Matevž Gačnik's Weblog

During exploration of high availability (HA) features of Windows Server AppFabric Distributed Cache I needed to generate enough load in a short timeframe. You know, to kill a couple of servers.

This is what came out of it.

It's a simple command line tool, allowing you to:

• Add millions of objects of arbitrary size to the cache cluster (using cache.Add())
• Put objects of arbitraty size to cache cluster
• Get objects back
• Remove objects from cache
• Has cluster support
• Has local cache support
• Will list configuration
• Will max out you local processors (using .NET 4 Parallel.For())
• Will perform graceously, even in times of trouble

I talked about this at a recent Users Group meeting, doing a live demo of cache clusters under load.

Typical usage scenario is:

1. Configure a HA cluster
   Remember, 3 nodes minimum, Windows Server 2008 (R2) Enterprise or DataCenter
2. Configure a HA cache
3. Edit App.config, list all available servers
4. Connect to cluster
5. Put a bunch of large objects (generate load)
   Since AppFabric currently supports only partitioned cache type, this will distribute load among all cluster hosts. Thus, all hosts will store 1/N percent of objects.
6. Stop one node
7. Get all objects back
   Since cache is in HA mode, you will get all your objects back, even though a host is down - cluster will redistribute all the missing cache regions to running nodes.

You can download the tool here.