Center for Research in Intelligent Storage (CRIS)

Temple University

CRIS is an NSF supported Industry-University Cooperative Research Center (IUCRC) focusing on the crucial issues of configuration and energy management in storage systems and more generally in data centers. Following the normal IUCRC model, research projects in CRIS are funded entirely by the industry members and are expressly designed to fulfill precompetitive research needs of the supporting members.

CRIS is currently supported by Intel, HP Enterprise, and Dell-EMC. The currently defined projects relate to intelligent storage caching, configuration management of storage gateways, hinting in storage systems, coordinated network and storage management, and access predictions across a storage hierarchy. Recently the focus of the IUCRC has been broadened to consider compute, storage and networking issues in hyperconverged systems in both data center and edge computing contexts.

CRIS Faculty:

Site Director: Dr. Krishna Kant, http://www.kkant.net

CRIS Students:

Current: Jit Gupta, Tanaya Roy, Lu Pang, Joyanta Biswas, Pavana Pradeep

Former: Madhurima Ray, Anis Alazzawe, Dusan Ramljak, Ibrahim El-Shekeil, Malek Athamnah

CRIS History:

CRIS is a site established at the Computer and Information Science (CIS) department of Temple University for an IUCRC at the Computer Science Department at University of Minnesota. The latter IUCRC is called CRIS (Center for Research in Intelligent Storage) and has been focused on storage system issues. Details on CRIS may be found at http://cris.cs.umn.edu. It is directed by Prof. David Du and is currently being supported by 10 industry members (http://cris.cs.umn.edu/sponsorlist.htm). Recently CRIS has transitioned to phase II, where it is augmented by two new sites, one at Temple University, directed by Prof Krishna Kant (http://www.kkant.net) and the other at Texas A&M University, directed by Prof. Narasimha Reddy (http://www.ece.tamu.edu/~reddy/). According to IUCRC charter, each new site must necessarily expand the scope of the research. The Temple University site is focused on the crucial issues of energy and configuration management in large scale storage systems and beyond.

CRIS Scope:

The CRIS site’s interests are also rooted in storage systems, although it has specific interests in the areas of energy and configuration management. In the past, our projects have largely been focused on intelligent caching that exploit machine learning on data access patterns. Other projects include an ongoing effort to predict accesses to storage hierarchy, again based on machine learning techniques, in order to facilitate intelligent tiering. Another project has examined the interplay between data center storage and network to proactively avoid network congestion that could result due to high intensity accesses to NVMe SSDs and other high-speed storage technologies such as Intel Optane that are becoming very common and can easily overwhelm the network. At the same time, the project also attempts to avoid scattering of active storage chunks throughout the data center so as to allow intelligent management of network energy during low and normal usage periods. Another project concerns the configuration management and optimization of cloud storage gateway where a small amount of local storage backed by remote cloud storage is used give impression of unlimited local storage. Yet another ongoing project concerns implementation of hinting in storage systems so that it si possible to implement end to end QoS in block storage systems. A final project examines the issue of quantifying “health-index” of various resources in a data center based on their configurations.

Overview of IUCRC Program:

Full details of IUCRC program are available on the NSF site (http://www.nsf.gov/eng/iip/iucrc/). IUCRC is an extremely successful program for bringing together industry and academia for conducting industry-centric pre-competitive research by academia. It works like a consortium and funded by memberships from industry. The membership fee directly supports the projects that the funding members help define. The main advantages of the IUCRC model to industry include the following:

(a) Availability of results from all projects supported under the umbrella IUCRC to all members. This has a great leveraging effect on the membership fee from any given member.

(b) Availability of academic expertise and resources (Ph.D. students & postdocs) to conduct industry specific precompetitive research.

(c) Grooming future employees for the member companies through close working relationship on IUCRC projects.

(d) Contacts and collaborations among member companies.

(e) NSF limits the overhead of membership money to 10%, which means that 90% of the money can be used to support projects.

According to NSF rules, the nominal IUCRC membership is $50K per year, but other arrangements are possible. NSF requires a minimum level of member supported funding for a center. The membership funds are committed annually, although there is generally an expectation (but not an obligation) that a member will stay with the center for at least 2 years.