Networking storage with Ethernet
Although much has been written about Storage Area Networks (SANs) over the past few years, only one in three IT shops has yet to deploy SANs, according to the April 2003 Ziff Davis Market Experts study on storage purchasing. Most of those SANs have been deployed in larger organizations, and for the most part, these storage devices are connected to each other and the systems they serve using FibreChannel wires, switches and routers.
Although FibreChannel, which solved the problems of distance limitations on SCSI cabling between disk and server, has been around for years, it is a very different networking infrastructure than is used for most of the rest of IT’s networks – namely Ethernet and TCP/IP. Although there are many benefits of having a separate network for storage, including better performance and easier troubleshooting, since you don’t have to separate out the network traffic from the storage traffic, there has been a significant downside: finding and retaining IT staff who understand how to keep FibreChannel storage networks happily running, as well as the additional investment in another network – and the additional hardware costs incurred in the switching gear to connect it all together.
But SANs aren’t the only way to get the benefits of networked storage. Network Attached Storage (NAS), which has been around even longer than SANs, is a quick, easy way to get additional file servers up and running on an existing TCP/IP Ethernet network. NAS devices can literally be installed in minutes, don’t require any special cabling or a new network, and are, usually, less expensive per gigabyte of disk, to deploy. Still, you have to be careful how you connect it – if a NAS device is attached on heavily used network segment, the storage traffic can impact the LAN traffic and vice versa. Also, for applications that demand block-level access to disk storage, such as Oracle or MS-Exchange databases, NAS devices won’t deliver the performance that good old fashioned direct attached storage (DAS) in the server cabinet offers.
Internet SCSI (iSCSI) adds to the somewhat confusing array of choices. With the advent of gigabit Ethernet (gigE), and the imminent availability of 10 gigabit Ethernet, our old friend Ethernet now offers raw performance in the same ballpark or better than FibreChannel (which typically runs at one or two gigabits/second). Just as FibreChannel offers a way to extend the distance between disk drives and their associated controllers and servers, iSCSI takes the same SCSI protocols and encapsulates them in Ethernet packets over TCP/IP, which mean you can now get the benefits of a SAN, including support for block-level transfers, without having to budget, staff for, and maintain a whole new network to get those benefits.
If you’ve already gone the FibreChannel route, then not to worry – new generations of FibreChannel adapters, switches, and directors are coming with even faster speeds. But for the majority of IT shops out there who haven’t gone the SAN route yet, and who are still feeling the budget pinch that the current economic climate has thrust upon us, there is now another route to explore, an Ethernet-only SAN.
Does this spell hard times ahead for FibreChannel? Not in the short term – in fact most FibreChannel switch vendors are adding iSCSI support to their hardware, just as companies like Cisco have added FibreChannel capability to their network switches. But for companies with limited IT budget and a need to get access to high performance, block-level SAN storage, iSCSI may be the way to go.