Liuba Shrira

Modular software upgrades for distributed systems

Upgrading the software of long-lived, highly-available distributed systems is difficult. It is not possible to upgrade all the nodes in a system at once, since some nodes may be unavailable and halting the system for an upgrade is unacceptable. Instead, upgrades must happen gradually, and there may be long periods of time when different nodes run different software versions and need to communicate using incompatible protocols. We present a methodology and infrastructure that make it possible to upgrade distributed systems automatically while limiting service disruption. We introduce new ways to reason about correctness in a multi-version system. We also describe a prototype implementation that supports automatic upgrades with modest overhead.

BuddyCache: high-performance object storage for collaborative strong-consistency applications in a WAN

Collaborative applications provide a shared work environment for groups of networked clients collaborating on a common task. They require strong consistency for shared persistent data and efficient access to fine-grained objects. These properties are difficult to provide in wide area networks because of high network latency.BuddyCache is a new transactional caching approach that improves the latency of access to shared persistent objects for collaborative strong-consistency applications in high-latency network environments. The challenge is to improve performance while providing the correctness and availability properties of a transactional caching protocol in the presence of node failures and slow peers.We have implemented a BuddyCache prototype and evaluated its performance. Analytical results, confirmed by measurements of the BuddyCache prototype using the multi-user 007 benchmark indicate that for typical Internet latencies, e.g. ranging from 40 to 80 milliseconds round trip time to the storage server, peers using BuddyCache can reduce by up to 50% the latency of access to shared objects compared to accessing the remote servers directly.

BuddyCache: cache coherence for transactional peer group applications

Universal network access has enabled a new class of distributed applications that allow collaborating peers to share cached data. Until now these applications had basic limitations: sharing was limited to read-only access, networks were limited to local area networks, and there was no support for fine-grained sharing or transactions. BuddyCache is a caching system for peer applications that provides updates to shared data from remote repositories. BuddyCache is the first system to provide transactional fine-grain coherence in high-latency networks. It provides consistent, low-latency access to shared objects cached by peers when consistency management requires coordination with remote repositories over high-latency wide-area networks.

Skippy: Enabling Long-Lived Snapshots of the Long-Lived Past

Decreasing disk costs have made it practical to retain long- lived snapshots, enabling new applications that analyze past states and infer about future states. Current approaches offer no satisfactory way to organize long-lived snapshots because they disrupt the database in either short or long run. Split snapshots are a recent approach that overcomes some of the limitations. An unsolved problem has been how to support efficient application code access to arbitrarily long-lived snapshots. We describe Skippy, a new approach that solves this problem. Performance evaluation of Skippy, based on theoretical analysis and experimental measurements, indicates that the new approach is effective and efficient.

Session 5: Overview

This session contained three papers loosely centred around low level system support for persistent computation.

Transactional Remote Group Caching in Distributed Object Systems

Peer group computing is an increasingly popular class of distributed applications enabled by universal access to global networks. The applications allow collaborating peers to share cached data, but up to now only read-only applications were supported. This paper describes BuddyCache, a caching architecture for peer group applications updating shared data. Earlier group caching approaches only worked in local area networks or did not support fine-grain coherence and transactions. BuddyCache coherence protocol is the first transactional fine-grain group coherence protocol for object repositories in high-latency networks. The main challenge in BuddyCache is how to provide low-latency access to consistent shared objects cached by peers when consistency management requires high-latency coordination with remote servers accessed over wide-area networks.