Cheap storage and sleepless nights

The low-end storage marketplace is rife with an ever-expanding selection of incredibly cheap, feature-rich NAS devices. With support for RAID, hot-swap disks, CIFS file sharing, iSCSI block-level access, and a multitude of other useful features, these devices are a godsend for small businesses with surging storage needs. But if you try to use these devices for the same roles as their expensive enterprise-class cousins, disappointment may result.

The storage Swiss Army knife

A great example of this type of device is the QNAP TurboNAS line, so I’ll use it as an example here. To be sure, the marketplace is packed with offerings from the likes of EMC’s Iomega, D-Link, Netgear, and numerous others — many of the points I make here apply equally to these devices.

The QNAP TurboNAS line is available in a wide selection of two- to eight-disk enclosures including both table-top and rack-mount form factors. The items ship without the required SATA disks, so the choice of drive is up to the end-user, making them very flexible in terms of capacity.

Using a Linux-based operating system, they support a cornucopia of features that even enterprise-grade devices rarely provide. These include the ability to offer files via any combination of CIFS, NFS, FTP, and HTTP, while also supporting multipath iSCSI block-level access. In addition, they support volume-level disk encryption; a full range of popular RAID types, including RAID6; and online RAID expansion, which makes it easy to upgrade your disks. Many of them are even approved for use with Citrix XenServer, Microsoft Hyper-V, and VMware vSphere in iSCSI-based clustering environments.

Even better, they work with a bunch of other interesting features, such as the ability to aggregate and store feeds from IP-based security cameras, act as a print server for USB printers, and utilize cloud storage platforms such as Amazon S3 for online data backup.

The feature list is a testament to the flexibility of the (largely) open source software these systems are built on. It’s easy to see how they could become the heart and soul of a small company’s network — and therein lies the problem.

Managing expectations

The key to using cheap storage — and not living to regret it later — is to understand its limitations and to build your storage design and disaster recovery plans accordingly. Sub-$5,000 NAS devices that offer SAN-like features bear very little resemblance to enterprise storage platforms in terms of hardware and software architecture. They offer comparatively anemic performance and significantly less hardware redundancy; they’re also subjected to far less quality assurance, and they rarely come with the same level of manufacturer support.

Those limitations don’t rule out these devices, but you need a realistic plan for them. This usually means building significantly more redundancy into your storage design than you might if you were using an already internally redundant enterprise storage platform.

For example, let’s say you’ve been tasked with providing a server and storage environment to a relatively small business — say, 30 or 40 users. A few years ago, you might have been able to get away with a single Microsoft Small Business Server for file sharing, email, and general domain services.

Today, I’ve seen very small businesses as widely varied as vet clinics, car dealerships, and dentist offices that find themselves in the position of needing four or five servers to get the job done — whether due to increasing demand for desktop virtualization, a growing number of server-based applications requiring dedicated servers, or realistic projections of compliance requirements.

That expanding stack of requirements, combined with the uptick in performance offered by servers, might push you toward using server virtualization to reduce the number of physical hosts. But if you pile all your VMs on a single host, you’ve left yourself with a glaring single point of failure. Instead, many businesses install a pair of servers and set them up in a cluster with shared storage — so either server can fail and business can continue unimpeded.

Companies of this size will rarely be able to afford even the lowest-end SAN, so that’s where our cheap NAS device might come into the picture. Provided that the limited number of low-performance SATA disks were able to keep up with the users’ transactional storage demand, an iSCSI or NFS-enabled NAS device might work perfectly — at a tiny fraction of the $20,000-plus price tag of an enterprise-class device.

But you can see what’s going on here: This architecture simply moves the single point of failure from the server tier to the storage tier. That might be OK with a storage tier composed of a battle-hardened storage device with redundant power supplies, redundant cache, redundant controllers, and a 24/7 on-site support contract, but without those features you’re wide open to a range of production-killing failures.

What kind of failures? Here’s a simple example I’ve heard about on multiple occasions. After being in production for a year or so, a company with a low-end NAS device suffers a power failure. After power is restored, the NAS doesn’t want to boot properly. After a few rounds of conversation with offshore online support techs (oops, no phone support included!), it turns out the device is working fine. It’s performing a simple disk consistency check. Unfortunately, because the onboard processor is so slow and the disks are so large, that check may literally take a day or two to finish — ouch.

Tips on avoiding downtime

In light of these reliability constraints, here are a few points to consider when determining how to deploy of lower-end NAS devices:

• Backups: Backups are always important, but they’re even more so when you use a low-end NAS. Whatever your approach (tape, disk-to-disk, or online) to providing backups, make sure you have a solid understanding of exactly how long it will take to restore your data after a catastrophic failure. Also, factor in the time to troubleshoot the problem and acquire any necessary replacement hardware.
• Buy a spare: If you’re going to depend heavily on a NAS device, consider buying a second one and deploying both at the same time. Also consider buying extra disks to have on hand if one of the active disks fail. Warranties for this class of device generally don’t include cross-shipment of a replacement NAS — and you’ll likely be on your own for replacing failed disks — so buying spare parts is often worth the trouble.
• Use replication: Many of these devices support periodic replication from one NAS to another, making redundancy much easier. Note, however, this kind of replication is never a replacement for backups. You still need offline backups to protect against data corruption (which could easily replicate to both devices) or physical loss of both devices due to a fire or similar disaster.
• Have a plan and test it: The most important step is to have a plan for what you’re going to do if the critical device dies in the most inconvenient way imaginable. Rock-solid backups and data replication aren’t much use in an emergency if you haven’t run through the process of using them — nobody wants to have to Google instructions on reconfiguring the iSCSI initiator on a virtualization host while the network is down. Run through the whole procedure before you need to do it and you’ll be ready.

Using low-end NAS devices will undoubtedly save you a ton of money, but you need to recognize that you are making an explicit decision to trade reliability and support for cost, which may result in downtime and data loss that completely overshadows any money you pocketed. Planning ahead and expecting the worst can reduce the risk and still saved you that big chunk of change in the end.

This article, “Cheap storage and sleepless nights,” originally appeared at InfoWorld.com. Read more of Matt Prigge’s Information Overload blog and follow the latest developments in storage at InfoWorld.com. For the latest business technology news, follow InfoWorld.com on Twitter.