When: 28-29 June 2011

Where: Edinburgh 

Who:

• Mircea Markus
• Manik Surtani
• Pete Muir
• Sanne Grinovero
• Dan Berindei
• Jonathan Halliday (only Wednesday, June 29) 
• Kevin Conner  (only Wednesday, June 29)
• Paolo Romano (INESC-ID, via Skype, Wednesday)
• Sebastiano Peluso (INESC-ID, via Skype, Wednesday)

Agenda

All times in British Standard Time (GMT + 1)

Tue (28th) 

• Infinispan 5.0
  • Where we are, what we need to finish this up
• Infinispan 5.1
  • Roadmap
  • Timelines
• OGM related stuff:
  • Need for sequences or alternative way(s) to create counters (related to use of CAS style methods on the Cache when in transaction)
  • Fine-grained-locking AtomicMap
  • Proper MVCC - not limited to REPL

Wed (29th)

• 9:30 Present existing locking model (Mircea)
• 10:30 Present the 3 locking improvements as described here and detailed (Mircea)
  • ISPN-1131 - Acquire (write) locks at prepare time even on local node
  • ISPN-1132 - Reorder lock acquisition to avoid deadlocks
  • ISPN-1137 - Locking optimization: only lock main data owner (dist only)
• 11:30 Cloud-TM and how this needs to be designed in (Manik)
• 12:30 Lunch
• 13:30 Eventual consistency and versioning entries

• All: review suggested design and brainstorm 

Followup
Following documents detail the final design:
- SingleNodeLockingModel
- OptimisticLockingInInfinispan
- LockReorderingForAvoidingDeadlocks