Talk:Virtualization for Unified CCE
Router/Logger Example 1 Configuration seems to be incorrect, please see below:
1) A typical UCM subscriber will support at most 500 agents. A UCM cluster has a historical maximum of 2,000 agents. There should be four, rather than two UCM clusters in the UCS Design.
2) A CVP Call Server (in this case Call + VXML) will support 900 calls, where calls are defined as (self service + queue + connected to agents). The server count is then calls / 900, rounded up. With 8,000 agents and 3,600 CVP ports, calls is equal to 8,000 + 3,600 = 11,600. 11,600 / 900 = 12.88888; rounded up to 13. For a geographically split core with N+N redundancy, 13 CVP Call + VXML virtual servers are required at each core. Only 4 per core are reflected in the example.
3) It seems another VRU PG pair is warranted, given the logic of #2, above.
4) Is it a valid design to have SIP dialers on four Agent PG pairs? Would it be better to put the SIP dialer alone on a VM or more than one VM?
5) Is one CVP Report Server per every four CVP Call + VXML servers really enough?
Respectfully, Dan Eccher ShoreGroup, Inc.
Question to Server spec for UCCE
HI Tems, As we know that Unified IP-IVR is supported with Unified CCE on UCS B-Series solution and on UCS C-Series with the model (UCS-C210-VCD2) only as guide of UC on UCS
Our customer want to deploy to IP-IVR and ICM using the UCS-C250-M2 of UCS server
Is it possible to deploy to UCS-C250-M2 ?
UCCE for 12,000 agents
For "Router/Logger Example 2" (Router/Logger - up to 12,000 CTIOS Agents) We need a confirmation for ESXi Server A-4 and ESXi Server B-4. The example shows: 4 HDS on A4 and 4 HDS on B4. From my point of view it is not possible for several reasons: - 4 HDS on the same blade is not possible. - Total amount of Vcpu by blade is more than supported. -- Lotfi AMARI
-Thanks for the update (Now, ADS/HDS/DDS on four different ESXi) -- Lotfi
Addition of B230 Blades
There is no mention of the change in sizing due to the newer blades?
Are we looking to add this soon?
PG co-res w/ Unity Connection
Can we get DocWiki updated to reflect co-res support with other UC apps including Unity Connection under the following conditions:
The rule is you must use the 1000 user OVA (1vCPU) for Unity Connection. In addition, any server hosting a Unity Connection VM must reserve 1 vCPU for use by the ESXi scheduler (this has always been required for Unity Connection).
So you will effectively have:
1 cpu – agent PG 1 cpu – CUCM 1 + 1 cpu – Unity Connection
For a total of 4 vCPUs for your desired workload.
The following section would be updated to reflect CUC in addition to UCM/CUP/IPIVR.
Note: EXCEPTIONS to the above VM co-residency table:
On a C-Series server, the HDS cannot co-reside with a Router, Logger, or a PG. PG (in CCE solutions up to 1000 CTIOS agents or 500 CAD agents) VMs can be co-resident with UCM/CUP/IPIVR VMs on the same ESXi host/serv
Thanks, Jon Nelson
Comments on Operating Conditions in IOPS Numbers - ECCs
In looking through the IOPS information recently, I realized there were operating conditions that all look sound...except for ECCs!
Why only 5 scalars? With the vast majority of deployments using CVP, I think this is far too low of a number. I have a spreadsheet I created some time ago to track all the configured "standard" ECCs for a CVP deployment, and it breaks down like this:
33 total ECCs
31 scalars, total size of about 338 bytes - most of these are for CVP only, and while they may not all be used, not configuring most of them will actually result in configuration errors in the PIM, so even if they are not used, configuring them could be considered "best practice". They don't necessarily amount to a lot of bytes (the ones that just prevent config errors), but they should certainly be included in a standard ECC count.
2 arrays (ToExtVXML and FromExtVXML), with 5 and 4 elements, respectively, each element (for both) typically sized around 100 bytes, totaling 900 more bytes between them.
That doesn't include any of the multichannel or Outbound ECCs.
I would strongly suggest that a more realistic ECC load be used for IOPS and other performance/scalability benchmarking.
- Bill Webb, Presidio Networked Solutions