我们想在生产环境中使用ISTIO进行工作量。这里记录了所需的CPU资源：
文档使用负载测试方案，该方案由每秒1000个请求组成，每秒通过Envoy Proxies（包括其他参数）。我们确实有较少的请求，当然，重要的是优化CPU资源以优化解决方案的成本。
因此，问题是：我们可以将CPU分配减少到代理代理与在情况下的参数减少成比例吗？ CPU分配的最低阈值？我在文档中什么也没有发现。当然，我们想保留代理添加合理的小延迟（类似于已记录的方案中的2.65毫秒）。
一旦运行ISTIO，我们就可以为解决方案执行一些负载测试；但是，ISTIO消耗的CPU额外资源将是决定是否使用它的相关点之一。因此，对此一些最初的想法将是很棒的。
We would like to use Istio for our workload in a production environment. The required CPU resources are documented here: https://istio.io/latest/docs/ops/deployment/performance-and-scalability/
The documentation uses a load test scenario consisting of 1000 requests per second going through envoy proxies (among other parameters). We do have a much lower number of requests, and of course it is important to optimize CPU resources in order to optimize the cost of our solution.
So, the question is: can we reduce the CPU assignment to the envoy proxies proportionally to the reduction of parameters in our scenario? There is a minimum threshold for CPU assignment? I have found nothing about this in the documentation. Of course, we would like to keep a reasonable small latency added by the proxy (similar to the 2.65 ms in the documented scenario).
We can perform some load testing for our solution once we have Istio running; but the CPU additional resources consumed by Istio will be one of the relevant points in deciding whether to go with it or not. So some initial idea about this would be wonderful.
发布评论