Setting on pool size could be a little tricky. Planning pool size before the implementation is good but it depends on whether we can predict the actual throughput of a business host and the performance of it. As for my own experience, the more effective way to use pool size is to use it as a performance tuning approach. Say, we deploy a business host online and perform test run for several rounds with actual or estimated loads. During the test, we'll need to observe the queue. For example, if a queue is increasing much faster than the others, it usually means the underlining business host is a performance bottleneck and we'll need to either optimize the business host or simply increase it's pool size as long as there are spare CPU cores to be used.
Before a project goes live we usually try it out for a period of time, during which we gradually determine the pool size of each business host to achieve the required performance.
Set your pool size according to the actual situation of the business, but the value should not be greater than the number of CPU cores of your server
Documentation here : https://docs.intersystems.com/irislatest/csp/docbook/DocBook.UI.Page.cls...
It also explains the actor pools size and about the considerations and trade-off.
Hope this helps.
Hi, Dongsheng,
Setting on pool size could be a little tricky.
Planning pool size before the implementation is good but it depends on whether we can predict the actual throughput of a business host and the performance of it.
As for my own experience, the more effective way to use pool size is to use it as a performance tuning approach.
Say, we deploy a business host online and perform test run for several rounds with actual or estimated loads. During the test, we'll need to observe the queue. For example, if a queue is increasing much faster than the others, it usually means the underlining business host is a performance bottleneck and we'll need to either optimize the business host or simply increase it's pool size as long as there are spare CPU cores to be used.
Before a project goes live we usually try it out for a period of time, during which we gradually determine the pool size of each business host to achieve the required performance.
__________________________________________________________________________________
pool size的设置并不会对每一个应用都一样。
在实施之前规划好池大小是好的,但这要看我们是否能预测business host的吞吐量和它的性能。
就我自己的经验而言,将pool size作为一种性能调整的方法比较有用。
比如说,我们在线部署一个business host,在实际或预估负载的情况下,进行几轮测试运行。在测试过程中,我们需要观察队列的情况。例如,如果一个队列的增长速度比其他队列快得多,通常意味着对应的business host是一个性能瓶颈,我们需要对它进行优化,或者,只要有空余的CPU核心可以使用,增加它的pool size通常就会带来可观的性能提升。
在项目上线之前我们通常会试运行一段时间,期间逐步确定每一个business host的pool size以达到所需的性能。