In: Computer Science
What is the defining characteristic of a transient server and how does it affect application design?
The defining characteristic of a transient server is that it may terminate any time after a brief advance warning period. Transiency has a number of unique characteristics that differentiates it from conventional fault tolerance. First, transient servers can become unavailable at a much higher rate than conventional servers, since their availability is controlled by the operational policies of the data-center or the cloud. Second, transient server unavailability is often preceded by an advance warning. And third, transient servers may be available with different cost/availability tradeoffs, requiring the use of 1 careful server selection using data-driven techniques. Once a transient server receives an advance warning, a data center must move any VMs (and their associated state) hosted on the transient server to a stable server to maintain their availability. Transient server revocations can cause loss of application-state, which can result in application downtimes, degraded performance due to failure recovery, and end-user dissatisfaction in general.