When a client makes a cachecash-curl request toward a publisher, spikes in CPU usage occur at a node that runs a cache in use. The gRPC service of ClientCache has outliners at handling time, especially in the 99th and 90th percentile of requests. It is OK at the 50th and 25th percentile of requests. However, in the case of concurrent cachecash-curl requests made toward a publisher, e.g., 200 clients at once, the gRPC handling time reaches 10 seconds.
When a client makes a
cachecash-curlrequest toward a publisher, spikes in CPU usage occur at a node that runs a cache in use. The gRPC service ofClientCachehas outliners at handling time, especially in the 99th and 90th percentile of requests. It is OK at the 50th and 25th percentile of requests. However, in the case of concurrentcachecash-curlrequests made toward a publisher, e.g., 200 clients at once, the gRPC handling time reaches 10 seconds.