Computer Systems and Networking

Bridging the gap between Computer Science and Electrical Engineering is the tangible hardware on which computation occurs and the tangible networks over which information flows. Research on modern computer systems and data networks concentrates primarily on speed, efficiency, and bandwidth, yet must also address the interface of the hardware to higher-level software, such as operating systems, database systems and compilers. And, of course, it must also be correct; verification of proper behavior is an essential part of computer systems design.

An increasingly popular target platform for many programming environments is clusters of workstations. Traditionally, resources on such systems are managed by operating system code. However, over the past few years, there has been sufficient evidence to indicate that efficient utilization of key resources, such as the network and the secondary storage, cannot be obtained through application-independent operating system code. Arvind Krishnamurthy is exploring an alternative strategy to resource management whereby the compiler and the operating system share the responsibility. While the operating system is well suited for implementing low-level mechanisms, the compiler is capable of generating application-specific code for managing resources since it has access to program level behavior. With the compiler and system software interfacing through a set of well-defined operations, the design provides a high degree of adaptability to changes in machine parameters and application characteristics. Database systems provide an environment for storage and retrieval of both structured and semi-structured data. Such systems were originally designed for use in business-type applications. Today, however, they are being utilized in many other application domains, including scientific computing, networking, bioinformatics, etc. Avi Silberschatz has been working in the field of databases exploring such topics as transaction management, real-time systems, high-performance databases, multimedia systems, approximate queries, and data mining.

Data networks allow computers to communicate with one another, and of course form the backbone of the Internet. Networks are becoming increasingly complex as the needs for speed, bandwidth, robustness, and security increase. Moreover, the complexity of these networks requires new network management techniques. Yang Richard Yang’s research focuses on the problem of congestion in large networks. Much like congestion on real highways, network congestion can be avoided by re-routing data through less congested portions of a network. But to do so requires designing networks that are more intelligent than conventional telephone switches, for example. Yang’s research is especially important in emerging multimedia applications as well as wireless and mobile networks. Avi Silberschatz has been concentrating on network management issues. In particular, the design and development of a new software system that automatically discovers the topology of Internet Protocol (IP) networks, their elements, and physical interconnections among them. Detailed knowledge of network elements and their relationship with one another is crucial for efficient network management tasks like fault isolation and root-cause analysis. Computer architecture research at Yale focuses on mechanisms for achieving high performance for a wide variety of programs. Mechanisms for running traditional C programs in parallel, as well as for running higher-level languages such as Java and Haskell, are studied.

At a lower level, Yiorgos Makris studies methods for testing and verifying the correctness of complex digital circuits. Key to his research are hierarchical methods for identifying and utilizing transparency behavior in support of reliable integrated circuit design. Of particular interest is application of these ideas to methodologies for quickly generating complex digital circuits, such as "systems-on-chip" (SOC) and "field-programmable-gate-arrays" (FPGA), and for hierarchical design validation and testing of micro-electro-mechanical systems (MEMS).

Faculty members in the Computer Systems and Networking area are Arvind Krishnamurthy, Yiorgos Makris, Avi Silberschatz, and Yang Richard Yang.