[ Date Index ] [ Thread Index ] [ <= Previous by date / thread ] [ Next by date / thread => ]
On Thu, 11 Jun 2009, Austin Gossmeyer wrote: > I have 3x17431M and 2x34857M drives on a Ubuntu 9.04 system. I am > wondering the best way to split them up. At the moment Ubuntu is > installed on one of the 17431M drives and I was thinking of using one > drive for squid logs and another for dansguardian logs. I have formatted > the other drives but I am not sure how to get them in use. What space > should I give the log files? Thanks in advance. In the olden days we'd look to optimise spindles and data paths, and sometimes even locations on disks for partitions - these days, it seems that not many people do this. I consider it a dying art myself, but I'm of the old way of doing things ... So if you want performance, you need to look at the ways the drives are connected. You don't say what they are - SCSI, SATA, IDE, or ... If they're IDE, then it's best to avoid 2 drives on the same bit of ribbon cable. Modern controllers will give you a performance gain over older ones when accessing both drives concurrently, but I have seen one faulty IDE drive block access to a 2nd on the same ribbon cable. (Drive was fine, controller board smoked) The same is true of SCSI, but to a lesser extent due to the way the bus is designed, and thet're generally better engineered. SATA is less prone to this as you only have one drive per cable (but I've seen a mobo controller fail, rendering 2 drives unusable )-: I doubt the drives are SATA due to their small size though (18G and 35GB) I'm suspecting they're older SCSI drives as they were popular sizes. So if they're all on one chain, then look to separate tasks on different spindles. You'll only get a real advantage if you're using fast SCSI though. So 5 drives - OS on one drive, squid cache on another, suqid logs on another and DG logs on a 4th, leaving the 5th free... However you might want to consider RAID or mirroring. If you want speed, then combine the 2 x 35GB drives via RAID-0, giving you one big 70GB drive at near double the speed of a single drive. However if one drive fails, you lose the lot. Maybe not that big a disaster for a squid cache - just remove the bad drive, re-format the remaining good drive, and start again... but if you want redundancy, then use RAID-1 which will withstand a single drive failure - squid data on the RAID drive, OS on one, logs on the other 2. Too many choices though - and without knowing your aims, it's hard to know what to suggest. You also don't say what the Internet b/w is... If it's ADSL or cable, (or SWGFL) then RAID-0 might be overkill as you'll still be network b/w bound... If Janet speeds then you might find the server is the bottleneck! Another thing to consider though - power consumption - who's paying the electric bill for those 5 (old) spinning devices which will run 24x7... You may well be better off with 2 new "green" drives in the long-run. I know I was when I had multiple old SCSI spinny things running at home.. Anyway - mounting them - just create top-level directories like /squid-data /squid-logs /dg-logs and mount the drives there and adjust fstab to mount them at boot time. You will need to fiddle with the squid and DG configuration files to place the data/logs there though. Good luck! Gordon -- The Mailing List for the Devon & Cornwall LUG http://mailman.dclug.org.uk/listinfo/list FAQ: http://www.dcglug.org.uk/linux_adm/list-faq.html