RSS
Email
Print
ShareStartups Bring Unified Storage to SMBs
Posted by Howard Marks on July 7, 2009
Every few years we hit a "Great Minds Think Alike" moment where products based on the same idea appear within months of each other. In Hollywood this gives us "Dante's Peak" and "Volcano" or "Deep Impact" and "Armageddon". Here in Byte and Switch's storage world we had Lefthand, Intransa and Equallogic all releasing iSCSI clusters and now we have startups Prana and Scale Computing both announcing scale out unified storage systems for the SMB market.
Now like a movie where a comet impact is about to create an extinction level event unless Tommy Lee Jones or Bruce Willis blows it up scale out NAS isn't a new idea. After all Isilon, Ibrix and HP with their ExDS9100 have been building scale out NAS systems for oil companies, movie studios and other deep pocket customers for years.
Both vendors have 1u systems that run a Linux variant to provide CIFS, NFS and iSCSI access to storage that's managed through a web interface. Both also support clustering so a group of nodes can support a single NAS name space.
Scale Computing uses SuperMicro's solid, if not cutting edge, 1U servers with 4 500GB or 1TB SATA drives as the basis for their system. Each node has 1 gigabit Ethernet ports, 1 faces the user network and the other a private inter-node communications net. Rather than using conventional RAID within each node Scale's OS stripes and mirrors data so there's two copies of any data block on two different nodes which allows system with at least 3 nodes to survive drive or node failures.
Comment by Steve Carter on July 8, 2009 2:58 PM
Howard, I do need to make a minor correction: disks in both Pranah's control units and expansion units ARE hot swappable, along with our controllers, power supplies, and fans.
Reply to this comment
Comment by Howard Marks on July 8, 2009 3:08 PM
I stand corrected. Hot swappable but you have to open the server as they're not accessible from the front.
Reply to this comment
Comment by Dave Walstad on July 8, 2009 3:59 PM
It is true the that cover of the unit needs to be opened to access the disks, but the system has been designed to do so. There are cable guides to maintain cable connection integrity and we have Intellectual Property that provides shock and vibration protection that has been proven in operational military environments.
Dave Walstad
VP Sales and Marketing
Pranah Storage Technologies
www.pranah.com
www.twitter.com/pranah
Reply to this comment
Comment by Mike Young on July 8, 2009 6:07 PM
The only immediate downside to this, in the case of Scale is the need for more replication. In this day of green storage, unnecessarily replicating data should be avoided as a general rule. It's one thing to replicate data between a local site and a remote one. That can at least be justified as part of a disaster recovery scheme. But replicating between nodes or drives between nodes isn't very efficient. And this has less to do with the cost of a disk drive than it has to do with the consumption of energy. Sure, we're not talking major economic disaster, but it does add up if you're successful.
I think you still have to ask yourself what it is you're trying to solve. 210MB/s for a 3-node cluster is hardly worth writing home about. We're able to sustain 250MB/s on the same platform, but only a single server. And we can sustain close to 2GB/s in a 3U form-factor. All of our data is seamlessly replicated to our cloud service. Any server, desktop, or client can access that data in block or file mode if the local appliance fails. If you're going to design around a disaster or critical failure, then really go for it. And our appliances retail as low as $1/GB. And we've been doing this stuff longer.
If you say the desired goal is non-stop local access, then their approach is a valid one. But again, this can still be solved with better efficiency and performance. Oh yeah, and with fewer components, hence less energy consumption. Just saying.
I can't comment much on Pranah's stuff, but I do hope to learn more about their solutions and approaches. Shock and vibe are always a concern with the faster disks. And heat can be quite a challenge too.
As far as your analogy goes, you're right on. I can remember when Netapp and Snap introduced iSCSI for free at the same time. This was back when others were trying to charge as much as $300/port. So I'm happy to see these two companies going at it. I'm sure some good stuff will come from these clusters.
Mike Young
CEO, Cachengo
http://cachengo.com
Reply to this comment