On Mon, Sep 10, 2012 at 21:05 UTC, <[email protected]> wrote: > I upgraded one of my servers to the development version (4.2.7p304), mainly > so I could play with the new monitoring mechanism. I gather that the > default settings for mru are primarily intended to preserve functionality > for rate limiting, not for more in-depth statistical analysis, otherwise I'm > not certain why maxage would be set at 64 seconds. > > Is there any reason not to use the mru functionality for more in-depth > client monitoring? Would the extra CPU/RAM hit from maintaining a large > database negatively impact time accuracy, or would the only drawback be the > increased consumption of system resources?
The only drawback should be the increased memory use -- the cost of maintaining the MRU list should be close to constant regardless of the list's size. > I've gone with the following settings for now, though I'll certainly scale > them back if they will negatively impact the accuracy I'm handing out to my > clients: > > mru initmem 8 > mru incmem 8 > mru maxmem 16384 > mru mindepth 1000 > mru maxage 86400 > > Another question: Approximately how much memory is consumed by a single mru > entry on a 64bit system? Use "ntpq -c monstats" and divide maximum addresses by maximum kilobytes to get the number of entries per kilobyte of memory. Divide 1024 by that result to get the number of bytes per entry. That math shows 72 bytes per entry on a 32-bit system. Cheers, Dave Hart _______________________________________________ pool mailing list [email protected] http://lists.ntp.org/listinfo/pool
