Excessive request queuing is another common cause of performance problems. The Request Log Analyzer can be used to detect the presence of excessive request queuing and can report on the performance of queued requests vs. requests that encountered no queue.

Request queuing is, in and of itself, not necessarily a bad thing. MDEX Engines are often more efficient at processing a small number of simultaneous requests quickly and then moving on to process requests that have been waiting in queue. In this model, because each individual request is processed quickly, requests are only in the queue for a very short time, and overall performance is good.

When the request queue gets very long, or when requests have to wait a long time in queue, the queue is a problem. In this situation, additional engine threads can help if the server has enough available resources (CPU, RAM, disk). Additional MDEX Engines in a load-balanced configuration will also help.

Note that the response differential metric - the difference between round-trip response time and engine-only processing time - includes time spent in both the request queue (waiting to be processed) and the response queue (waiting to be written back to the client). It is not possible to determine from the request logs exactly how long requests spend in the request queue alone.


Copyright © Legal Notices