Please find the features planned for third milestone below:

*Load Balancer*
- Introduced load balance statistics publishing interface.
- A reference implementation of the above interface was done for HAProxy
extension.
- Added support for dedicated load balancing for services. This applies to
both Stratos load balancer and HAProxy extension.
- Added support for dedicated load balancing for subscriptions. This also
applies to both Stratos load balancer and HAProxy extension.
- Fixed failover logic: Support for retrying requests on all members of a
given cluster if the selected member did not respond for a request.

*Auto-Scaler*
- Introduced partition group model. This functionality could be used for
applying autoscaling algorithms on separate partition sets.
- Divided policies into two main types: Autoscaling & Deployment. Here the
autoscaling policies define how autoscaling logic is executed and
deployment policies define in which partition the next member is spawned.

*Stratos Manager*
- Cleaned up and enhanced artifact synchronization, so that both repository
initial checkout (git clone) and repository updates (git pull) are done
through the java based deployment synchronization client.  git.sh script is
not used anymore.
- Git repo credentials are encrypted/decrypted using a key, which is
generated per cartridge subscription.

*Cartridge Agent*
- Publish health statistics periodically to CEP (load average and memory
consumption).

*CEP Extension *
- Analyze health statistics sent by cartridge agent and publish inactive
member information (faulty members) to heath-stat topic in message broker.


Many Thanks
Imesh


On Mon, Nov 25, 2013 at 1:51 PM, Imesh Gunaratne <[email protected]> wrote:

> Hi All,
>
> We are now finalizing the third milestone of 4.0.0 release. The idea is to
> freeze the code-base today evening and start the integration tests.
>
> *[Important]*
> Please hold any other modifications which were not planned for M3 to be
> committed after this milestone is announced.
>
> Many Thanks
> Imesh
>
>
>
>
>

Reply via email to