This is an old revision of the document!
The performance requirements for VPOP3 depends on many factors, so we can't give specific details for any scenario.
Interpreting Disk I/O is a more complicated task, as there are many factors.
First, look at the “Avg Disk Queue Length”. If this is pretty constantly above 1 or 2 (or the number of striped disks you have in a RAID 0/10 array) then performance is probably being limited by the hard disks. The Disk Queue is the queue of disk operations that Windows is waiting for. In an ideal world, the Disk Queue would always be at zero, because the disks would be processing requests as quickly as Windows makes them, but this is unlikely to happen in the real world.
So, if the Avg Disk Queue Length is higher than expected, the next thing to look at is the “Disk Transfers/sec”. Different types of disks will be able to process transfers at different rates. Sometimes this is called IOPS (I/O operations Per Second), so you may be able to find information on the rated IOPS for your particular disks/controllers.
A typical 15000 RPM SAS (SCSI) drive will be capable of around 175-210 IOPS, a typical 7200 SATA drive will be capable of around 75-100 IOPS.
If you put these into a RAID 0 or RAID 10 array with a decent hardware RAID controller (not the cheap RAID controllers on desktop motherboards), then the IOPS may be spread across the striped drives, so a 4 drive RAID 10 SAS array would be capable of around 350 IOPS
SSDs (Solid State Devices) are capable of much more , typically over 5000 IOPS, even up to 250,000 IOPS.
As an example, one of our VPOP3 hosted service servers was running with a RAID 10 array on SATA drives, with no BBWC. This encountered performance issues, the IOPS was around 170 IOPS, and the Avg Disk Queue was regularly over 3 or 4 for long periods of time. We moved this to a new server with a RAID 10 array on 15000 SAS drives with a BBWC, the IOPS has increased slightly to 230 IOPS, but the Avg Disk Queue is now about 0.2, and the performance is much greater.
If the Avg Disk Queue is high, but the IOPS is not too high for the disks, then the issue may be disk fragmentation. If the disk is highly fragmented, then the disk subsystem will take longer than normal to process requests, so defragmenting the disk is a good thing to try.
For example, 50 users with VPOP3 Enterprise (using IMAP4) or 100 users with VPOP3 Standard (using POP3)
For example, 250 users or more, with VPOP3 Enterprise (using IMAP4), this is the spec of PC we'd be buying:
With this many users, the main limiting factor will be the disk I/O
On a heavily loaded server it may be worth moving the database to a separate physical disk/array from the rest of the PC. For instance, you could have a SATA RAID array for the system and programs, and a (more expensive) SAS RAID array for the database.
Using separate partitions on the same physical disk will slow things down, rather than speed them up, so that is not recommended.