- AboutThis should describe the systems research collaboration, and present the overall research goals of the new group.
- PeopleHere are the different labs in the SRC…
- PublicationsA page where you will find categorized publications!
- ProjectsA page where you will find our projects
- ResourcesVarious resources for prospective students, current students, alumni. Maybe put something here about life in NYC and at Columbia…
Proceedings of the 10th Annual ACM International Conference on Mobile Computing and Networking (MobiCom 2004), September 2004
We present MobiDesk, a mobile virtual desktop comput- ing hosting infrastructure that leverages continued improve- ments in network speed, cost, and ubiquity to address the complexity, cost, and mobility limitations of todayâ€™s per- sonal computing infrastructure. MobiDesk transparently virtualizes a userâ€™s computing session by abstracting under- lying system resources in three key areas: display, operating system, and network. It provides a thin virtualization layer that decouples a userâ€™s computing session from any particu- lar end-user device, and moves all application logic to host- ing providers. The virtualization layer decouples a userâ€™s computing session from the underlying operating system and server instance, enabling high-availability service by trans- parently migrating sessions from one server to another dur- ing server maintenance or upgrades. We have implemented a prototype in Linux that works with existing unmodified ap- plications and operating system kernels. Our experimental results demonstrate that MobiDesk has very low virtualiza- tion overhead, can provide a full featured desktop experience including full-motion video support, and is able to migrate usersâ€™ sessions efficiently and reliably for high-availability, while maintaining existing network connections.
Proceedings of the 11th World Conference on Medical Informatics (Medinfo 2004), September 2004
Organizations are beginning to recognize that health care providers are highly mobile and optimal care requires provid- ing access to a large and dynamic body of information wher- ever the provider and patient are. Remote display protocols (RDP) are one way that organizations are using to deliver healthcare applications to mobile users. While many organi- zations have begun to use RDPs to deliver real-time access to health care information to clinicians, little formal work has been done to evaluate the performance or the effectiveness of thin-client computing with health care applications. This study examines the performance of wireless thin-client tablets with two web-based clinical applications, a text-centric, graphics-poor EMR and a graphic-rich image analysis pro- gram. The study compares the performance of two popular RDP implementations, Citrix and Microsoft Remote Desktop, with the performance of a traditional web browser in a wire- less environment. For both applications, the RDPs demon- strated both higher speed and reduced bandwidth require- ments than the web browser.
Proceedings of the 10th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD 2004), August 2004
This paper proposes a novel Data Envelopment Analysis (DEA) based approach for model combination. We first prove that for the 2-class classification problems DEA models identify the same convex hull as the popular ROC analysis used for model combination. For general k-class classifiers, we then develop a DEA-based method to combine multiple classifiers. Experiments show that the method outperforms other benchmark methods and suggest that DEA can be a promising tool for model combination.
Department of Computer Science, Columbia University Technical Report , CUCS-027-04, July 2004
Rapid improvements in network bandwidth, cost, and ubiq- uity combined with the security hazards and high total cost of ownership of personal computers have created a growing mar- ket for thin-client computing. We introduce THINC, a remote display system architecture for high-performance thin-client computing in both LAN and WAN environments. THINC transparently maps high-level application display calls to a few simple low-level commands which can be implemented easily and efficiently. THINC introduces a number of novel latency-sensitive optimization techniques, including offscreen drawing awareness, command buffering and scheduling, non- blocking display operation, native video support, and server- side screen scaling. We have implemented THINC in an XFree86/Linux environment and compared its performance with other popular approaches, including Citrix MetaFrame, Microsoft Terminal Services, SunRay, VNC, and X. Our ex- perimental results on web and video applications demonstrate that THINC can be as much as five times faster than traditional thin-client systems in high latency network environments and is capable of playing full-screen video at full frame rate.
Group Ratio Round-Robin: O(1) Proportional Share Scheduling for Uniprocessor and Multiprocessor Systems
Department of Computer Science, Columbia University Technical Report , CUCS-028-04, July 2004
Proportional share resource management provides a flexi- ble and useful abstraction for multiplexing time-shared re- sources. We present Group Ratio Round-Robin (GR3 ), the first proportional share scheduler that combines accu- rate proportional fairness scheduling behavior with O(1) scheduling overhead on both uniprocessor and multipro- cessor systems. GR3 uses a novel client grouping strat- egy to organize clients into groups of similar processor allocations which can be more easily scheduled. Using this grouping strategy, GR3 combines the benefits of low overhead round-robin execution with a novel ratio-based scheduling algorithm. GR3 can provide fairness within a constant factor of the ideal generalized processor shar- ing model for client weights with a fixed upper bound and preserves its fairness properties on multiprocessor systems. We have implemented GR3 in Linux and measured its per- formance against other schedulers commonly used in re- search and practice, including the standard Linux sched- uler, Weighted Fair Queueing, Virtual-Time Round-Robin, and Smoothed Round-Robin. Our experimental results demonstrate that GR3 can provide much lower scheduling overhead and much better scheduling accuracy in practice than these other approaches.
Ottawa Linux Symposium 2004 (OLS 04), July 2004
Ph.D. Thesis, Department of Computer Science, Columbia University, June 2004
Proportional share resource management provides a flexible and useful abstraction for multiplexing time-shared resources. We present Group Ratio Round-Robin (GR3), a proportional share scheduler that combines accurate proportional fairness scheduling behavior with O(1) scheduling overhead. GR3 uses a novel client grouping strategy to organize clients into groups of similar processor allocations which can be more easily scheduled. Using this grouping strategy, GR3 combines the benefits of low overhead round-robin execution with a novel ratio-based scheduling algorithm. We analyze the behavior of GR3 and show that it can provide fairness within a constant factor of the ideal generalized processor sharing model for client weights with a fixed upper bound. GR3 can be easily implemented using simple data structures. We have implemented GR3 in Linux and measured its performance against other schedulers commonly used in research and practice, including the standard Linux scheduler, Weighted Fair Queueing, Virtual-Time Round-Robin, and Smoothed Round-Robin. Our experimental results demonstrate that GR3 can provide much lower schedul- ing overhead and much better scheduling accuracy in practice than these other ap- proaches.
Proceedings of the 1st International Conference on Autonomic Computing (ICAC 2004), May 2004
Complexity reduction in workload management is driv- ing the development of goal-oriented workload managers (WLMs). Simultaneously, server consolidation of work- loads with dynamically changing resource demands calls for these WLMs to be increasingly efficient in manag- ing resources. We propose the Class-based Kernel Re- source Management (CKRM) framework, implemented in Linux, for operating systems to these requirements.
Proceedings of the 13th International World Wide Web Conference (WWW 2004), May 2004
Web applications are becoming increasingly popular for mobile wireless PDAs. However, web browsing on these systems can be quite slow. An alternative approach is handheld thin-client com- puting, in which the web browser and associated application logic run on a server, which then sends simple screen updates to the PDA for display. To assess the viability of this thin-client approach, we compare the web browsing performance of thin clients against fat clients that run the web browser locally on a PDA. Our results show that thin clients can provide better web browsing performance com- pared to fat clients, both in terms of speed and ability to correctly display web content. Surprisingly, thin clients are faster even when having to send more data over the network. We characterize and analyze different design choices in various thin-client systems and explain why these approaches can yield superior web browsing per- formance on mobile wireless PDAs.
Proceedings of the 1st International Workshop on Data Processing and Storage Networking: Towards Grid Computing (DPSN 2004), May 2004
We present a novel cluster architecture that unifies switch, server and storage processing to achieve a level of price-performance and simplicity of application development not achievable with current architectures. Our architecture takes advantage of the increasing disparity between storage capacity, network switching on the one hand, and processing power of modern processors and architectures on the other. We propose the use of Network Processors (NPUs), which can apply simple classify/act/forward operations on data packets at wire speeds, to split processing of operations such as complex database queries across a network. We quantify the theoretical benefits of such an architecture over traditional server-cluster approaches using warehouse database queries as a motivating application. We also discuss the challenges such an architecture presents to programming language design and implementation, performance analysis, and security.