hesela.devStorage availability
Storage availability is the ability of a storage service to provide the required data access to a specified client under stated operating conditions.
A preserved copy does not ensure an operational path from every host. Specify the client, operation and tolerated failures before making an availability claim. [1]
Availability percentages need a measurement contract. The cited object-storage metric averages five-minute eligible-error ratios, rather than simply dividing all errors by all requests; its idle-interval convention also matters. [2]
The companion analysis gives two synthetic traffic patterns with the same total requests and errors but different interval-weighted results. Neither result establishes outage duration, end-to-end application success or a durability probability.
A status-based metric does not by itself establish completion before an application's deadline; evaluate the latency distribution separately. [3]
Relations
- related
- Data durability, Failure domain, Recovery time objective, I/O timeout, Fail-slow, Tail latency
Citations
- SNIA: NVM PM Remote Access for High Availability, May 2019, section 4.1, pp. 13-15 (non-normative whitepaper)
- Amazon S3 Service Level Agreement, updated 28 November 2023, Definitions
- Dean and Barroso: The Tail at Scale, CACM 56(2), 2013, pp. 74-80, DOI 10.1145/2408776.2408794
Sources & caveats
- The synthetic comparison is arithmetic, not service telemetry or a credit-eligibility assessment. No aggregation across different request types is inferred.