EXERT 2012

"Exascale Systems, Mega Datacenters, Warehouse-size Computers" - all these buzzwords describe the massive compute infrastructure that has fueled the meteoric rise of Amazon, Google, Facebook, Yahoo!, and other cloud computing/Internet services and enabled unprecedented scientific discoveries through high-performance computing.  Designers of the next-generation of data intensive applications and large-scale compute infrastructure face enormous challenges (improving performance, guaranteeing quality of service, using energy efficiently, provisioning power, maintaining reliability, controlling temperature, ensuring manageability, etc.)  Modern computer system design is built on a framework of quantitative evaluation and research methods, such as performance monitoring, workload characterization, analytic modeling, detailed simulation, and prototype evaluation.  However, the methods, tools and techniques we use to evaluate a single component, system, or cluster do not scale to the massive scope of evaluating quantitatively warehouse-sized systems.  Moreover, because these facilities cost millions and execute organizations mission-critical applications, it is difficult for researchers to gain access to production facilities and real workloads to inform their efforts.

The Exascale Evaluation and Research Techniques Workshop (EXERT) is a forum to stimulate further cutting-edge research in quantitative design, evaluation, and research methods for large-scale systems and enable an exchange of tools, methods, data, and experience gained in real facilities. The list of topics of interest to the EXERT workshop, includes (but is not limited to):

- Analytic modeling and simulation of extreme scale systems

- Benchmark suites for internet services and cloud computing workloads

-  Tracing and characterization of large-scale workloads

- Cost, TCO, reliability, and lifecycle analysis techniques

- System- and facility-level thermo-mechanical design and evaluation methods

- Performance/energy/temperature measurement and analysis tools

- Tools and techniques for facility-scale monitoring and instrumentation

- Scalable/parallel "cluster-on-a-cluster" prototyping/simulation techniques

-  Statistical methods to extrapolate results from scaled-down experiments

- Data, characterizations, or experience reports from production system deployments

In addition to original research articles, EXERT invites experience reports documenting new tools, measurements of production systems, or public data sets and benchmark suites.  Experience reports will be judged on their potential usefulness to the community rather than their novel research contribution.

Submission Guidelines:

Paper Submission Date: EXTENDED Thursday, January 5th, 2012

Papers should be double column, single-spaced format and up to six pages in length