Chi H. Ho

Ph.D. Student
Department of Computer Science
Cornell University

Office: 457A Rhodes Hall
Phone: (607) 255-7203
Email: chho@cs.cornell.edu

Education:

The University of Texas at Austin

Bachelor of Science in Computer Sciences, Highest Honors and Special Departmental Honors

Bachelor of Science in Mathematics, Highest Honors

Austin Community College

Truong Pho Thong Nang Khieu

Highschool Diploma.

Research Interests:

I am interested in systems, especially in their reliability, scalability, and performance.

Publications:

Nysiad: Practical Protocol Transformation to Tolerate Byzantine Failures (NSDI 2008).

(with Robbert van Renesse, Mark Bickford, and Danny Dolev.)

Making Distributed Applications Robust (OPODIS 2007).

(with Danny Dolev and Robbert van Renesse.)

A Method for Introducing Fault Tolerance to Random Network Failures

In Parallel Programming Libraries (undergraduate honors thesis, UT-Austin, 2005).

In Preparation:

RINGS: Reliable Interactive Group Sessions.

Course Research Projects:

Scaling Attraction Force Computation (research project for course Advanced Database, 2007).

Abstract: In many games and simulations, objects can assert attraction forces on one another from a far distance. Traditionally, the brute-force approach is used to compute the total force on each object. This approach is precise, but it scales too poorly to be used in practice when the number of objects is large. In this paper, we present a heuristic algorithm to compute total attraction forces. The heuristic is exponentially faster than brute-force, while maintaining a low average error. Moreover, the heuristic algorithm allows the user to control the maximum error by trading off some performance.

Simple Message Logging Protocol (research project for course Advanced Systems, 2005).

Abstract: This paper presents a simple optimistic message-logging protocol that works without any guarantees about message delivery. Unlike traditional optimistic message-logging protocols, the protocol does not suffer from domino effect and multiple rollbacks. Furthermore, the protocol achieves optimal recovery and simple garbage collection, while imposing light overhead on error-free execution—which is almost as little as uncoordinated checkpointing.

Other Research:

Abstract: Critical systems should tolerate both disasters and Byzantine failures. The replicated state machine approach can provide this high degree of reliability and availability, but these protocols use several communication steps and these can become a performance concern when long-distance links are used, as is necessarily the case for disaster-tolerance. We propose Demeter, a new disaster-tolerant replicated state machine that uses the optimal number of replication sites and execution nodes, and is optimized for using as few long-distance communication steps as possible. We designate one of the locations as the main locations, and clients operating from this location can complete read or write operations in a single WAN round-trip