Wiki
Clone wikignrs / Minutes 2012-10-04
Status Report
- Demonstrated initial version of k-replica insertion/lookup for GNRS server.
- Server code needs significant clean-up. Rob and Feixiong will review the code, spending a day on a quick-and-dirty refactor. Long-term rewrite will be discussed later.
Discussion
- In MF meeting, privacy/security-performance trade-offs are really important. We should offer multiple models for privacy/security possibilities and evaluate them.
- If all of the GUID mappings/requests are in all of the ASs, then "everyone" knows where you are. For high-mobility, perhaps this is a reasonable exchange, but it may be that GNRS performance will suffer if greater privacy is desired.
- Message semantics need to be defined. Example: update(OLD, NEW)
- What happens when we "replace/update" an entry that isn't present?
- How do we "refresh" a TTL for a mapping? Do we reuse Insert or do we have a separate Renew message?
New Tasks
- Come up with request/message scenarios? What are the most likely situations?
- Define some proposed semantics for various messages and the interactions.
- Need to get our hands on a workload/network trace that we can use to build simulations.
- Think about a mobility model for TTL prediction.
Updated