Last modified: April 10, 2001 11:30 AM PDT
Commentary: Cast a skeptical eye on iSCSI
The idea of accessing cheap SCSI storage, available from many suppliers, over IP Ethernet networks rather than more expensive Fibre Channel solutions, is an interesting one. However, to fully replace an FC solution, iSCSI needs sophisticated software that, so far, has been lacking.
Also, Cisco Systems' price for its new iSCSI-based SN5420 router, reported to be $27,000 per router (for single-port protocol conversion; this does not include the price of Gigabit Ethernet switch ports), is much too high given the lower data-transfer speeds of IP compared to FC. The SN5420, which Cisco announced on Monday, will link Gigabit Ethernet networks to FC-based storage-area networks (SANs).
Cisco's storage network plans are clearly driven by the rapid expansion of the storage market. During the past few years, storage has evolved from being an integrated component of a server to include separate boxes attached to a server or directly to the network. This trend, combined with the likelihood that storage demand will increase by an order of magnitude during the next three years, makes storage a particularly attractive target for Cisco.
Part of Cisco's strategy has been to partner with storage vendors such as Brocade that provide FC-based storage systems for wide-area network storage access and replication services. However, the introduction of the SN5420 represents a sharpened focus for Cisco (in campus environments) and places it in competition with Brocade and other vendors in the long term.
Here's mud in your iSCSI
iSCSI could enable organizations to build generic, plug-and-play SANs over Gigabit Ethernet networks. However, a drawback of this approach is that it is compute-intensive, demanding as much as 40 percent usage of the CPU to manage the storage traffic. FC systems typically use dedicated processors for this task.
See news story:
Cisco, Lucent jump into storage networking
Another current issue with iSCSI is it lacks comprehensive operating system support, limiting options to Microsoft's Windows 2000 and some variants of Linux. But the biggest issue concerns software. The additional software functions that enterprises require to support many business applications will not be transparent to many iSCSI implementations.
Management is becoming the primary selection criterion for storage solutions. Focusing only on the raw dollars-per-megabyte evaluation will cause more problems than iSCSI can solve. Users should be wary of the early iSCSI storage implementations from both IBM and Cisco, because the technology suffers from numerous shortcomings.
Users certainly should monitor developments in iSCSI, but must realize that the trend to carry SCSI storage traffic over an IP network is immature. Multiple proposals are vying to become the standard, and we do not expect a resolution before 2002. The ability to carry SCSI storage traffic over an Ethernet network for communications within data centers will not mature until 2004-05.
IT organizations should continue with native FC solution deployments and wait for the evolution of iSCSI costs, interoperability, operating system support and approved standards. However, bringing up iSCSI in discussions with storage vendors is an important way to understand how they are planning to deal with a variety of storage-access protocols--including iSCSI, FC, FCIP, SoIP and TCP/IP.
Meta Group analysts Sean Derrington, David Cearley, Val Sribar, Dale Kutnick, Jack Gold and William Zachmann contributed to this article.
Visit Metagroup.com for more analysis of key IT and e-business issues.
Entire contents, Copyright ? 2001 Meta Group, Inc. All rights reserved.