Optimizing search performance
Martin White · Facet eBooks · 2018
In this chapter: ■ The parameters of search engine performance ■ Search engine tuning ■ Using web analytics software ■ Data privacy issues Introduction The previous chapter outlined the elements that go to make up a search application, whether for a website or an enterprise. A major challenge for the organization is then to make sure that the entire system works as effectively as possible. For better or worse, Google has set some standards for search performance, but these are mainly in terms of systems performance and the time taken to run a query against an index of the spidered web. Users of websites in particular, but also users of enterprise search applications, use Google as a benchmark when judging search performance. Optimizing search performance requires attention to be paid to three different but related areas: ■ the performance of the search system ■ the usability of the search application ■ the extent to which the search provides the information being sought. Search system performance Some search vendors display the time taken to undertake a search as a metric of search performance. Obviously system response speed is important, but it is doubtful whether the user actually cares how long it took so long as it took no longer than they expected, even though they would not be able to specify in tenths of a second how long that is. Among the factors that need to be tracked on a regular basis are: ■ average response time for different types of query (which requires a definition of query types) ■ average time taken to retrieve and then present individual documents or groups of documents from each server ■ time taken to update the indexes (measured in megabytes per hour). The initial indexing can be quite a prolonged exercise. A representative time might be of the order of 10 Gb/hour. Subsequent updates to this will be much quicker. In an enterprise application the system response times for queries and document presentation are determined by the organization's server/network architecture. This may vary with the time of day, the day of the week and many other factors, because certainly as far as the network is concerned the latency will be affected by other activities taking place over it.