This preview shows page 1. Sign up to view the full content.
Unformatted text preview: k with eventual consistency? Switch gears to implementation techniques for cache coherent memory Table of: write through/write back vs. coherence: none, TTL-‐based, callback 1) no caching (very early dfs, novell) 2) write through, TTL (DNS, web) 3) write back, TTL (NFS) 4) write through, coherent (AFS) 5) write back, coherent (coda, ivy) Illustrate behavior of each quadrant: Start simple. No caches – what if every RPC goes to server? What semantics does that provide? (linearizability) TTL – time to live. Allow client to use copy for some period of time. After TTL, invalidate the cached copy, so the client goes back to the server to get the latest version. What semantics does this provide? (eventual) One tremendous advantage to TTL cache coherence – no state needed at the server. The server does not need to keep track of who has a copy, since they will each time out in turn. [Anecdote: DNS uses TTL cache coherence, but client checks only when name is used. USGS web site becomes very slow, every time an earthquake hits California, beca...
View Full Document
- Spring '14