need to purchase a server for virtualization. confused about 10K RPM disks and near-line SAS 7.2k RPM.
do not require any redundancy for data hence RAID0 will do.
choose disks 10K SAS for OS array and slower 7.2 K drives for data, does this config require two controllers? (PERC 6i)
need approx 2 TB of data space, am confused by number of drives that a controller can handle and the form factor (2.5 inch vs 3.5 inch), since real estate is no problem rack or tower is fine
experts please answer.
thanks in advance
updates: mid-range performance is acceptable, the server hosts approx 25+ VM's(windows only) and may have few SQL Server DB…..no heavy i/o, but occasionally.
Best Answer
lets separate the issues:
SAS vs NL-SAS is very simple. While SAS drives are the proper SAS (i.e. SCSI drives with a serial connector and access protocol), NL-SAS is really a SATA disk with a SAS connector and access protocol. The advantage of NL SAS vs SATA is in the connector and protocol, because while you can use SATA disks connected to SAS controllers, you will suffer performance hits in addition to the disk slowness, because there will always be a protocol conversion between SCSI (the protocol SAS uses) and ATA (the protocol IDE and SATA use) So in the end, SAS vs NL-SAS is just the matter of RPMs, while SAS vs SATA was RPMs plus overhead
3.5" vs 2.5" is even easier - you choose between larger (in size, not capacity) and cheaper drives, and smaller and more expensive drives. The caveat of the larger and cheaper ones is the fact that you can fit much less of those on a backplane. I have two IBM 1U pizzaboxes in my server room, one can hold up to two 3.5" drives, and the other - up to 6 2.5" drives. This can not only give you more capacity, but can bring you up to a higher spindle count, which is a major factor when you need disk performance (and with VMs you probably will)
PERC6i will be able to handle whatever you put in that server, just create several raid arrays.
I would strongly advise against raid-0 no matter what the requirements are. It is simply too risky, especially if in case of a failure you'll end up restoring or recreating several VMs, instead of a single physical machine