Will this benchmark measure energy and/or energy efficiency?

This is being discussed in the working group. 

Will this benchmark run on really small clusters?

Yes. You'll be able to run with a small number of nodes in a cluster, but it may not be a very representative scenario. For instance, in memcached type workloads, one node will get dedicated as the server and the other as a client and therefore limit the number of clients a server may be able to serve.

What architectures will this benchmark run on?

We want to make this as hardware agnostic as possible, therefore, there should be no limitation to the architecture that this benchmark can be run on. That said, it will definitely run on all the major processor architectures.

Will this benchmark only measure latency and throughput as two values or as a curve with different sizes of datasets?

What gets measured and reported will depend on the benchmark workload. For memcached, we will report latency and transactions per second. For graph analytics, it may be time to completion (or convergence) of the graph.

Currently, we plan to have a single dataset per benchmark for the official scoring version. But we want to architect the benchmark such that the end user can run any number of datasets of varying sizes.

Which operation systems will be supported?

Although it's not 100% concrete at this time, we are targeting major Linux distributions for now. Fedora Core, Ubuntu, etc.

What is the target release date for the EEMBC ScaleMark Benchmark Suite?

We are targeting end of Q4 for availability for EEMBC members for final testing. Our goal is to make this publicly available in the Q1, 2016 timeframe.