With this change we proactively fail map updates when serialization failures can occur and immediately notify the caller
Change-Id: I62a8a84731b9c2a6eeff7fa6f8336dc74234bf30
Stream processes by Lambda queries in collection objects has a known performance bottleneck. For each create and update link data operations, this kind of queries are frequently used and this causes inefficient topology discovery operation. For to solve that problem i analize thread dump during discovery process and i saw that especially getAllProviders method uses high cpu. than i change the provider search operation with additional map data which holds linkkey and its providers. by this way provider search operation not only gets faster but also uses minimum resource. Before that change, i can discover only 4XX number of switches at one controller, but later number grows to 15XX switches
Change-Id: I65ed71b7f295917c818b2f9227d0fc070aa4a29b
- Do not include known peers in heartbeat message and instead each node relies on cluster metadata to add/remove nodes from inventory
- Ignore heartbeats from nodes that are removed
Change-Id: Ia20ce84ad88aa6c723a21af1fe7f6899821181ed
- Use execute instead of submit so that uncaught Exception will
be dealt and logged by ExecutorService's handler.
- Use component's own logger
Change-Id: I761264aea00748980929b5048e111756776dd2f6
In this submission:
* Setup/teardown multicast route according to SinkAdded/SinkRemoved event
- ingressVlan and egressVlan is configurable through network config
* Change behavior of OFDPA VLAN assignment
- Always use the VLAN in metadata if present
* Bugfix of writing immutable object
NOT in this submission (coming soon):
* Error handling (e.g. link/device failure recovery)
Change-Id: I9be11af04eb2d6456b865c7e59e96cc02370f846
- Additional log on error
- Allow count=0 using CountDownCompleter
- test case to detect the issue (@Ignored by default right now)
- other bug fixes found along the way
Based on patch by Madan@China
Change-Id: I7d6cb8c214052859900ef7ee0337a7e1c8a9d295
- Create update list using Stream, so that
hasPendingUpdates() can short cut on first update found.
- Might improve performance when there is large number of updates in a Tx.
Change-Id: I20820b7212c642315a620d715b5c750e35d31dd0
- Using osgi_jar for all modules except apps
- Adding target for installation of core modules to .m2 repo
Change-Id: I65d9e3a9567918e275191724842d7d9b1f485128
- Fast path of transactions updating a single key.
- Bug fix in StoragePartitionClient where we were always creating a CachingMap
Change-Id: Ide117fba34fd12a9ff4aabd5fb7a21952bae672b
- Dropping MutexExecutionService as there are now better alternatives
- Dropping New from class names that were added during transition phase
Change-Id: If0cdd3321081c3f8fda81441ef2c84549b616edd
