Americas

  • United States
by Anne Skamarock

N+I panel debate routing traffic through main data network

Opinion
May 20, 20033 mins
Data Center

* What N+I panelist and attendees make of putting storage on data networks

At Networld+Interop 2003 last month, I moderated a panel on IP storage which discussed whether or not routing storage traffic through your main data network makes sense. There were some interesting questions raised by the audience as well as perspectives shared by the panelists.

The panel was made up of the CTOs of Stonefly Networks (Bill Huber), Network Appliance (Steve Kleinman), Andiamo/Cisco (Tom Edsall) and HP (Randy Haagens). The executives brought perspectives from traditional data networking and new IP-based storage networking. A quick poll of the audience showed they were primarily management-level customers looking to implement storage-area networks (SAN) using IP technologies.

Right off the bat, everyone agreed that the primary influence in how well a data network performs has to do with how well the network is architected.  Poorly architected networks may support your current performance requirements but the design flaws would be exposed as further SAN traffic is added. Kleinman was the first to point out that this has been the case in networks as appliances have been added to network-attached storage (NAS), which has been accessing storage via the IP network for more than 20 years.

The second point brought out was that the addition of storage networking through the use of the recently approved iSCSI protocol to your primary network depends greatly on the application accessing the storage. One application discussed was databases.  Databases or applications built on top of databases require high levels of performance. In this case, it was agreed that the use of a separate “storage-only” network would prove the best design for the database application. On the other hand, e-mail doesn’t usually have the high performance requirements, so using iSCSI on your current data network would probably work well. Of course, all additional traffic to any network will tend to lower performance so that current usage levels should always be taken into consideration before adding traffic.

The attendees were very interested in how “real” or ready, the iSCSI products are today.  This got the panelist excited as three of the four companies represented currently have iSCSI products available.  All agreed that iSCSI is here and available to provide alternative solutions to business problems. As well, they described how the other IP storage protocols such as FCIP and iFCP could be used for extending Fibre Channel networks across IP.

In summary, IP is a viable direct-attach storage networking option being used in customer’s production sites today. Over the next six months to a year, expect to hear IP storage announcements from many of the large storage vendors, especially in the midmarket space.