Axis 2 provides many desirable features, axis 2 migration in SRM should be high priority.

Pnfs Manager load in CMS:
  Most of the load comes from PnfsManager
  Pnfs Manager counting of the PnfsManager-pnfs back-end operations can be helpful. FNAL will implement it.
  After that we can compare behavior of this interface between 1.9.1 and 1.9.2.
  Tigran: Write into dCache changed, AL and RP is now stored in layer 2. 
  Possible solution, batch modifications to layer 2 and write it only once. 
  It appears that layer 2 modification on create did not exist at all before 1.9.2, so setting AL/RP would need to 
  be added to a later layer operation, setSize. This will require either addition of this info to setSize message coming from pool, which is modification of the pool/pnfsmanager interaction which will require a simultaneous upgrade of the pools. 
 Another solution is to extract the AL/RP from dir tags at the moment of the setSize execution, which still is an increase of the load on pnfs, but possibly not as bad as setting layer 2 twice, and allow for upgrade of pnfs manager without pool upgrade.
  Proposal, modify the setSize so that it does not require a pool upgrade, and add a new or use a different message for setting size and AL/RP simultaneously will allow the new pools to reduce the impact of AL/RP storage to the maximum possible extend. 
  The solution to a problem of added load generated by AL/RP storage will be coming within days.

  Migrate to BerkleyDB pnfs.
  Paul: Run PnfsDump on pnfs before migration, it helps find multiple idiosyncrasies that had developed in the pnfs over the time. 

JMS in dCache.
  Swedish T2 runs a production system with JMS used for cells tunnels. Gerd gave a detailed explanation of the current support for JMS messaging in dCache and his ideas for further expansion of the functionality.