Microreboots in Three-Tiered Internet Systems

Title: Microreboots in Three-Tiered Internet Systems

Research Question: Can microreboots, a technique for restarting fine-grained components of software systems, improve the availability of distributed interactive applications?

Methodology: The researchers used an open-source J2EE application server augmented with fault injection, instrumentation, and the ability to microreboot individual beans (application components) and functional subsystems like the Web server and Java Servlet Pages. They compared the effectiveness of microreboots to full reboots and server process restarts.

Results: The study found that microreboots are nearly as effective as full reboots but are less disruptive in terms of downtime and lost work. Microreboots reduced the number of failed user requests by 65% and perceived downtime by 78% compared to a server process restart. The researchers also showed how to replace user-visible transient failures with transparent call retry, resulting in a slight increase in end-user latency during recovery.

Implications: The use of microreboots can significantly improve the availability of applications hosted on a rich middleware platform. Since microreboots are less disruptive, they can reduce the number of end-users affected by a failure and mask some transient failures, leading to lower recovery times. This makes microreboots a promising technique for improving the availability of distributed interactive applications.

Link to Article: https://arxiv.org/abs/0403007v1 Authors: arXiv ID: 0403007v1