Riverbed Granite: Enterprise Riches or Fool's Gold?

I've been spending some time wrapping my head around Riverbed's product introduction last week (you try  to decipher this product sheet) and I would certainly welcome your input. My instinct tells me we've got a creative new direction from Riverbed (kudos there), but one that today's enterprise should think twice (and then three times...) about . Here's why.

Disclosure: Silver Peak is one of my clients and I own stock options in the company. Silver Peak competes with Riverbed.

THE ANNOUNCEMENT

The announcement consisted of three introductions The Steelhead CX  Family appliances replaces Riverbed's xx50 series of mid- to small office optimizers (the CX family scales from 6 Mbps to 90 Mbps) adding more capacity and improving throughput.  The appliances are targeted at organizations that only want WAN optimization in their branch offices. The Steelhead EX Series appliances add a multi-function, enterprise-class branch office appliance to the Riverbed mix. If an organization wants what Gartner calls a Branch Office in the Box (BOB) then the EX will be the pick. The appliance runs Riverbed's Virtual Service platform (VSP), which is a Riverbed version of VMWare, enabling the Steelhead to run a virtual appliance.

Perhaps the most interesting product, though, was the Steelhead Granite. This is another storage optimization product, similar to Riverbed Whitewater.  Granite enables organizations to maintain data as SANs  in the data center and push that data out to the branch. Granite is built around two devices Granite Core, a physical or virtual appliance in the data center, and Granite Edge, a service running on a Steelhead EX in the branch office.

EVEN MORE PROPRIETARY HARDWARE FOR THE BRANCH

My concerns with the announcement are multifold. Partners and enterprises, for one, will be befuddled by the blistering array of products from Riverbed. While there's value to being exhaustive in your product set, it shouldn't be exhausting to identify the right product.

IT need also wonder about Riverbed's hardware upgrade train. How long will it be before these appliances are also superseded by newer models, forcing them to purchase yet more hardware? And how much more difficult will it be to get off that train when the storage department is also vested in the WAN optimizer?

Bundling applications into what's effectively a proprietary VMWare server unnecessarily locks IT into the hardware refresh cycle that they avoid in the data center. An alternative solution would be to load virtual appliances and VMware on off-the-shelf server hardware. This avoids the vendor lock-in yet still provides for consolidated deployment.

BLOCK-LEVEL STORAGE - WITH A TWIST

At the same time, though, Steelhead's Granite solution is unique and offers an interesting architectural twist on branch office storage. Block- and file- level access have generally inhabited different parts of the organization. Block-level storage with its flexibility and performance has traditionally been the domain of SANs and the data center. It's heavily used by databases, Exchange, VMWare and for booting up servers.  File based access is generally simpler and more cost effective and better suited for mass file storage or VMWare with NFS.

Increasingly the two approaches are converging with Riverbed, in this case, saying that Granite brings file-level intelligence to block-level data. The Granite Edge device presents to applications as an iSCSI target and then relies on intelligence in the Granite Core device to preload the Granit Edge with blocks likely to be requested by the user. Granite users are able to retrieve data from the local cache, not over the WAN, improving performance and in theory enabling IT to further consolidate their servers into the data center. In the event of a WAN failure, the Granite appliance will keep the change and synchronize them with the new version once the connection is reinitiated.

GRANITE PROBLEMS

It remains to be seen, though, as to how important Granite will be to most businesses. IT already has relationships with storage vendors, such as EMC, Dell, HDS, and NetApp, who can deliver similar capabilities by leveraging existing WAN optimizers.

What's more block-based applications, such as databases and Exchange, are already optimized with good client-server design. Adding a WAN optimizer further improves the line performance in many ways, not the least is enabling second-time Gets to be pulled from cache. All of this and without a proprietary branch office box.

Pre-population of the cache and local file operations certainly sounds intriguing, but this too carries its own risks. Granite provides local access to files even in the event of a WAN failure and then synchronizes changes upon WAN restoration. Yet, such architectures have caused problems in the past. File versions invariably fall out of synch, leaving users and the IT to have to manually reconcile differences. Is that really the approach today's IT pros want to take with their local storage?

Bottom line: Riverbed delivers a hardware refresh and a bold new direction. But the continued use of proprietary appliances and the blurring of the lines between IT and storage silos,  will make adopting the technology very difficult for IT. The limitations on agility and lock into a hardware refresh cycle should be top-of-mind for any IT manager looking at Granite for their organization.