On Sun, 2006-10-29 at 12:19, James Ward wrote: > In an effort to reduce the time spent doing the BackupPC_nightly, I > upped the number done in parallel from 2 to 8. Now the system is 99% > wait. What is the best mix, half wait and half CPU bound?
Disk head motion is bound to be the bottleneck here and that's pretty much single threaded. The only large improvements could come from sorting the operations in inode order (didn't someone do that some time ago?) or spreading the disk operations over more heads with raid0 or LVM. If you are using raid5, that's probably the main problem. > I'm concerned about performance because I am backing up over 200 > machines and so far, I'm only getting around to each machine about > once a week. Any performance tuning ideas you can give me would be > appreciated. As I said before, I also eliminated the blackout window. The other thing that might be improved is the backup speed. Rsync can be slow even for incrementals if the runs span huge numbers of files. Splitting them into separate filesystem or directory runs might help. If that's not possible, adding RAM to the server might help. If you have the bandwidth, tar might be faster on the incrementals - and you can stagger the fulls or try to get them on weekends. -- Les Mikesell [EMAIL PROTECTED] ------------------------------------------------------------------------- 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 _______________________________________________ BackupPC-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/backuppc-users http://backuppc.sourceforge.net/
