Storage Wars 2 v. 2

By Jason Dowd Vice President of Pinnacle Computer Services

www.pinncomp.com or 1-812-476-6662

In our last exciting installment https://city-countyobserver.com/2013/08/19/storage-wars-part-1-2/, we discussed RAID. Using RAID it was very common by the end of the 90’s to see servers with so much storage that the drives had to be placed in an external unit and connected to the computer via a thick, bulky, and highly length constrained SCSI cable.

If you have an external hard drive that you attach to your computer via a USB cable, that’s pretty much the idea except that your small, inexpensive external USB drive probably has more storage capacity than the disk arrays of even large database servers in the late 90’s. Mine is 2 Terabytes with transfer rates of nearly 1 Gbps, and it cost me less than $150.

Around this same time in the late 90’s, someone had the following clever idea. Instead of having every server connect to its own, separate disk array, what if we just had one big disk array to which every server connected?

That is the basic ides of a storage area network, or a “SAN” as it is known in the industry. Now, when we configure logical disks in the array, we also specify which connected server gets to use that logical disk. Also, those bulky, length constrained SCSI cables have been replaced with standard copper and optical networking cables which can be run for much greater lengths.

Currently, SAN’s are a mature technology and are pervasive in large organizations with adoption in medium-sized organizations growing at a feverish pace. However, in spite of maturity and wide adoption, the SAN space is currently brimming with innovation mostly from small and niche players scrambling for a piece of the storage pie. At the same time there are also a few large vendors who, due to questionable business practices and/or loss of technological leadership, are seeing significant loss of market share.

336TB of Storage in 5U of rack space. Just look at that baby!

As such, this is an extremely exciting space to watch. But for anyone who either owns a SAN or is thinking of investing in one, it also makes the going quite treacherous. The main thing you need to realize is that five years from now, this space is going to be littered with the smoking corpses of the vendors who didn’t make it, and when that time comes, you probably won’t want to own one of their products.

Unfortunately, just because a vendor is large and stable with a broad install base doesn’t mean they have the best product or necessarily even a very good one. And some large vendors in this space as well as others have an interesting revenue model: every few years, you have to buy all new equipment and pay for the professional services to migrate you to it. If this seems unethical, that’s because it is, but a surprising number of large companies you’ve probably heard of seem to live by precisely this model.

We can’t give you too much specific vendor guidance in this article, but we will explore several key decision points for SAN selection and implementation.

 “Fibre Channel or iSCSI? How about both!”

How about both! When connecting a SAN to servers, these are the two basic technology choices: iSCSI and Fibre Channel. There are other options, but we don’t see much of them. iSCSI seems to have some obvious advantages as it runs over standard networking equipment while Fibre Channel requires its own, separate infrastructure which is expensive. However, when it comes to performance vs. price per port, Fibre Channel is still the king. But if you want your Fibre Channel infrastructure to be fault tolerant, it will cost you more.

One of our favorite implementation strategies is to provide primary connectivity between servers and SAN via Fibre Channel and use iSCSI as a failover. This is a configuration that works very well and is quite cost effective. However, your SAN must support Fibre Channel and iSCSI at the same time. Most do, but some don’t.

“Fibre Channel and iSCI battle it out for world domination”.

Also important here: if you decide to brave the world of 10Gbps Ethernet for your iSCSI connections, make sure you get input from your SAN vendors about exactly what Ethernet switches you should be using. This technology is still finicky as well as pricey.

Firmware Upgrades

Always ask a perspective SAN vendor about their firmware upgrade process. Some vendors will require that you buy more storage than you need so you can migrate data away from disk enclosures before upgrading their firmware if you want to keep your data available during the upgrade process. The biggest issue with this is that migrating such large amounts of data can take days or even weeks, and you will have to spend that much time for each enclosure you need to upgrade. So a process that can be easily done in an hour with no downtime for some SANs can be a process that takes months on others.

Some SAN vendors’ firmware upgrade process.

NAS or File Server?

“NAS” stands for “Network Attached Storage”, but that isn’t really very descriptive. SANs present data to servers in a form indistinguishable from a hard drive. A NAS on the other hand presents data to clients or servers in a form indistinguishable from a file server. The two are very different and most SANs come with no NAS functionality. However, many SAN vendors offer an additional piece of equipment that serves the data on the SAN up as a NAS. That is, in a form readily digestible to users.

Our experience, however, is that this is largely a waste of money, and that you are probably better off just building a file server or two and attaching them to the SAN like any other server. Then let them take care of the presentation to users.

There is certainly more that can be said on the current and anticipated future of this space, but for that you will have to join us for “Storage Wars – Part 3” coming soon!