[Bug 1091] pmapi extension for regularly scheduled fetches

[bugzilla-daemon@xxxxxxxxxxx]

Comment # 1 on bug 1091 from Frank Ch. Eigler

As a random additional data point, a cross-network pmlogger instance
running with a default pmlogconf indicates, under strace:
- apprx. 900 bytes sent per fetch request
- a delay of 60 milliseconds awaiting a response
- a write of the result to disk within 30 microseconds

At the pmcd end, under strace:
- those 900 bytes are received
- after 300 microseconds, the PMDAs start collecting data
  (doing too much work, as noted at bug #1067 etc.)
- sending overall response of some 9000 bytes back 56 milliseconds after
  receiving request
  (without pmcd running under strace, it is considerably faster)

... an additional 4ish milliseconds remain unaccounted for in terms of
network/kernel latencies

---

      You are receiving this mail because:

• You are on the CC list for the bug.