Stas Bekman wrote: > So there are two different issues here: > > 1. patching the scoreboard to collect and store this extra info > > 2. patching mod_status to use this extra info
true > since mod_status in the extended mode is already doing a lot of work, > how much of an extra overhead will the port/ip info will add? A hard question to answer without benchmarking. Where does the scoreboard growth stop? I assume we would support IPv6, so this would mean increasing the size of worker_score by the size of a v6 sockaddr struct, which is 24-28 bytes. The worker_score is 208 bytes now on my Linux box. It's a Pentium II with a 32 byte cache line size. This means that each request would touch a minimum of 8 cache lines rather than 7 cache lines for the worker_score, i.e., it would cause at least one additional cache miss per request on my box. Greg
