John Lane

Department of Computer Science
The Johns Hopkins University

John Lane
313 NEB
Johns Hopkins University
Baltimore, MD 21212
johnlane at jhu dot edu

Resume

Research Interests

I am interested in large-scale distributed systems, and I currently work in the Distributed Systems and Networks Lab at Johns Hopkins University. Dr. Yair Amir is my PhD adviser. My dissertation work is on survivable replication systems that scale to wide-area networks. These systems use distributed algorithms to establish a total agreed order on updates submitted by clients. Each replica executes the same ordered stream of updates. I have also worked on a reliable multicast system that scales to large numbers of groups.

In addition to protecting against external attacks, survivability engineering focuses on preserving system functionality when security fails and an attacker gains control of one or more parts of the system. A malicious adversary may attempt to use the servers that he has compromised to disrupt the system. Servers controlled by an attacker are referred to as Byzantine. Our group developed Steward, the first Byzantine fault-tolerant replication architecture suitable for deployment in multiple local-area sites distributed across a wide-area network (i.e., Internet scale). The physical machines in each site function as a logical entity or logical machine. Each logical machine plays the role of a single participant in a benign fault-tolerant wide-area replication protocol. The machines in each site use several Byzantine fault-tolerant local protocols to agree on the content of outgoing wide-are messages. Threshold cryptography prevents malicious servers from misrepresenting a site. Steward has been tested in an eighty server deployment and it underwent a successful DARPA evaluation during which an independent team of attackers attempted to break the system. Steward survived all attempted attacks.

We have also developed a new architecture that achieves performance only slightly worse than Steward while cleanly separating the protocol used to implement the logical machine and the wide-area protocol run on top of the logical machines. One of the main benefits of this architecture is that it provides customizability of the fault-tolerance approach used within and among the sites. This architecture leverages the state-machine approach to construct logical machines and uses several optimizations to reduce wide-area bandwidth consumption and amortize the cost of computationally intensive cryptography.

Claudiu Danilov and I are developing and adding a reliable multicast protocol to Spines, an overlay network created by Claudiu and Yair Amir. Our multicast service will provide reliable FIFO ordering. We are focusing on scaling to thousands of groups each having a relatively small number of members. We hope to include this service in a future version of Spines.

Biography

I received my BA from Cornell University in 1992 with a major in Biology and a concentration in Neurobiology and Behavior. During my time at Cornell, I worked with Dr. Owen Hamill studying the effects of a series of diuretics on stretch activated ion channels using patch-clamping. After graduation, I went to Johns Hopkins University to work in a systems neuroscience lab run by Dr. Ken Johnson and Dr. Steven Hsiao at the Krieger Mind Brain Institute. They study the somatosensory (tactile sensory) system in both humans and animals, and specialize in the development of complex data collection systems and robotic tactile stimulators. I transitioned from conducting experimental work to designing and building data analysis tools and research infrastructure. In 2003, I entered the PhD program in Computer Science at Johns Hopkins.

Education

Computer Science Publications

Other Selected Publications