With the exception of IBM's DS300, what EMC, IBM and NetApp's iSCSI systems have in common is that they're built on top of existing NAS systems. That's a logical approach for a vendor to take, says Randy Kerns, senior partner at the Evaluator Group. From a vendor's perspective, "you already have the Ethernet," and all you have to do is "take out the NFS/CIFS layer and access the LUN directly," he says. "Really, it's a very straightforward implementation." For users too, an all-in-one NAS and iSCSI system may make sense, Kerns says. "It gives you a measure of flexibility...

But while an all-in-one iSCSI/NAS array may be convenient and easy to use, it does present some tradeoffs, says Tom Major, vice president of marketing at LeftHand Networks, which makes IP SAN storage systems. From the get-go, "all-in-ones" may force you to dedicate storage resources as either block or file, prohibiting you from dynamically allocating storage resources. And like a regular NAS appliance, an iSCSI/NAS appliance suffers from poor scalability, Major says. A better approach for users who need both block and file access, he says, may be to implement a dedicated iSCSI device as well as a NAS gateway, such as one built on top of Windows Storage Server 2003.

Going forward, that's the approach Major believes the big boys of storage will take. Already, EMC is rumored to be coming out with an iSCSI-only version of its DX100, probably later this year. In the meantime, the NAS/iSCSI combo, says Major, is just an "easy way to dip their toe in the water."