Spec me a storage solution..

mrbios

mrbios

mrbios

Soldato
Joined
7 Jun 2003
Posts
16,131
Location
Gloucestershire
This is the current setup...

1 x Infortrend Eonstor A16E-G2130-45 with 8 x 2TB SATA drives - This is an iSCSI box with 4 1GB Connections
This has 2 RAID6 arrays, one is used for all the schools storage split in to two partitions - one for home folders and the other for all the other data. The other array is also split into 2 partitions both used for VMware storage (ALL our servers run off this array)

1 x Infortrend Eonstor A12U-G1410 with 8 x 500GB SATA drives - This is a directly connected SCSI box
This is connected directly to an ML350 and is used purely for DFSR (all data on the A16E is replicated to this box) backups, and shadow copies.

All our servers are Server 2008R2, iSCSI and 2008R2 seems to have an issue with shadow copies over iSCSI connected devices hence the way we're doing this, the A12U is incredibly slow and the backups have occasionally have issues thanks to this box.

This is what i want to achieve:
1) Improve backup time/speed while keeping the current system of replicated storage and working shadow copies - The A12U is very old now, willing to scrap it.
2) Split the VM storage off the same box as the student+teacher data on to it's own storage device
3) Give redundancy to the new VM storage device

Now i'm not too clued up on storage devices so I'm after some advice from people here who have some level of expertise in this area, all advice appreciated :)

Environment is a school (as you may have guessed reading the above) so we're not made of money, just need to try and cover the aims.

EDIT: i should probably add that the backup program in use is Backup Exec 2010 R3 with SP1
EDIT2: I should also add that we only need 1.5-2.5TB of data storage and 1-2TB of VM storage space (after raiding) - so 12 or 8 bay units would suffice.

EDIT3: Was looking at the "ESVA E10-2130" for my VMs, not a clue what the cost is of that but it looks good and fault tolerant, plus i find infortrends kit to be quite reliable. http://infortrend.com/uk/products/models/ESVA E10-2130
 
Last edited:
Surely someone here can help with this? It's right up the alley for a lot of you i'd have thought.

I got a price back of £11k on the ESVA E10-2130 and that's without drives, a fair bit out of my price range so I'm now looking at this instead: SBRiDS-S12E-R2140 at a touch under £5k without drives online (may get it cheaper once i've got a few quotes though)

Redundant controller, redundant PSU, redundant cooling, 2 backup batteries for the cache. Looks like it's got itself quite well covered, question is what drives would you lot buy for something like this? I'll be running around 30-40 VMs spread across 5 hosts off it with the prospect of expanding in the future.

The VMs include an SQL server, exchange server and a file server among other things and it supports 6GB/s SAS drives or SATA drives with an added MUX board. Unsure if it's really worth the extra getting SAS drives or if it'd even gain any added benefit from being able to use 6GB/s drives?
 
Thanks growse, good to see HP may provide an option, I'm quite a fan of HP kit (currently all our servers and switches are proliants and procurves) so if they provide something comparable to what i've already looked at i'll have a few things to think about :)

Currently awaiting a price from my supplier for an Eonstor DS S12E-R2140 with all the bells n whistles (remote replication looks like it could provide me an added layer of protection)
 
Teledude said:
NetApp FAS2020 - About 15k :)
Click to expand...

A bit over my budget sadly, though netapp equipment normally is :p

Trying to keep it to under £10k excluding hard disk costs.

I think I'm getting a better idea of what i want now.... The storage local to my ML350 i want to act as a replication partner for both my file store and my VM stores i've decided. I have DFSR sorting the file store replication between the two boxes so this obviously isn't hardware dependant, but i'd like a faster more reliable replacement box to replace the old SCSI eonstor while also being compatible with VM replication to and from the main box and this backup box.

.... Did that make sense? Sadly the VNXe you recommended Vanilla in order to achieve the same thing would require me to purchase 2 of them as they have to be the same for remote replication to work, this appears to be the case with most storage solutions except for a company called nexenta, anyone used anything by them they can recommend?
 
Yamahahahahaha said:
This one stood out for me.
What kind of loads are you seeing on the SQL server? - Can be quite high at times, SQL Services can max out the VMs memory allocation during busier periods (which is 12GB)

How how many mailboxes are you running on the Exchange Server? How large are the mailboxes? Which version of Exchange is it? 2000-2500 mailboxes totaling around 100GB

I can imagine the I/O from these are butchering those 2TB drives. RAID6 already suffers large write penalties, combine this with a relatively slow spindle speed, it's a perfect storm. Consider keeping the Exchange on local disks if you have a spare host. Every last one of our servers is now a VM, each host is part of a cluster enabling vMotion between the hosts so i'd like to aim on keeping that all enclosed within the same system as we have currently

I think growse' approach to the situation would be wisest. Have a bunch of 2.5" disks in RAID10 for the intensive work, then a larger slower array for storage. I think my VMs - excluding the SQL and exchange ones - are vastly more reads than writes so i was thinking of a RAID6 for these and compromising on performance - no one notices the poor performance from exchange luckily - but once moving my VMs off the current A16E i was thinking that one gets hammered with writes throughout lessons by staff and students so i was thinking of a RAID10 for that one
Click to expand...

Replys in bold, thanks for the help so far :)

My primary aim is to gain failover for the VMs AND the storage in a remote location, of which i currently have failover for the storage in a remote location but in an ideal world the storage failover would also supply space for the VM failover (doesn't matter if it was dog slow so long as it worked while the main unit was repaired for example) Performance is a secondary objective.
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom