Cisco Support Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Community Member

high availability UCCE

Hello I need to know what is the best practice to do High Availability with UCCE in scenario LAN-to LAN, what is the requirements for that?

What is the bandwidth  for the private and public?

anyone can help me?

Thanks.

1 ACCEPTED SOLUTION

Accepted Solutions
Green

Re: high availability UCCE

santiago_figueroa@mrd.net.mx

Hello Geoff the reason is for that Site2 is Headquarters and Site1 is branchoffice.

That's not a very convincing argument. Let's make sure I have this correct:

1. At site 1 (the branch office) there are 2 gateways. At site 2 (headquarters) there are ZERO gateways.

2. At site 1 (the branch office) there are N agents. At site 2 (headquarters) there are ZERO agents.

3. At site 1 (the branch office) there is a DC/DNS for the ICM Domain. At site 2 (headquarters) there is no DC/DNS

4. At site 1 (the branch office) there is an AW/HDS. At site 2 (headquarters) there is no AW/HDS.

5. At site 1 (the branch office) there are M non-agent phones. At site 2 (headquarters) there are P non-agent phones.

6. The publisher will be at site 2 (headquarters). If the WAN is down, no changes can be made to phones at the contact center.

As it is, splitting the sides makes little sense, so there must be something about the future layout that you haven't told me.

Let me know N, M and P.

Let me know the BHCA.

Regards,

Geoff

15 REPLIES
Green

Re: high availability UCCE

The SRND spells out the details and the requirements. This is quite a big topic, so you would need to be more specific.

Bandwidth between what and what - RoggerA and RoggerB? What part concerns you?

Regards,

Geoff

Community Member

Re: high availability UCCE

Hello Geoff,  I have to move the one PROGGER-B and CVP2 at SITE 2,in this moment a I have all in SITE1 ,

I attachement the my scenario that I want.

Thanks.

Green

Re: high availability UCCE

You need to go through the various paths and estimate the bandwidth.

Private

Rogger A to Rogger B

PGA to PGB

Public - Normal Operation

  • Under normal operation, agent phones will be bound to the CUCM at site 1 (so local)
  • JTAPI on ProggerA talking to the CTI Manager on CUCM at site 1 (so local)
  • 2121 gateways communicating with CVP2 - a number of different message streams (VXML, HTTP). Standard branch office so no problem.
  • The CVP PIM on site 1 will often be against the CVP at site 2 (or vice versa) so the PIM is talking to CVP across  the WAN. Sensitive to latency.

Public - Abnormal Operation

    • Under certain situations, the JTAPI at site 1 will be talking to the CTI Manager at site 2. This is "clustering across the WAN" and there are very strict requirements.
    • HDS replication from ProggerB to the AW/HDS at site 1 if RoggerA is down
    • Agent phones may be bound to CUCM at site 2 if subscriber at site1 fails

    12MB is way above what is required, but you need to separate the private traffic out. It cannot travel over the same wide area network.

    You will have to consider QoS - when the PG is talking to a Rogger at the other side, this is very sensitive. You will have latency requirements.

    How far apart are the sides? What is the reason for separating the servers?

    Regards,

    Geoff

    Community Member

    Re: high availability UCCE

    Thanks Geoff , Now I have a Questions the Private can use to link WAN, and Public to use the LAN to LAN?

    can I used for Private E1 link?

    Thanks.

    Green

    Re: high availability UCCE

    Sorry, I know English is not your first language, but I can't understand your question.

    For a deployment your size (you haven't provided any numbers so that's a guess), a single E1, independent of the 12mbs WAN link, would be sufficient for the private link. Get this from another vendor if possible, or an independent E1 from the same vendor, and run it through different switches inside your premises to the demark. You are trying to ensure that the Private does not go down at the same time as the WAN.

    "the Private can use to link WAN, and Public to use the LAN to LAN?"

    Sorry. It does not compute.

    Regards,

    Geoff

    Community Member

    Re: high availability UCCE

    Thanks Geoff I have attached the my scenario with that redundancy , can you see and tell me if it seem correct?

    Thanks.

    Green

    Re: high availability UCCE

    There's not much I can say, since you have not provided any numbers, but it's clear that you have a dedicated private link, which is what the A2Q team would be looking for when you submit your design.

    I always find that these diagrams look better if you put the private link on the inside. See attached.

    Please explain why you want to split the site up into two sides. What is the purpose? If site 1 is down, the CC can continue but there are no agents and no calls coming in, so what is the reason.

    Regards,

    Geoff

    Community Member

    Re: high availability UCCE

    Hello Geoff the reason is for that Site2 is Headquarters and Site1 is branchoffice.

    This are the numbers.

    inbound


    AVg Talk time sec is 210,Warpup time sec 60, percentage Trafic % 100, trans% 100%, conf% 10,Internal Transfer 10


    Outbound

    AvgTalk Time 180, Answered transfer to agent 50%

    Thanks.

    Green

    Re: high availability UCCE

    santiago_figueroa@mrd.net.mx

    Hello Geoff the reason is for that Site2 is Headquarters and Site1 is branchoffice.

    That's not a very convincing argument. Let's make sure I have this correct:

    1. At site 1 (the branch office) there are 2 gateways. At site 2 (headquarters) there are ZERO gateways.

    2. At site 1 (the branch office) there are N agents. At site 2 (headquarters) there are ZERO agents.

    3. At site 1 (the branch office) there is a DC/DNS for the ICM Domain. At site 2 (headquarters) there is no DC/DNS

    4. At site 1 (the branch office) there is an AW/HDS. At site 2 (headquarters) there is no AW/HDS.

    5. At site 1 (the branch office) there are M non-agent phones. At site 2 (headquarters) there are P non-agent phones.

    6. The publisher will be at site 2 (headquarters). If the WAN is down, no changes can be made to phones at the contact center.

    As it is, splitting the sides makes little sense, so there must be something about the future layout that you haven't told me.

    Let me know N, M and P.

    Let me know the BHCA.

    Regards,

    Geoff

    Community Member

    Re: high availability UCCE

    Hello Geoff, I forget to comment you that if agents will exist in the site2, now you can help me with a doubt  in ICM Setup do I have to change the configuration of PG in both sides to say that the callrouter is remote WAN? is correct that? if this correct that other thing ICM setup has  to change?

    Best Regards.

    Community Member

    Re: high availability UCCE

    Hello Geoff can you help me? According to my design I have a question in part of PG configuration,  I have a  the Private link(2Mbps) through WAN and Public link 10 Mbps .

    my doubt is that you value I should put here ?

    that I choose LAN or WAN according to my design?

    which the recommendation is?

    Side A

    Call Router is local ?

    Usable Bandwidth Kbps

    Call Router is Remote ?

    Usable Bandwidth Kbps

    Side B

    Call Router is local ?

    Usable Bandwidth Kbps

    Call Router is Remote ?

    Usable Bandwidth Kbps

    Thanks.

    Green

    Re: high availability UCCE

    There will be different configurations on side A and side B.

    When running setup on PGA - the router connection on side A is local, side B is remote

    When running setup on PGB - the router connection on side A is remote, side B is local

    Use the default values.

    I am still waiting for you to provide numbers I requested.

    Regards,

    Geoff

    Community Member

    Re: high availability UCCE

    Hello Geoff, I have a doubt over Split brain in scenario where the same phisical cable are used to private and public netrork this in scenario of LAB, but I need to know what is time of tolerance or TTL to said the event Split brain just happened?

    In other works if phisycal cable is down what is the time to said split brain just happeden?

    Thanks.

    Green

    Re: high availability UCCE

    As soon as the dual failure occurs, each side is each thinking it is the active side. They are writing to their Logger independently.

    Things function fine, until the problem is solved and the duplex nature is restored and the Loggers try to sync. This can corrupt the database.

    I can't say I have ever deliberately tried to do this to see what happens and I guess I am guilty of taking Cisco at their word - but it seems reasonable.

    Regards,

    Geoff

    Re: high availability UCCE

    The two sides of the Central Controller should never be active at the same time while they are not talking to eachother. This is where the concept of 'majority of PGs' comes in. I.e.each PG pair will end up registering to one end of the Central Controller, one of them will have more than 50% of the PGs registered and go In Service, the other one will be Isolated Disabled.

    This concept used to be 'perfect' in a world where both sides of a PG pair used to be at the same physical site as they had to sit physically next to the ACD that they controlled anway. The concept is not that great anymore in these days where you can split your CUCM cluster across a WAN link.

    You could have a scenario with two sites where :

    Site 1 : RouterA, PG1A, PG2A

    Site 2 : RouterB, PG1B, PG2B

    The issue with that is that if you have a total outage between both sites, PG1A and PG2A could end up talking to RouterA separately and PG1B and PG2B talk to RouterB separately. Both Routers would have 100% of PGs connected and go In Service.

    This will work jolly well for the duration of the outage. But it causes problems with replication of historical data. Going into the details would take us a bit far, but after your network / ICM recovers from a fault like the above, you're likely to have issues with replicating historical data to the HDSs attached to one of your Loggers (typically LoggerB). And you'd probably need manual intervention to resolve the situation.

    In short, if you will be splitting PGs across the WAN it's more crucial than ever that you have fully segragated public and private links between your Central Controllers.

    Cheers,

    Kris

    1228
    Views
    5
    Helpful
    15
    Replies
    CreatePlease to create content