The Riverbed Blog (testing)

A blog in search of a tagline

Riverbed’s Universal Data Store

Posted by riverbedtest on January 5, 2009

Riverbed has been able to scale its disk-based WAN optimization solution for thousands of customers, including in some enterprise networks that rank among the largest and most complex IP networks in the world.  One of the key reasons why the Steelhead appliance solution has been successful in very large deployments is Riverbed's Unversal Data Store architecture.  Riverbed is the only major WAN optimization vendor to implement a disk-based storage model that allows each Steelhead appliance to store each unique string of data only once, regardless of the number of peer devices that it is communicating with.  In contrast, Riverbed's main competitors including Cisco WAAS, Blue Coat ProxySG/MACH5, and Juniper WXC all use a per-peer data store.  In other words, with each of these competitive products there is a separate data store used to optimize data transfers to each peer device.

Graphically, this can be shown in the following diagram.  Riverbed's universal data architecture store allows the central Steelhead unit to store byte-level data patterns in a single shared data store–the "Universal Data Store"–that is used to optimize WAN transfers to all peer Steelhead devices.  A file that is fetched by users at 10 different sites is only stored once in the central Steelhead's shared "Universal Data Store."


On the other hand Riverbed's competitors use what we describe to be a "per-peer" data store architecture.  Under this approach, the central WDS device communicates to each peer WDS device using a separate independent data store.  If the central WDS device is communicating to peer devices at 10 different remote branch offices, then the central WDS device then there will be 10 separate data stores in the central WDS unit.  This is graphically shown in the diagram below.


The implications of the these differences are significant in a large network.  When using a WDS solution such as Cisco WAAS or Blue Coat ProxySG/MACH5, the central WAN optimization device will have difficulty scaling.  If users at 10 different sites access the same file from the data center, then the central WAAS device will need to store that file 10 separate times.  If users at 100 different remote sites receive an email, perhaps an email with a 10MB attachment, then that email and its attachment will be stored 100 separate times in the central WAAS unit.  In this case, the central WAAS device will consume storage 100 times faster than the equivalent Riverbed solution.  As additional branch offices are added, the amount of central storage allocated to optimize WAN transfers for each branch office will have to shrink even further, further reducing the overall effectiveness of the WAN optimization solution.

In a different blog I wrote late last year, I talked about the tragic consequences of limiting the evaluation process for WDS solutions to an isolated lab environment or limite-scale production pilot.  The per-peer data store is only one issue among several that don't manifest themselves in small-scale POC's and isolated lab testing.  The issue only becomes significant–and painfully so–in a large-scale production rollout. 

For these reasons, I continue to advise anyone evaluating WAN optimization solutions to talk to references of other customers who have deployed the various vendor solutions at scale.  You should have access to reference contacts in a confidential 1:1 consultation.  A vendor with a WAN optimization solution that can truly scale should have no problem supplying these references; be wary of any vendor who cannot supply references, or who refuses to allow you to talk confidentially to them.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: