Monitoring the Microsoft FAST Search Server 2010 for SharePoint

eG Enterprise embeds a specialized Microsoft Fast Search monitoring model for monitoring the performance of the FAST Search Server 2010 for SharePoint.

Layer of Fast Search

Figure 3 : The layer model of the Microsoft Fast Search server

Each layer of Figure 1 above is mapped to a wide variety of tests that collect a wealth of performance information related to the FAST Search server, Using these metrics, administrators can find quick and accurate answers to the following performance queries:

  • How much time elapsed between item initialization on the server and the completion of indexing? Was too much time spent? If so, where did the slowdown occur? - did it occur before initialized items were placed for submission? did it occur after submitting batches for processing - if so, was it at the first attempt to submit? or was it when waiting for an acknowledgment for submission?
  • Did too many batches fail even before initializing?
  • Which crawl collection is currently crawling too many web sites/web links?
  • Is any content distributor taking too much time to process a batch? If so, which one is it?
  • Are any document processors invalid? If so, which content distributor are they registered with?
  • Which content distributor is not utilizing its document processors effectively?
  • What is the current state of each of the document processors on the server?
  • Is the workload of the web analyzer evenly distributed across all its link processing components/lookup databases?
  • Are there adequate sites and URIs in the lookup database to service link information requests?
  • Is any API queue overloaded with pending operations? Which indexer is such a queue attached to?
  • What is the load on each indexer partition? How well does every indexer partition index the items on it? Is any indexer partition experience processing delays?
  • Has a handle leak been detected in any query dispatcher?
  • Have too many queries to any query dispatcher been enqueued? If so, which dispatcher is it?
  • Have any queries to a QR server failed?
  • Is any QR server experiencing latencies when processing queries?
  • Did any query timeouts occur in the indexer?
  • How much time does the indexer service take to process a single query? Is it too high?

The answers to these queries will enable administrators to proactively detect performance bottlenecks (if any).