MFS-даунтаймы в крупных сервисах

• Google App Engine Incident #19007
• Amazon SimpleDB Service Disruption
• Cassandra Overload because of hint pressure + MVs

Принцип constant work

• Reliability, constant work, and a good cup of coffee

Как работать с ретраями

• Fixing retries with token buckets and circuit breakers
• What is Backoff For?

Полезные статьи про MFS

• Metastable Failures in Distributed Systems
• Metastable Failures in the Wild
• Metastability and Distributed Systems