The CTP version of Velocity had been released a while back.

It certainly has great benefits.

Here's a link to Bart's post going over this release, worths the check.

From Bart's post:

Introduction

But first... what's in a name? Multi-tiered distributed applications are common-sense nowadays and with cloud computing within reach the need to build scalable distributed services has never been bigger. One of the core aspects in enabling those scenarios is to have intelligent caching of objects, not only to reduce the number of accesses to the underlying data source but also to boost availability by employing scale out techniques. Obviously, developers want to be able to do all of this without having to worry about the complexities that this brings, having to deal with load balancing and availability themselves. That's where Velocity comes into play.

The core idea is very straightforward: we have a cache that behind the scenes is distributed and replicated across a bunch of machines called the cluster. Storing data in the distributed cache is as easy as calling some Add or Put method, and retrieving it is as easy as calling Get. With some creative stealing from the documentation we end up with the following picture:

An important thing to emphasize is the fact the cache clients deal with regular .NET objects all the time and don't have to worry about storing those objects. Indeed, .NET serialization takes care of the rest. There are more concepts to it such as cache eviction policies (when objects are removed from the cache, such as least-recently used or LRU), the distribution mechanism where simple clients just contact the cluster "in the cloud" through any cache host and get redirected to whatever host the object is available on versus routing clients that have awareness of object placement through a routing table. Other important pieces include the supported concurrency models and associated locking mechanisms but let's not go there in this introductory post.