A little over 15 years ago we started to see the creation of the backup appliance market with the creation of the deduplication backup target appliance. Since that time, we’ve gone from having no backup appliances to having so many of them that IDC and Gartner created market categories to track them. They brought a lot of value to a lot of customers and some of them have also added to the bottom line of a lot of backup vendors. This led to pressure on some of the independent software vendors to come out with their own backup appliances. The question is whether or not that’s a good idea.
Backup Appliance Types
There are two main types of backup appliances: those receiving backup data and those creating backup data. Each of these two main types break down into sub-types, and each of these types of appliances is designed with very different goals in mind.
The Deduplication Backup Appliance
The first type of backup appliance is the type that created the market in the first place – the deduplication disk target. The initial purpose of these appliances was to bring the benefits of disk backup to traditional backup products, at a cost close to that of tape – without requiring backup software vendors or customers to change much about how they did things.
These purpose-built systems are typically built on an Intel platform running some type of Linux kernel upon which runs the deduplication software. They typically receive backup data over the network via an NFS, SMB, or Virtual Tape Library (VTL) interface, as well some custom interfaces designed just for this purpose. This category includes products like EMC Data Domain, Quantum DXi, HPE Storeonce, Exagrid, Falconstor, and NEC Hydrastor.
Regardless of interface, deduplication appliances made the concept of electronic vaulting a reality, since they understood data at a much more granular level than even an incremental backup job, only storing changed blocks of data.
This enabled replication of full and incremental backups, which was something simply not possible before the advent of deduplication. This allowed for the first time customers to have an onsite and offsite backup without touching a tape or using a human. Since many of the lost data scenarios of years past involved humans, being able to have onsite and offsite backups without involving a man in a van has changed the world in incredible ways. It moves backup from being just for operational backup and closer to being capable of providing full disaster recovery services. The ways that different products are meeting DR needs are yet another area where these products are differentiating themselves.
Inline vs. Post Process
Another way that one can sub-categorize the deduplication backup target market is whether they deduplicate data as it is being written to the appliance (i.e. inline) or after it’s first been written to a staging area (i.e. post process). The main advantage of the inline approach is that it does not need the staging area or the additional IO and processing power to write data once and deduplicate it later – although “later” is usually seconds later. The main advantage of the post-process approach is during recovery of recent data, since it can be read from the staging area in its native format. This is most noticeable during instant recoveries, where the backup storage is being used as a datastore for the virtualization engine and VMs are being run directly from the backup storage in order to quickly bring them online.
That’s a summary of backup appliances meant to receive backup data. Let’s take a look at those that are designed to create backup data. They also come in two different types.
Backup Software Vendors Become Hardware Vendors
Faced with the popularity of the deduplication disk target appliances, backup software products sought to develop their own appliances. They did this in two different ways. The first was essentially a tight partnership between a software and a hardware vendor. In this strategy software vendors partner with hardware vendors capable of producing the physical appliance, and simply packaged (pre-installed) their product with an appliance sold by their partner. These products were aimed at giving the customer a single SKU to purchase and a single point of contact for support. While convenient for initial implementation the customer is free to move to other products as time goes on.
Other backup software companies chose to develop their own physical appliance using rebranded white box vendors. These appliances allow the backup software vendor greater control over the hardware components that make up the appliance, and also allow them to support their own customer directly. This is the type of appliance that the title of this article is referring to. Should backup software vendors produce their own backup appliance, or should they partner with others for this purpose?
A House Divided Against Itself
Software companies acquiring or developing their own appliance – versus working with partners to do the same thing – often end up limiting customer choice. The first reason this is the case is different customers want to purchase their appliances from different vendors. Some may prefer big-box names with full support, where others prefer white box vendors they support themselves. Some prefer to build their own systems with best-of-breed components, where others prefer systems where everything is already pretested. Some may prefer target deduplication appliances using the inline approach for its space advantages, and others may prefer the post-process method for its recovery advantages. But when a single company owns both the software and the hardware, customers have fewer choices.
What happens when a software company starts producing its own appliances is it starts competing with its own channel. When you compete with the channel, sometimes the channel does not want to play ball. Sure, a big-box vendor will still sell its server to a client of a particular software company. What might not happen is the software company partnering with the big-box vendor to create a joint appliance, because then it would be helping its partner to create a product that competes with its own appliance. The result is that if a customer wants to buy an integrated appliance from the big-box partner, they might not be able to do so.
A similar problem is once a backup software company has its own appliance, it starts competing with companies that were once partners. It makes it more difficult for other appliance vendors to partner with that same company to do things that would increase its value. Because again, the backup software company that owns its own appliance doesn’t want to help its competitors look as good or even better than its own hardware product. It wants you to buy both its backup software and backup hardware.
It also creates an incentive to create tie-ins to that product that only work with that product. Looking outside the backup world, we see a perfect example of that with Apple’s MacOS and MacBook line. Since Apple made the decision to only support MacOS on Apple-produced hardware, they have made hardware decisions for their customers, choosing Lightning over USB-C, DVD over BluRay, and hardwired RAM and CPUs over upgradeable ones. Some of their latest OS features are also tied to their hardware, such as the touch bar features in the new Macbook Pro. So even if you figure out how to run MacOS on non-Apple hardware, you won’t be able to take advantage of these new features without an Apple touch bar.
You see the same challenges in the data protection industry. A perfect example is what if Company X supports live recovery (a great feature), but only if you buy its appliance? You might prefer to use a post-process appliance for this feature since you can run your VMs from the landing zone for better performance.
Alternatively an in-line deduplication solution could optimize their offering so the need to dedupe on the fly is not noticeable to the users. For example it could leverage more CPU power, more ram or flash caching of instantly recovered data.
But what if company X offers live recovery but doesn’t have a post-process product or an optimized inline product; it has an non-optimized inline product. This leaves the user two choices. Don’t buy company X’s backup product (which you like), or buy its backup product with someone else’s appliance – and miss out on a really important feature (live recovery.) Either way, the vendor’s choice to produce hardware and software has limited customer choice. One might even argue that this was the point. It’s called vendor lock-in.
Be Careful What You Ask For
Some customers are pushing independent software vendors to come out with their own appliances. Be careful what you ask for, because you may create something that you weren’t looking for. What happens if your favorite backup software product suddenly starts putting out its own appliance? At a minimum, other appliances from other vendors may cease to exist. Why should a big-box vendor partner with your backup software product only to compete with them? What happens when someone at the helm starts realizing that by creating a tighter integration between the software and the hardware, the company increases its bottom line enough to make that worthwhile? You may find yourself forced to buy a different hardware product than you want in order to keep the software that you like.
Flexibility and customer choice should be king. Backup software products should partner with big-box vendors and white box vendors to create products that are sold via those partners allowing each company to do what it does best. This increases choice and gives customers that are looking for a ready-made appliance exactly what they’re asking for – without giving them what they weren’t asking for.
Sponsored by Veeam