The future setup will be a VMware infrastructure with 5-6 systems running ESX server
3-4 systems with different webservers and with the 1-2 as sql server.
I'm looking for the right ISCSI box for this setup, which provides:
- Fast ISCSI data acces for the sql server (lots of I/O)
(I don't worry about the webserver applications)
- reliable
- disaster recovery (i understand there aren't clustering possibilities)
the questions:
1.Does the following setup work in the way that I think?
one storage 16 bay chassis with 2 raid controllers (unified storage SATA/SAS). 1 with
raid 6 SAS disks for fast acces for sql and 1 with raid 6 SATA disks for "normal" data.
2.does a quad core Xenon processor with 8Mb cache give me significantly more
performance?
3.How many ethernet adapters are recommended?
Right now I wan't to use 4 Gb adapters, 2 for faill over operation (is this supported?)
4.I can't find information about open-e making use of HBA with TOE cards when placed in
the storage sever (the open-e device). Is this correct? and why isn't it supported?
I would expect a performance boost
5.Can I replicate a snapshot to another open-e server (low end raid 5 storage box)
directly from the open-e software? (no filesystem copy and so on, just keep it block
based)
And is this possible without an extra server/system?
6.Is there performance data available for open-e (prefer preferrably from a independent
source)
1) Yes you can set up your RAID configuration like this and have 2 separate Volume Groups to support your application needs.
So VG0 (RAID 6 for SAS/SQL) and VG1 (RAID 6 for SATA/DATA)
2) When utilizing many Snapshots and other functions like Volume Replication, then yes. If not then this could be done with dual core/4Mb.
3) Yes, you can have 2 using Balance RR and if using Volume Replication the maybe bond with 802.3ad or Balance RR or use more ports for you bond.
Download our free 30 day trial cd to test.
4) We do not support TOE cards; we have had poor performance values. But who knows maybe next year this might improve - but no development has been in place.
5) Not really as you would need another iSCSI-R3, then use Volume Replication for the mirror Target from the Source Server. On the Destination Server you would create a Snapshot volume and set a task. Then use one of the MS Initiators to back up from the Snapshot.
But just to back up directly onto another volume, no. We are coming out with iSCSI backup functions in October of this year if you can wait until then.
6) Very soon with Intel also search on the Forum to find speeds and performance.
Download our free 30 day trial cd to test.
Thanks for the quick reply.
May I inquire when the intel performance documents are available?
The test cd is a chicken and egg problem. If we decide to use the open-e solution we will buy a nice configuration for it. But we don't have a high performance storage server on the shelf for testing.
So I really want(and need) to see the performance documents so I know what to expect.
Then I can request the required budget ;-)
And more questions:
- Is the upgrade from the current version to the version in oktober without any upgrade cost?
- Why can i have only one active snapshot with the 64 bit version per LV. (and is this gonna change in the next release(s))
- Why should I go for the 64 bit instead of the 32 bits?
I understand, but Intel has not released this until next month, so we do not have this yet.
Our updates to all products are free. Not sure about what you mean concerning only one Snapshot??? You can have 10 per Logical Volume (meaning you can have several Logical Volumes). You do not have to use the 64bit; the only benefit is if you are needing more then 16TB per single Logical Volume.
Next month is only a few days away. I can wait ;-)
The 'one snapshot' remark was concerned the relaese notes you posted in an other thread:
I will only paste the line with the known issues concerning my remark
Open-E iSCSI-R3 Enterprise Release Highlights:
* 2007-07-03 ver. 1.21 release
KNOW ISSUES:
· When using 64bit system mode there is limitation to 1 active snapshot per LV and 10
active snapshots overall.
By the way, I did download the demo because I already can try the basic performance on a older dell server and try the load balancing and failover.
(and of course see how the rest works)
I see where the confusion is, I was looking at the new release 2.30. Correct in 2.21 this was an issue. But I have not confirmed this for myself with new version.
i have a system i built for similar purposes.. SQL 2005 for a sharepoint cluster - iometer is reporting about 110mb/s both read/write but when i run iozone (iozone -Razb c:\fileout.xls -f T:\myscsivoltest.lo1) things only look good on larger writes.
still fumbling through options in open-e. im pretty sure that everything else is tweaked and all that remains is the ietd.conf and IP options left to do.
PS - when benchmarking my disk locally Im getting 592mb/s read and 397mb/s write - enough to saturate 2 gig ethernet links.