Michael, > > Really... Ok, I just assumed it would help. How could it not? FBSD > > does that good (or bad?) a job with disk processing? Wouldn't the > > combination of fast cpu and ram disk bound processing be faster?
Sven Schuster writes: > I don't know if FreeBSD makes a good job or not (but I would think > so :-) ), I'm more on the linux side. But what I think linux (and > probably FreeBSD too) would do in case of the temporary files > amavisd-new creates is that those files are never written out to > disk due to their short lifetime but instead kept in memory. Yes, this is the general idea. Similar can be achieved by turning off synchronous writes of fs metadata on traditional file systems. Provided that a partition is dedicated to temporary and scratch files and nothing else, it would be safe to do so. The same holds for RAM cache in cheap IDE disks, even without battery backing. Certainly this would be a no-no if other data is stored on such disk, but is perfect for scratch storage, and fast too, not sacrificing system memory. Dedicating system RAM to temporary file system seems like a waste of good resource. The so called 'free' RAM is not really unused on modern operating systems, but is put to good use for caching file system data etc. Dedicating it to a ram file system makes it dedicated to such use only and most of the time such partition would wastefully be mostly empty, except when some exotic mail explodes to large size during decoding. It would make me nervous to have temporary directory live on a too small disk partition. > > I guess you need the ram to increase the innodb log sizes and buffer > > sizes for mysql so that the bayes and awl work faster.? Not much sense > > in increasing $servers beyond 4 * (CPU|CORE|HTT) count is there? > > Increase DNS caching size of using lots of RBL's? > > I don't know how to tune mysql for bayes and awl usage. But for > for amavisd-new, I think it's more memory bound than cpu bound, > so your "equation" of $servers = 4 * cpucore might not be the > best one. A good value for $max_servers should be more estimated > by the amount of memory you have in your system (not taking into > account things like mysql running on the same host etc.) > Those things were probably discussed quite a few times in the > past, so you might want to take a look in the archives. As for > dual core cpus, I think in the end it comes down to something > similar like a smp (multi cpu) system. Indeed a 4*cpu may be on a low side. A factor like 8 or 12 may be more like it, provided the memory size is sufficient. See http://www.ijs.si/software/amavisd/amavisd-new-magdeburg-20050519.pdf for some benchmarks, most of them showing througput vs. $max_servers (on a single processor system). Mark ------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ AMaViS-user mailing list AMaViS-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/amavis-user AMaViS-FAQ:http://www.amavis.org/amavis-faq.php3 AMaViS-HowTos:http://www.amavis.org/howto/
