Design issues for nntp storage manager
davidsen at tmr.com
Tue May 29 21:13:05 UTC 2001
In article <20010525113207.A14463 at arun.cam.uk.internal>,
Russell Vincent <russellv at uk.uu.net> wrote:
| It would probably be reasonably easy to pre-load the cache (with
| a bit of extra code), but I have never worked out the need to do
| pre-loading of caches. The whole point of a cache is that the first
| person loads the article into the cache and the next articles are
| found there. The spool retrieval time is pretty negligable on a
| well built network anyway. I am aware of some installations that
| just turn off caching altogther to reduce disk IO on the reader
| frontends. They claim it works very well in that mode.
If cost is no object, that's true. But if you are paying for
bandwidth, you don't hang your remote servers off dedicated OC24's, they
have limited bandwidth back to the server, like maybe only 10Mbit or so.
Now if you isolate groups which have high readership, it's nice to
prefeed those because a regional server does follow the local day/night
cycle, even if the master server is more evenly loaded.
By preloading articles you know you will need you can reduce spikes as
someone hits a newsgroup the first time. Well designed sometimes
involves diddling the application to be more cost effective.
bill davidsen <davidsen at tmr.com>
CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.
More information about the inn-workers