I wrote a web spider with HttpClient API and run it to crawl some http/2 websites. It worked well at beginning, but then was too much slower. From jstack, many HttpClient-Workers were trying to allocate HeapByteBuffer.

Could we do any improvement on it?