By now, most IT storage professionals understand the potential advantages of building a storage network based on good ol' IP (as opposed to Fibre Channel): lower cost, use of existing personnel and expertise, easier management, etc. The question is no longer why or how you should use IP; it's when
I'm talking about building a local storage area network (SAN) by transporting block-level SCSI commands over TCP/IP via the emerging iSCSI standard, as opposed to using long-distance IP networks to connect existing Fibre Channel SANs. Using IP to transport storage I/O over MANs and WANs (via the emerging FCIP or iFCP standards) is a no-brainer, but using iSCSI for local SANs still has hurdles.
Ignoring the price and performance questions for now, let's look at the hardware pieces that would be required to build an end-to-end iSCSI SAN.
Switches-You should be able to use standard, low-cost Ethernet/IP switches in iSCSI SANs (one of the key cost-savings arguments of the pro-IP crowd), but performance may be an issue. Again, the verdict is out.
Target devices-This part of the puzzle may take awhile. IBM is the only big vendor shipping a "pre-iSCSI" disk array (the 200i), but even IBM admits that performance is not up to the requirements of most storage applications. Smaller vendors such as 3ware are also shipping pre-iSCSI arrays. But it probably won't be until the second half of 2002, at best, that you'll have high-performance options from multiple brand-name vendors. The same goes for "native" iSCSI tape libraries.
In the absence of iSCSI target devices, you can use routers or switches that convert from Fibre Channel or SCSI to iSCSI/IP, but that may topple the low-cost equation. It works, but it's expensive in the context of a local SAN.
Two more hurdles remain: The iSCSI standard is not expected to be ratified until the second quarter (and that's just a guess), and interoperability between iSCSI products from different vendors is still not proven.
So, when will you be able to build an iSCSI SAN? You could do it today, but it's unlikely that performance will be adequate, and the potential cost savings are questionable. I think that mid-2002 is a good bet for experimental, departmental iSCSI SANs. Momentum will pick up in the second half of the year, with 2003 perhaps being "The Year of iSCSI."
In this issue
In the meantime, read up on IP storage, starting with this month's Special Report, "IP storage and Fibre Channel: Competition and co-existence," by John Webster (see p. 21). The article presents an excellent overview of the various IP storage standards and their relative positioning with/against Fibre Channel and-down the road-InfiniBand.
To spur healthy debate, we've also added an Opinion piece on the subject: "IP storage: Separating hype from reality," by Jeff White and Bill Peldzus (see p. 48).
If you have an opinion, send it to me at email@example.com.
Dave Simpson, Editor-in-Chief
InfoStor appoints new publisher
Mark Finkelstein recently joined our team as publisher and will head up all sales and marketing efforts, expand the InfoStor franchise, and work with editor-in-chief Dave Simpson on the future direction of the publication.
Before joining PennWell Corp., InfoStor's parent company, Mark was most recently co-founder of TechOnLine, an e-learning engineering and IT portal. In addition, he was vice president of sales and business development.
Previously, Mark was sales director, business development, at Softbank-Ziff-Davis, a large publishing firm and trade show producer, where he was primarily responsible for managing sales for all Networld+Interop events worldwide. He also had held sales and business development positions at Reed Exhibitions, where he was responsible for their DSPx, Intermedia, and NepCon events.
Please join us in welcoming Mark to the growing InfoStor team!