[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

[Linux-cluster] Low cost storage for clusters (was: Setting up a GFS cluster)

Hello again Brendan,

We have recently deployed the following environment for a customer:
* Dell EMC AX-100 storage through fibre channel connections;
* 2 DELL PE-2800 servers with 1 QLogic HBA each, direct attached to the storage;
* Around 400 GB of raw disk space on the AX100's SATA disks;
* Red Hat Enterprise Linux v3 Update 7 (this was before the release of U8);
* Red Hat Cluster Suite HA solution installed.

This environment has yet to prove to be effective, but when it goes to production, it will support about 100 concurrent users accessing Dataflex files and applications, both through Telnet sessions and through file sharing using Samba (Visual Dataflex running on top of Windows machines).

I believe the performance will be quite acceptable, even though the storage uses SATA disks.

I've seen some storage solutions using even ATA HDDs with very good performance. On such example is Apple's X Server RAID storage, which uses ATA (not SATA nor SCSI) disks. It is even certified for Red Hat Enterprise Linux (and for RHCS too, I believe). It uses copper fibre channel connections, though.

In my opinion the secret with these cheaper solutions is on storage processors. If they hold a reasonable amount of cache memory, they can compensate eventual latencies with the hard drives.

We have also had a good degree of success using Dell EMC's CX-300 and CX-500, using fibre channel connections.

Finally, I think iSCSI is a good thing if you don't need high performance I/O. I previously suggested you iSCSI in place of Dell's PV220S (which lacks write cache in Cluster mode) because an iSCSI solution with full cache functionality would give you similar or better performance than the PV220S solution.

With iSCSI, we haven't seen very high performance I/O rates. In one case, a customer employed an IBM storage (don't remember the model), which was accessed via iSCSI using a Cisco iSCSI-to-fibrechannel switch. Even with trunking enabled on the iSCSI NICs on the servers, performance was not outstanding as the storage itself would give if the servers were direct attached or connected through an SAN. We used in this case standard NICs and Linux's traditional iscsi-initiator-utils (a sourceforge project), so we could have better performance if we employed dedicated iSCSI NICs (such as the QLogic models).

Maybe someone will disagree with me about a "not so good performance with iSCSI", but we (our company) see iSCSI as a good solution for lowering costs when you have a reasonable amount of servers and don't want to invest on a full redundant SAN solution (a QLogic HBA card costs around USD 2000.00 here in Brazil, for example).

I hope this gives you some extra information about storage choices for your cluster deployments, ok?

Best regards,


Brendan Heading escreveu:
Celso K. Webber wrote:

Maybe a Dell|EMC AX-100 using iSCSI could a better choice with a not so high price tag.

Sorry for the long message, I believe this information can be useful to others.


Please don't apologize, your reply was very informative and you've probably just saved me a big pile of cash :) I've never been happy with the PERC RAID controller that I currently have, and as part of moving to the cluster I intend to use Adaptec stuff all the way. However it sounds like that won't fly with the PowerVault in any case.

I saw the EMC AX-100 but I've got one significant problem with it - the disk storage inside is SATA. The server that I'm dealing with is essentially a build server used to do regular parallel builds, and it may be simultaneously in use at any one time by up to 20-30 users. I've been looking at iSCSI but I find it very strange that nobody seems to sell iSCSI-to-SCSI boxes.

Do you think an SATA-based array is likely to hold up under these conditions ?

A less important reason here is that I've got a bunch of 15000rpm 150GB & 300GB SCSI320 disks connected to the Dell boxen at the moment that I was hoping to reuse. It would be nice to reuse them, but I don't have to.

Thanks again,


Linux-cluster mailing list
Linux-cluster redhat com

*Celso Kopp Webber*

celso webbertek com br <mailto:celso webbertek com br>

*Webbertek - Opensource Knowledge*
(41) 8813-1919
(41) 3284-3035

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]