Research Overview

The Scalable QoS Provision Architecture project aims to design a scalable, differentiated services WAN architecture for QoS-sensitive applications with elastic requirements using aggregated flow control and class-based label switching. From a theoretical perspective, our work provides a theory of aggregate-flow scheduling which complements the well-known per-flow theories. Akin to real-time scheduling theory showing the existence of optimal real-time schedulers (e.g., EDF, RMS), aggregate-flow theory shows that there are optimal aggregate-flow schedulers which provide the theoretical foundation for differentiated services. The Purdue Infobahn QoS Testbed is a private IP-over-SONET backbone comprised of nine Cisco 7206 VXR routers. Dr. Park has agreements with Cisco which allows IOS -- Cisco's router software -- to be modified to implement the class-based label switching algorithms for IPv4 (and IPv6) developed in the Network Systems Lab.

• H. Ren and K. Park. Toward a theory of differentiated services. In Proc. IEEE/IFIP IWQoS, pp. 211-220, 2000.
• H. Ren and K. Park. On the QoS provisioning power of optimal aggregate-flow scheduling. Technical Report CSD-TR-00-021, Purdue University, Dept. of Computer Sciences, November, 2000.

The Multimedia and Self-Similar Traffic Control project designs next generation end-to-end Internet protocols for QoS-sensitive transport of real-time -- MPEG I & II video/audio -- interactive, and bulk data traffic (file transfers). The traffic controls which follow the multiple time scale traffic control framework achieve significant performance improvement by exploiting large time scale predictability structure present in self-similar and heavy-tailed Internet workloads. The multiple time scale traffic control framework is able to mitigate the reactive cost of feedback traffic controls in broadband wide area networks characterized by a high delay-bandwidth product by engaging predictability information exceeding the time scale of the round-trip time. These protocols have been implemented both for TCP and UDP, and run over UNIX and Windows NT.

• K. Park and T. Tuan. Performance evaluation of multiple time scale TCP under self-similar traffic conditions. To appear in ACM Transactions on Modeling and Computer Simulation, 24:152-177, 2000.
• T. Tuan and K. Park. Multiple time scale redundancy control for QoS-sensitive transport of real-time traffic. In Proc. IEEE INFOCOM '00, pp. 1683-1692, 2000.

The DUNES project aims to provide seemless distributed scheduling by transparent load balancing, pull/push caching, communication-sensitive load balancing, and distributed real-time scheduling. QoS-sensitivity is achieved by gang scheduling which decides where -- load balancing -- and how much -- real-time scheduling -- a task is allocated with respect to available distributed resources. DUNES is implemented as an extension of Solaris UNIX (portable to Linux). Our recent work allows us to characterize distributed system workloads that are most amenable to communication-sensitive load balancing.

• J. Cruz and K. Park. Towards performance-driven system support for distributed computing in clustered environments. Journal of Parallel and Distributed Computing, 59:132-154, 1999.
• J. Cruz and K. Park. Towards communication-sensitive load balancing. In Proc. IEEE International Conference on Distributed Computing Systems, pp. 731-734, 2001.

The Adaptive Network Security and Fault-Tolerance project integrates network security services with efficiency and QoS such that processing overhead is effectively managed and a user-specified trade-off is achieved. Critical security services (e.g., confidentiality, authentication) -- user programmable -- are proactively affected, while other user-selected services (e.g., intrusion detection, DoS monitoring) are reactively handled. AdSec is a prototype system built on top of SNMP using distributed agent technology which combines proactive and reactive service protection. More recent work includes scalable solutions to distributed denial of service attack prevention based on a novel approach called route-based distributed packet filtering.

• K. Park and H. Lee. On the effectiveness of route-based packet filtering for distributed DoS attack prevention in power-law internets. In Proc. ACM SIGCOMM '01, pp. 15-26, 2001.
• K. Park and H. Lee. On the effectiveness of probabilistic packet marking for IP traceback under denial of service attack. In Proc. IEEE INFOCOM '01, pp. 338-347, 2001.

The research projects are supported by several grants and funding assistance from government and industry (NSF ANI-9714707, ANI-9875789 [CAREER], ANI-0082861 [ITR], ESS-9806741, EIA-9972883, DARPA ATO FTN, PRF, Santa Fe Institute, Sprint, Xerox), including equipment and software support (Cisco, FORE, and Intel).

Recent Publications   ||   Project Publications