natural language processing, artificial intelligence, and other fields. Our group (Powerset) develops a semantic search engine that aims to answer queries not only by matching keywords, but by actually matching meaning in queries to meaning in Web documents. Compared to typical keyword search, semantic search can pose additional engineering challenges for the back-end and infrastructure designs. Of these, the main challenge addressed in this paper is how to lower query latencies to acceptable, interactive levels. Index-based semantic search can include numerous synonyms, hypernyms, multiple linguistic readings, and other semantic information, both on queries and in the index. In addition, some of the algorithms can be super-linear, such as matching co-references across a document. Consequently, many semantic queries can run significantly slower than the same keyword query. Users, however, have grown to expect Web search engines to provide near-instantaneous results, and a slow search engine could be deemed unusable even if it provides highly relevant results. It is therefore imperative for any search engine to meet its users ’ interactivity expectations, or risk losing them. Our approach to tackle this challenge to exploit data parallelism in slow search queries to reduce their latency in multi-core systems. Although all search engines are designed to exploit parallelism, at the single-node level this

Abstract. In this paper, we describe University of Otago’s participation

Caching of query results is an important mechanism for efficiency and scalability of web search engines. Query results are cached and presented in terms of pages, which typically include 10 results each. In navigational queries, users seek a particular website, which would be typically listed at the top ranks (maybe, first or second) by the search engine, if found. For this type of query, caching and presenting results in the 10-per-page manner may waste cache space and network bandwidth. In this article, we propose nonuniform result page models with varying numbers of results for navigational queries. The experimental results show that our approach reduces the cache miss count by up to 9.17 % (because of better utilization of cache space). Furthermore, bandwidth usage, which is measured in terms of number of snippets sent, is also reduced by 71 % for navigational queries.This means a considerable reduction in the number of transmitted network packets, i.e., a crucial gain especially for mobile-search scenarios. A user study reveals that users easily adapt to the proposed result page model and that the efficiency gains observed in the experiments can be carried over to real-life situations.

Although most large-scale web search engines adopt the standard DRAM-HDD storage hierarchy, the usage of hard disk is greatly limited by its long read latency. On the other hand, NAND Flash memory is 100x faster than hard disk

Précis is a class of cache managers that stores and retrieves the final results of a time consuming operation that may use data from disparate sources. Examples include the results of an aggregate query, serialized instances of a data structure that contains results of different queries and computations used to generate a dynamic HTML page, etc. It is a high throughput, low latency key-value cache manager designed to scale a large data intensive web application. Example systems include memcached and COSt AwaRe (COSAR) cache manager. In this paper, we present the implementation details of COSAR and its variants of LRU, LRU-2, and GreedyDual-Size cache replacement techniques. We compare COSAR with memcached that employs the classical LRU replacement technique, showing the merits of its proposed replacement techniques. 1

Web search engines are facing formidable performance challenges due to data sizes and query loads. The major engines have to process tens of thousands of queries per second over tens of billions of documents. To deal with this heavy workload, such engines employ massively parallel systems consisting of thousands of machines. The significant cost of operating these systems has motivated a lot of recent research into more efficient query processing mechanisms. We investigate a new way to build such high performance IR systems using graphical processing units (GPUs). GPUs were originally designed to accelerate computer graphics applications through massive on-chip parallelism. Recently a number of researchers have studied how to use GPUs for other problem domains such as databases and scientific computing [9, 8, 12]. Our contribution here is to design a basic system architecture for GPU-based high-performance IR, to develop suitable algorithms for subtasks such as inverted list compression, list intersection, and top-k scoring, and to show how to achieve highly efficient query processing on GPUbased systems. Our experimental results for a prototype GPU-based system on 25.2 million web pages shows promising gains in query throughput.

Due to the rapid growth in the size of the web, web search engines are facing enormous performance challenges. The larger engines in particular have to be able to process tens of thousands of queries per second on tens of billions of documents, making query throughput a critical issue. To satisfy this heavy workload, search engines use a variety of performance optimizations including index compression, caching, and early termination. We focus on two techniques, inverted index compression and index caching, which play a crucial rule in web search engines as well as other high-performance information retrieval systems. We perform a comparison and evaluation of several inverted list compression algorithms, including new variants of existing algorithms that have not been studied before. We then evaluate different inverted list caching policies on large query traces, and finally study the possible performance benefits of combining compression and caching. The overall goal of this paper is to provide an updated discussion and evaluation of these two techniques, and to show how to select the best set of approaches and settings depending on parameter such as disk speed and main memory cache size.