The SunSCALR Framework for Internet Servers

Internet servers need to be highly-available, inexpensive, and scalable. These goals are often connicting and most designs meet, with limited success, only few of them. In this paper we describe the SunSCALR framework that achieves these goals by combining proven technologies, careful system design, and engineering trade-oos. It uses a distributed, self-stabilizing algorithm for status monitoring and failure detection, and IP failover for automatic re-connguration. SunSCALR provides high-availability against message loss, host crashes, and scheduled downtime, and allows on-they addition and removal of hosts. We present detailed performance of SunSCALR. It can provide 10 second failover latency (i.e., better than 99.999% availability if machines fail for 2 hours/month). SunSCALR based products have been in use within Sun and are also available in the market.