The Dependable Computing Systems Laboratory (DCSL) at Purdue University investigates the question of how to build dependable, heterogeneous, large-scale distributed systems.
“Dependability meets Data Analytics, and at Large Scales”
The above sums up our current research direction. We work on software systems to enable them to perform their functionality in the face of natural and malicious failures. We apply and adapt data analytic techniques to work with the noise of computer systems and at large system scales. Current application domains come from distributed software systems, embedded systems, cellular systems, and bioinformatics.
Since many business and life critical functions are being performed by distributed systems, they need to be dependable while meeting their performance goals. Thus, there is need for smart error detection, diagnosis, and recovery protocols. Since many of these systems operate on vast amounts of data and the patterns of errors or normal operation are approximate and noisy, we have to adapt leading-edge machine learning tools to these systems problems. There is also need for architectures that can combine dependability and security aspects without significantly degrading performance and do this in an adaptive manner, adapting to different user requirements and different runtime environments. This is our mission at DCSL.
Our application contexts come from various domains, many from our industrial colleagues. These include: security-critical enterprise (with Missile Defense Agency, Northrop Grumman and Lockheed Martin), mobile and cloud platforms (in collaboration with AT&T and IBM), large-scale scientific clusters and applications (in collaboration with Lawrence Livermore National Lab and Argonne National Lab), and cyber physical systems (in collaboration with GE Global Research Center and Sandia).
Our goal is to do cutting edge practical systems work that makes a splash in the academic community and a difference in the wider technology sphere through strategic collaborations with industrial partners. DCSL is the founding lab within the Purdue College of Engineering Center for Resilient Infrastructures, Systems, and Processes (CRISP).
- May-June 2019: Our work with Department of Energy/Lawrence Livermore National Lab (DOE/LLNL) on approximating scientific computation through reducing precision of floating point variables gets accepted into ICS (International Conference on Supercomputing). Acceptance rate was 45/193 = 23.3%. The work shows how you can quickly find which variables to reduce precision of and to what extent, while bounding the accuracy loss. [ PDF ] [ WWW ]
A related piece of work won the best paper award at the International Supercomputing Conference (ISC) held in Frankfurt, Germany, June 16-20. [ HPCWire news story ]
- May 2019: Our center SMART-WHIN had its industry day on May 16 when all our industrial sponsors came to visit. DCSL graduate researchers, Heng Zhang, Mustafa Abdallah, and Edgardo Barsallo presented posters and a demo on their work on reliable wireless networking with IoT devices on farms and industrial manufacturing facilities. Their work also showed how to do in-sensor analytics and visualization of the data for all our industrial stakeholders.
- Update: [April 2019] The camera ready paper is now released. [ PDF ]