TCP/IP storage devices

[MadRat]

I see the need in small businesses to have network storage that doesn't require servers or setting up propriety network cables. It would be nice to be able to grab a storage center (so to speak) and just plop one end of a patch cable to it and the other to the LAN's central hub. We have routers and switches that can do simple communications work, why not a simple storage center?

I'm guessing that its not exactly possible right now due to the propriety of current OS's (=Microsoft) to directly connect standalone storage devices on a LAN without setting up a server. What would be the stumbling blocks to making something like this possible?

The advantages:

1. No cost for a Server OS - Save big $$$
2. Simple connections - standard Cat5 would be nice
3. Simple setup - internal flash memory firmware setup would suffice
4. Simple upkeep - hang on the tree and leave it alone

I doubt it would take much to get this idea going. Their would be a pretty minute load on the brains of the device, so very low-cost processors (perhaps even an ARM) would suffice. The TCP/IP protocol could be used, with its NIC ignoring subnets outside the network. Some sort of security could be setup with a client for just about any OS could be used. While it may not be the most secure way of storing information, it wouldn't necessarily need to be for top secret data. The idea would be to keep the price down and to keep its use simple enough for just about anyone to setup.

 

[Sahakiel]

I think you're referring to network attached storage devices. I do believe several companies make some including Maxtor.

 

[SCSIRAID]

As long as you're not trying to boot off of it.... its simple. Netbios (PC/AT timeframe) allows this. NAS devices, SNAP server etc or even a W2K client doing sharing would do the trick. The real issue is performance... File level protocol has its limitations. Block level over TCP is in development now and is called iSCSI.

 

[spidey07]

yeah, called network attached storage. Been out for maybe 4 years I think.

Most run many flavors of application layer protocols like Microsoft SMB/netbios, NFS and raw FTP. The advantages are simplicity and cost. The disadvantages are limited flexibility by being just a file server. Never really caught on though.

 

[Peter]

Yep, these have been around for quite a while. LSI, Quantum (now Maxtor), Seagate, Adaptec,
all the big names in SCSI offer turnkey solutions for this. And so do the big and small names in server
equipment.

Two approaches, it's either an actual file server (serving files and directories, name says it), or a rather
stupid mass storage thingy doing SCSI-over-IP.

regards, Peter

 

[MadRat]

The cheapest system I could find was $1200 for only 60gb. Not a very economical solution if you ask me! I think it could be done for far less. These other systems appear to be servers in the true sense, not really standalone drivebays. The brains shouldn't have to be any larger than a PDA, and so the rest of the device should be a powersupply, NIC, and HD. I'm thinking it shouldn't be more than a 1U box for tree mounts, or a shoebox for desktops.

 

[SCSIRAID]

Peter,

Just curious but why would you classify SCSI over IP (aka iSCSI) as stupid?

 

[spidey07]

Here's my take on iSCSI and FiberChannel SANs.

iSCSI will eventually win because it relies on already tried, true and well managed network technology like ethernet and IP. Not to mention it will run on the same network if need be.

It is very inefficient (not to mention expensive) to have a switched SAN network comprising of VERY expensive fiberchannel switches when you can do the same thing with iSCSI and gig ethernet with better performance AND WAY cheaper cost. Storage centers like EMC, HP and Hitachi will eventually just be another node on a 1 gigabit or 10 or 100 gigabit IP ethernet network complete with all the service and latency guarantees needed for true network based storage.

I can't wait for it. FiberChannel standards bodies are dragging their feet and will suffer the same pitfalls of ATM because of it.

 

[SCSIRAID]

FC switches are artificially expensive. When iSCSI gets closer to reality I believe you will see the price point plummet. That will put the two in the same cost arena. iSCSI will still need to add the storage management capabilities to its tool suites to compete with FC. FC has reasonable fabric management today but not nearly as good as ethernet. I think it is going to be a good fight. The longer iSCSI delays the more entrenched FC will get. In the long run I agree that iSCSI will probably win. Let the games begin!

 

[Peter]

SCSIRAID, that's technically stupid as in "not much brain in there, just shifting raw storage data,
the file system handling is done elsewhere".

Definitely _not_ a stupid design, nor a stupid choice as a network storage solution. Just a rather
un-intelligent piece of hardware.

regards, Peter