The convergence of computing, communications and storage grows ever stronger with the introduction of new storage-area-networking (SAN) technologies. In my last two columns, I discussed the exciting though difficult-to-predict market for Infiniband silicon. Now, let's look at the emerging iSCSI silicon opportunity.
Whereas Infiniband offers an elegant solution for server clustering and in-building SANs, iSCSI is more focused on bridging SANs to local-area networks. Though interested parties are excited about the prospect of using iSCSI to bridge SANs to LANs, the reality is that iSCSI is most valuable for bridging one SAN to another over a wide-area network.
The beauty of iSCSI is that it transparently exposes storage systems without any upgrade or modification to the massive installed base of Internet Protocol networking gear. All that's needed is a gateway box or line card. Sound familiar? Well, it should; it is pretty much the same selling point as voice-over-IP. However, this scheme is even better, because low latency is not as critical for storage data as it is for voice.
TCP/IP offload engines are one of the most critical ingredients in making iSCSI a success. Intellectual property for TCP/IP offload will find its way onto silicon used both for iSCSI initiators (iSCSI terminology for a client adapter card) and gateways. This intellectual property is especially critical in initiators, since TCP/IP processing consumes about 1 Hz for every bit of data. So, without an offload engine, a 1-GHz processor would be at 100 percent utilization when sending 1 Gbit/second of data.
Just how big is the storage network processor market used in iSCSI systems? Gartner Dataquest analyst James Opfer predicts that in 2005, a total of 5.74 million iSCSI ports will ship. This breaks down as 5.4 million initiator ports, 90,000 gateway ports and 250,000 iSCSI-to-Fibre Channel converter ports.
Storage network processor vendors are targeting 2003 pricing of about $75 per 1-Gbit/s port. With 25 percent price erosion, that works out to $42 per port in 2005. This implies an iSCSI storage network processor opportunity of approximately $250 million in three years.
It is helpful to keep two things in mind. First, this forecast is simply middle of the road. There is potential upside as well as downside. Second, just as in communications network processors, this opportunity will be split between ASICs and ASSPs, with ASICs mostly likely dominating.
Jeremey Donovan (jeremey.donovan@gartner.com) is Chief Analyst at Gartner Dataquest.
RSS
RSS| 
Digital| 
Mobile
