Latency-optimal fault-tolerant replication

Add to your list(s) Download to your calendar using vCal

• Piotr Zielinski, Inference Group
• Wednesday 01 February 2006, 15:00-16:00
• Committee Room, Cavendish Laboratory.

If you have a question about this talk, please contact Phil Cowans.

In this talk, I will present some of my PhD research in distributed agreement protocols done at the Computer Laboratory, Cambridge.

Slides: http://www.cl.cam.ac.uk/_{pz215/files/disc.pdf http://www.cl.cam.ac.uk/}pz215/files/msgeneric.pdf

To improve the fault-tolerance of a distributed system, a central server can be replaced with many identical replicas. Client requests are then broadcast to and executed by all these. To remain consistent, all replicas must execute client requests in the same order.

In a simple approach, clients send their requests to a distinguished replica, which then broadcasts them in some order to all others. This algorithm is fast; a client request reaches all replicas in only two communication steps. Unfortunately, it is not fault tolerant; a crash of the distinguished replica will block the entire system. Fault-tolerant solutions to this problem exist, but they all require at least three steps, even when no failures occur.

In this talk, I will present new algorithms that combine the resistance of fault-tolerant solutions with the efficiency of fault-intolerant ones. Two facts are exploited: (i) requests from different clients are usually independent, and (ii) networks normally deliver messages to all recipients in the same order. In the absence of failures, if one the above conditions holds, these algorithms require only two steps, and three otherwise.

Keywords: Atomic Broadcast, Generic Broadcast, optimistic techniques, replication, fault-tolerance, Consensus

This talk is part of the Inference Group series.

This talk is included in these lists:

• All Cavendish Laboratory Seminars
• All Talks (aka the CURE list)
• Biology
• Cambridge Neuroscience Seminars
• Cambridge talks
• Centre for Health Leadership and Enterprise
• Chris Davis' list
• Committee Room, Cavendish Laboratory
• Featured lists
• Guy Emerson's list
• Hanchen DaDaDash
• Inference Group
• Inference Group Summary
• Interested Talks
• Joint Machine Learning Seminars
• Life Science
• Life Sciences
• ME Seminar
• ML
• Machine Learning Summary
• Neurons, Fake News, DNA and your iPhone: The Mathematics of Information
• Neuroscience
• Neuroscience Seminars
• Neuroscience Seminars
• Required lists for MLG
• School of Physical Sciences
• Stem Cells & Regenerative Medicine
• Thin Film Magnetic Talks
• dh539
• rp587
• yk373's list

Note that ex-directory lists are not shown.