I ran a set of preliminary tests using 32 concurrent searches. This maintained a workable system, without failing. For some reason, while I was running the tests, I continued to get better response than the siege test was reporting. I'm not sure how that could be.
I also verified that with 32 concurrent tests, and java VM of only 1Gb of heap space will run out of memory. I therefore updated this to 4Gb. I also noted that the elasticsearch instance will restart properly if it does run out of space, so we are somewhat protected against DOS attacks.
I ran a set of preliminary tests using 32 concurrent searches. This maintained a workable system, without failing. For some reason, while I was running the tests, I continued to get better response than the siege test was reporting. I'm not sure how that could be.
I also verified that with 32 concurrent tests, and java VM of only 1Gb of heap space will run out of memory. I therefore updated this to 4Gb. I also noted that the elasticsearch instance will restart properly if it does run out of space, so we are somewhat protected against DOS attacks.