StorageTek Announces Linux Based Storage Solution

njcoder writes "TechWorld has a report on StorageTek's new content-based disk and tape storage product. 'The software runs on a cut-down Linux kernel on dual Xeon nodes in a meshed network. The content-addressable store (CAS) makes the system more efficient than standard CIFS and NFS access, but supports all three.' The disk based storage uses SATA drives."

Comments

Inverse for SHA 256?

[MarkByers]

IntelliStore uses the NIST-developed SHA 256 algorithm

A stored object can have its hash re-created and if it differs from the original hash then the data has changed. It can be regenerated from the hash.

It seems that they have developed in inverse function for the SHA 256 algorithm. Nice!

Seriously though, I would be interested to see how this works. Some sort of brute force attack? Or do they store some extra error-correction bits?

Not Solaris?

[Anonymous Coward]

I thought Sun had taken them over!

Re:Not Solaris?

[njcoder]

The thing uses a stripped down version of the linux kernel. It doesn't need the whole OS. Linux is better suited to things like that because you can custimize it more. Maybe when OpenSolaris comes out next week people can start doing that with the Solaris kernel.

In any case, I don't think the deal has been finalized yet and contrary to what most people say on here, Sun isn't anti Linux. At least not to the extent some people portray them to be. They haven't really said anything worse than what other unix vendors have said, including IBM and HP. Sun uses linux in the support processor for their opteron servers, they sell and support RedHat and Suse Linux on their x86 servers, they have a linux desktop distro, they provide a lot of their software for the linux platform, they contribute to the linux kernel as well as other important oss projects, etc.

Re:Not Solaris?

[griasr]

in such cases id always stick to linux. who knows what sun is about to do in the future?

Re:Not Solaris?

[turgid]

who knows what sun is about to do in the future?

Well, Sun certainly doesn't.

/me ducks.

Wow, that's expensive.. lemme do ~10 times better.

[freality]

"A 4TB starter IntelliStore costs around $75,000 (£42,000) with each additional terabyte costing $9,000 (£5,000) A set of additional compliance functions adds $15,000 to the cost (£8,400)."

Better:

- Storage: SuperMicro 15-bay disk array with 2 RAID-core controllers (2 RAIDs), SCSI-attached, for a total raw space of 12 disks (-1 for each RAID, -1 for OS). 2 fast processors and a bunch of RAM. Mount all data under /mnt filesystem.
Size: 12*400GB=4.8TB
Approximate cost: $5-10K depending on disks, processors and RAM.

- File-server: mathopd static web-server pointing at /mnt filesystem. 10-20k transactions per second for small files. Fast transfer of large files.

- Database: Nutch open-source search engine capable of indexing 40M pages per 1GB/RAM. Like the article says "millions of objects" now, "billions in the future". Point nutch at mathopd and watch your "content-based" storage come online. You can even get an RSS feed of newly added items.

- Offline: Dell PowerVault 8 tape changer, SCSI attached + mtx for automatic tape changing. + 5k

Now, a web search engine isn't a database, at least not off the shelf. But with this configuration you can afford a l33t programmer for half a year and still come in under the price of the StorageTek solution. Plus, once you've customized it, your capacity upgrades are much cheaper. And I bet it's faster. Dunno though, I bet you can't evaluate most of the relevant system parts from StorageTek before you make a buy decision (unlike the system above). ;)

Content Addressed Storage

[Intron]

Data is accessed by content rather than a hierarchical filesystem based on unique filenames. Has anyone used a system like this and can explain advantages/disadvantages? Would this ever be used on home PCs?

Re:Content Addressed Storage

[gedhrel]

CAS products typically identify objects by a hash of the contents. They're usually accessed via an API: you give it the stuff to store (and, optionally, a retention policy) and you get an opaque ID in retrun. You can get the content back if you have the ID, so IDs normally are capabilities (in the authorisation sense).

These things often sit behind document or other content-management systems. The advantage is that you only store one copy of the data; the disadvantage? Well, mostly, it's just a different way to store data. You wouldn't "use" this directly as a filesystem on a desktop. It might, however, form a part of the storage layer for an HSM system. So you need to understand what it is before you realise how you might use it.

Various companies have CAS systems (it's a fairly new storage buzzword - think EMC's is called "Celerra", for instance) and you'll find whitepapers explaining what these are. Don't expect to be amazed :-)

Re:Content Addressed Storage

[gedhrel]

Meant to add: "Content Addressable Storage" does NOT mean you can look up objects by their metadata attributes, normally. That kind of functionality is usually a part of the application that uses CAS and sits on top of it (although IIRC the Celerra which I mentioned earlier does have some kind of metadata search facility).

Re:Wow, that's expensive.. lemme do ~10 times bett

[superpulpsicle]

When it comes to enterprise storage, the heavy cost is not associated with the equipment. As expensive as a tape library might sound, you only buy it once. It's the support and service bills to follow that STK capitalize on.