Now that Im reviewing storage products for eWEEK Labs, Ive noticed that at least a couple of the same myths I saw in other product areas are also infesting the world of storage. The most common is that you can solve all your storage needs by simply plugging everything into the company Ethernet.
Why would people want to use their production network to support their storage needs? Well, its cheaper than a dedicated Fibre Channel SAN (storage area network)—probably a lot cheaper. With IT budgets even more challenged than they were in the past, being cheap can be a strong attraction.
As Ive said, this myth has been around before, but not always for storage. A couple of years ago, for example, I was testing IP PBXes, and one of the primary selling points was the ability of these devices to use your existing Ethernet network so you didnt have to pull phone wires. This would then save you money on adds, moves and changes.
While you may have been able to use your production network to provide the physical layer for your phone system, that didnt necessarily mean it was cheap, and it didnt mean it was a good idea. For example, many corporate Ethernet switches cant provide the prioritization that voice traffic needs. Likewise, now, they may not be able to provide the POE (power over Ethernet) needed by many IP phones, and they may not support the signaling method used by your phones. At the very least, this could mean youll need a new Ethernet switch for your phones. A new switch may mean new cable runs. You can see where this is going.
You see the same problems in storage. While the POE issue hasnt cropped up, the fact is that high-performance storage can require a lot of bandwidth—perhaps the entire 1G-bit capacity of a Gigabit Ethernet port. You cant share this Ethernet cable without hurting performance. You may also find that your existing Ethernet switch cant support the full performance needs of your storage network, either, because it doesnt have the backplane capacity or because it cant support the larger-than-usual frame sizes that some Ethernet-based storage solutions use to reach full performance.
With storage, the issue goes beyond just bandwidth. In some cases, the addressing, failover design or load sharing effectively requires either a separate switch or a switch that can be logically divided so that it can appear to be separate. While many—perhaps most—enterprise-class Ethernet switches can handle this last challenge, youre still going to need prioritization and the like, and youre not going to be able to share the same cable with the rest of your network.
With Ethernet-based storage, as with VOIP, this isnt always made clear when youre looking at a product to meet your needs. Its also easy to get drawn into the low-cost trap when youre looking at NAS (network-attached storage), Ethernet SANs or even server-based storage. You can probably use your corporate network for these services, and it may even function. But you can bet youll be getting less-than-optimal performance. In addition, you may find unanticipated costs ranging from specialized infrastructure to new cabling. Your savings could be a lot less than you thought, while your performance suffers, making the assumption of cost savings elusive at best.
Senior Analyst Wayne Rash can be reached at firstname.lastname@example.org.