locked
OCS R2 Enterprise Voice failover mechanism? RRS feed

  • Question

  • Hi everyone,

    I am looking among the documentation-sets about a detailed explanation, how OCS R2 solves the problem of high availability regarding multiple mediation servers and multiples PSTN gateways.

    I know the basic idea: if multiple gateways are configured for a single route, they are used in a round robin loadbalanced way. What bothers me, is the way that OCS recognizes that one of the breakout points are wrong.

    Scenario1): OCS FE -> mediation server connection is down / service down
    Scenario2): OCS FE -> mediation server connection and service is fine, but connection between med and GW is down
    Scenario3): OCS FE -> mediation server connection and service is fine, med -> GW connection is fine, but PSTN trunks are down

    Is OCS R2 in all the above cases able to detect and problem, and failover to additional routes, or will just drop the call with error?
    Thursday, August 27, 2009 3:24 PM

All replies

  • Calls in progress that are not conferences will remain in progress with the loss of a med server or FE server because the RTP stream is point to point.  The "High Availability" with multiple med servers and multiple gateways only addresses new call routing and that is addressed by utilizing multiple MED servers as gateways on the Routes tab.

    I guess if I understood what you are trying to accomplish I could prob help you better... 


    Mark King | C/D/H | MCTS:OCS | MCSE: Messaging | MCITP:Enterprise Administrator | CCNA
    Thursday, August 27, 2009 4:20 PM
  • Hi Mark,

    all my questions relates strictly to OC -> PSTN calls.

    My experince shows, that if med server crashes or service stops, all active calls flowing through that particular server, are disconnected. I expected exactly this behaviour.

    My goal is to have 0.0 % rejected-call ratio, so no call gets rejected because 1 of the 10 redundant med server or the connected pstn gateways is inaccessible, and the built-in failover mechanism is not able to choose any other perfectly working med server.

    I want to discover all the possible use-cases, where I have to expect that builtin failover may think that the destination number cannot be called, and instead of choosing a different available med server, it rejects the call without making all the effort to connect the call via a different routing path.
    Thursday, August 27, 2009 5:27 PM
  • Im sorry, you are correct.  You can withstand the loss of a FE server but not a MED server (while I don't understand why you lose the call just because the SIP signaling is gone)  but I digress... 

    Now I understand what you are trying to accomplish in your scenarios above.  Here is what I believe is the relevant information from TechNet. 

    "More than one Mediation server (and therefore media gateway) can be listed in a single route. Office Communications Server 2007 routes call to the Mediation servers in a round-robin fashion as a way to balance the traffic across the gateways. If a gateway fails or is taken out of service for maintenance, the corresponding Mediation server will send back a response to Office Communications Server 2007, which will immediately attempt to route the call to another Mediation server (and therefore a different media gateway) in the route's list.
    After 10 attempts to route calls to a failed media gateway, Office Communications Server 2007 will subsequently scale down traffic to that media gateway until it becomes responsive again. If the media gateway continues to fail to respond after an additional 10 attempts, Office Communications Server 2007 will stop routing calls to that media gateway until it becomes responsive."

    Mark King | C/D/H | MCTS:OCS | MCSE: Messaging | MCITP:Enterprise Administrator | CCNA
    Thursday, August 27, 2009 6:02 PM
  • What mark said is correct, also on that note, some of the gateways have built in load balancing that is not very well known of.

    On the dialogics for example you can configure multiple voip hosts on a single gateway and configure a monitor to verify it is active, this would allow for redundancy on outbound and inbound calls, depending on the capabilities of your gateways.


    Randy Wintle | MCTS: UC Voice Specialization | WinXnet Inc
    Thursday, August 27, 2009 8:32 PM
  • Randy,

    I have a situation, that built-in OCS failover cannot handle, that iss why the my investigation started:

    - multiple mediation servers accross many countries
    - routing configured to achieve least-cost routing in countries, where we have local PBX
    - last-resort routing, that tries all the installed med. servers as the last step, if all the previous routing attempts failed.

    We have a telephony provider issue in one of the offices, so we cannot dial out from that office. Calls going via that gw to outside of the office are rejected by OCS. Even if we have multiple other breakout points to that particular destination number (none of them are as optimal as the local breakout point) but this single failure breaks the mechanism OCS to try other gateways configured for this destination. I would need a list of response codes, that triggers the failover, and other codes, that suggest OCS not to try other paths.

    I know the loadbalacing features of dialogic, but if the solution is to try a different med. server (probably in a different country). Loadbalancing locally in the gw does not help in this situation.

    I hope that makes my problem clear to you.
    Friday, August 28, 2009 9:04 AM