Test the distributed system for performance and configuration issues.

Scale network services by multiplexing all nodes. Based on architecture, evaluate system to scale CPU, Network, and Disk in terms of availabilty to remote nodes.

"distributed systems" can entail many different things.  Resilancy (and redundancy) are a key.  Can your distributed system survive the failure of a node (or multiple nodes)?  You might look at Hadoop - as an example.

Is your distributed systems model one of all nodes doing the same thing?  Is it divided into nodes that do "back end" processing vs "front end" processing?  Key areas in testing will *always* be communications between nodes (networking), processes (sockets, shared memory, ...).  Will storage be shared, or kept isolated to each node?

Spandan Veggalam has made good points, one should understand the architecture of the distributed system you are looking to test so that the failure points can be identified.

The performance of distributed systems are related to several factors:

- At what level the used middleware handles the fault tolerance problem

- At what level the middleware manages the load balancing problem.

- Measuring the impact of network latency on the performance of distributed system. - Security is also an important point

- At what level it accepts heterogeneous nodes

- ...