Infrastructure

Stress testing of components helps define operational limits of the system and can identify areas of inefficiency.

Typical read operations:

• Each of the MN implementations should be evaluated, with emphasis on high access operations such as listObjects, getSystemMetadata, getChecksum, and get

• CN throughput for user facing operations should be tested - resolve, getSystemMetadata, get, isAuthorized, listObjects, search, getLogRecords, hasReservation

Typical CN operations:

• What is the rate of change to system metadata that can be sustained across three coordinating nodes?

• What is the rate of synchronization of new content that can be sustained?

• What is the typical range of latency for new science metadata to be fully propagated across CNs?

Should also consider failures - stress test until failure and observe how graceful or otherwise thing turn out.