royle@iuvax.cs.indiana.edu (Keenan Royle) (11/22/90)
My configuration: Sparcstation 1+, 12 Meg RAM, 30 Meg swap. Cnews, nntp, 3 full feeds, 12 partial. My in.coming spool is running terribly behind, can I make runnews process articles faster? I observed that it was processing 50 articles per minute, always. Looked through runnews and tried changing the 50 in the sed script to 100 and to 200. Still processed 50 articles per minute. ( 100 every 2 minutes or 200 every 4 minutes, you get the idea ) Figured it must be a limit somewhere in relaynews, but I grep'ed for every pattern I could think of and could not find any 50s or sleep()s. Are there any parameters I can set to make it process more news faster? I'm currently 9000 articles behind after only 14 hours and I had to kill off the nntpds. HELP! -- Keenan Royle royle@cs.indiana.edu postmaster@cs.indiana.edu royle@iubacs.bitnet
henry@zoo.toronto.edu (Henry Spencer) (11/22/90)
In article <73545@iuvax.cs.indiana.edu> royle@iuvax.cs.indiana.edu (Keenan Royle) writes: >My in.coming spool is running terribly behind, can I make >runnews process articles faster? I assume you mean newsrun? >I observed that it was processing 50 articles per minute, always. >Looked through runnews and tried changing the 50 in the sed script >to 100 and to 200. Still processed 50 articles per minute. There is no deliberate rate limiting; that would be a silly thing to do. Are you perhaps using dbm rather than dbz? That's a mistake, dbz is much faster. (How old is your C News?) Is your NNTP dropping articles into in.coming one file per article, instead of in batches? If so, look no further for the performance problem. Throughput is enormously higher for batches. -- "I'm not sure it's possible | Henry Spencer at U of Toronto Zoology to explain how X works." | henry@zoo.toronto.edu utzoo!henry