SI2-SSE: TestRig 2.0

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

Advances in computation, modeling, and digital collection have led to an exponential growth in the size of scientific data sets. This 'Big Data' is transforming the process of scientific discovery. However, many scientific workflows require that the data be transferred, via networks like the Internet, to an optimal location for analysis and collaboration. Unfortunately, scientists often encounter performance problems on these networks. To resolve these problems, researchers depend on the expertise of engineers at Network Operations Centers (NOCs) for diagnosis and resolution. Resolving these problems often requires a cycle of interactions between the user and engineer that can last for days if not weeks, seriously hindering the scientific workflow. This project, known as TestRig 2.0, will short circuit this cycle by deploying a simple, easy to use system that will automatically collect a wide variety of important network diagnostic information. TestRig 2.0 will distribute dynamically generated test environments that are tuned to the specific needs of NOC engineers. TestRig 2.0 will automatically runs a series of tests against the infrastructure, collect the results, and transfer them back to the NOC engineer. TestRig 2.0 will also conducted a coarse analysis of the collected data so as to give the engineers some initial insight into the collected metrics. In this project, researchers at Pittsburgh Supercomputing Center (PSC) propose to build a system, TestRig 2.0, that will quickly and easily gather a wide range of important network diagnostic information. This system includes a centralized management server that dynamically creates and configures client optical disk image files (as defined by ISO 9660). This server will be at PSC; it will provide TestRig 2.0 services to multiple geographically diverse NOCs via a free subscription model. These bootable images will provide a known good TCP stack configuration and system environment. TestRig 2.0 will use this baseline environment to perform tests against the perfSONAR monitoring infrastructure and also collect other pertinent data. The resulting data will undergo an initial automated coarse analysis by the client in order to provide engineers initial insight into the data. The resulting analysis and raw data will then be packaged and transferred back to the appropriate NOC for analysis. The client ISO creation process will include a unique dynamically generated public/private key pair. This key pair will be used for the authentication of the client ISO, distribution management, and user/test management. The rapid collection of pertinent data will enhance the process of scientific discovery by allowing researchers to focus on their research, instead of running tests for network engineers. By providing a framework for the use and development of post-collection data analysis, the system will lower the bar required for effective network diagnostics across a wide range of support personnel.