Deployment Guide for IM and Presence Service on Cisco Unified Communications Manager, Release 9.0(1)
IM and Presence multinode deployment planning
Downloads: This chapterpdf (PDF - 1.63MB) The complete bookPDF (PDF - 6.21MB) | The complete bookePub (ePub - 1.26MB) | Feedback

IM and Presence multinode deployment planning

Contents

IM and Presence multinode deployment planning

Multinode Scalability Feature

Multinode Scalability Requirements

IM and Presence Service supports multinode scalability:

  • Six nodes per cluster
  • 45,000 users per cluster with a maximum of 15,000 users per node in a full Unified Communication (UC) mode deployment
  • 30,000 users per cluster in a deployment with High Availability turned off
  • Administrable customer-defined limit on the maximum contacts per user (default unlimited)
  • The IM and Presence Service continues to support intercluster deployments with the multinode feature.

Scalability depends on the number of clusters in your deployment. For detailed VM configuration requirements and OVA templates, see Virtualization for Unified CM IM and Presence at the following url: http:/​/​docwiki.cisco.com/​wiki/​Virtualization_​for_​Unified_​CM_​IM_​and_​Presence

Supported End Points

The multinode scalability feature supports the following end points:

  • Cisco Unified Communications Manager (desk phone)
  • Cisco Unified Personal Communicator Release 8.x (XMPP client)
  • Cisco Jabber
  • Third-Party XMPP clients
  • Cisco Unified Mobile Communicator
  • Microsoft Office Communicator (Microsoft soft client)
  • Lotus Sametime (Lotus soft client)
  • Third-Party Interface clients
  • Lync 2010 Client (Microsoft Office Communicator client)

Multinode Performance Recommendations

You can achieve optimum performance with the multinode feature when:

  • The resources on all IM and Presence Service nodes are equivalent in terms of memory, disk size, and age. Mixing hardware classes results in nodes that are under-powered, therefore resulting in poor performance.
  • You deploy hardware that complies with the hardware recommendations.
  • You configure a Balanced Mode deployment model. In this case, the total number of users is equally divided across all nodes in all subclusters. The IM and Presence Service defaults to Balanced Mode user assignment to achieve optimum performance.

IM-Only Deployment

The IM and Presence Service supports an IM-only deployment. This type of deployment supports up to 25,000 users per node and up to 75,000 users in an IM and Presence Service cluster.

High Availability deployments

The IM and Presence Service supports High Availability deployments.

Cisco recommends that you configure your IM and Presence deployments as High Availability deployments. Although mixed mode deployments are permitted, for example High Availability subclusters and non High Availability subclusters in a single deployment, we do not recommend this configuration.

You must manually turn on High Availability in a subcluster. You can achieve a High Availability deployment by configuring the Balanced Mode (Redundant High Availability) or the Active/Standby Redundant High Availability deployment models, and turning on High Availability in your deployment.

Subclusters

The multinode feature introduces the concept of a subcluster. A subcluster is a single IM and Presence node, or a pair of IM and Presence nodes, where each node has an independent database and set of users operating with a shared availability database that is able to support common users.

In a single-node deployment within a subcluster, there is no High Availability failover protection for users assigned to the node. In a dual-node deployment within a subcluster, if you turn on High Availability in the subcluster, users have failover protection; each node acts as a backup for the other node allowing clients to fail over in case of outages of components or nodes. When you turn on High Availability in a subcluster, all users in the subcluster have redundancy and full failover capabilities.

User assignment

To allow users to receive the availability and Instant Messaging (IM) services on IM and Presence, you must assign users to nodes, and subclusters, in your IM and Presence deployment. You can manually or automatically assign users in a IM and Presence deployment. You manage user assignment using the User Assignment Mode parameter on the Sync Agent on IM and Presence.

Balanced mode (default) assigns users equally to each node in the subcluster and attempts to balance the total number of users equally across each node. Active-Standby mode assigns all users to the first node of the subcluster, leaving the secondary server as a backup. None mode results in no assignment of the users to the nodes in the cluster by the sync agent. The default mode is Balanced.

If you choose manual user assignment, you must manually assign your users to nodes, and subclusters, using the System Topology interface in the Cisco Unified Communications Manager IM and Presence Administration.

Cisco Unified Personal Communicator sign-in and redirect

The IM and Presence Service supports the ability to redirect a Cisco Unified Personal Communicator client application to the IM and Presence Service node to which the user is assigned (home node). The redirection feature is supported in intracluster and intercluster deployments. In both types of deployments, redirect occurs automatically when the client application signs in. After the user successfully signs in to the home node, Cisco Unified Personal Communicator caches the node name. As a result, redirection occurs only once, unless a user is reassigned.

See the following examples to gain a better understanding of the various redirect scenarios. In the examples, Cluster1 is assumed to be an IM and Presence Service Release 9.0 cluster and Cluster2 is a 7.x or 8.x cluster.

Figure 1. Intercluster and Intracluster Redirect Diagram



In the preceding figure, Cluster1 has three nodes, a publisher (C1Node1) and two subscribers (C1Node2 and C1Node3) and has an intercluster peer relationship with Cluster2, which contains a publisher (C2Node1) and subscriber (C2Node2). Several different redirect scenarios are possible:

  1. A Cisco Unified Personal Communicator user is assigned C1Node1 as a home node and attempts to sign in to C1Node2. C1Node2 automatically redirects the Cisco Unified Personal Communicator client to C1Node1. In this scenario, High Availability is disabled in Subcluster1. If High Availability is enabled in Subcluster1, C1Node2 will process the login request. No redirection occurs.
  2. A Cisco Unified Personal Communicator user is assigned C1Node3 as a home node and attempts to sign in to C1Node1 or C1Node2. Regardless of whether High Availability is enabled in Subcluster1, C1Node1 or C1Node2 redirects the Cisco Unified Personal Communicator client to C1Node3. High Availability rules do not apply here because C1Node3 is part of Subcluster2.
  3. A Cisco Unified Personal Communicator user is assigned C2Node1 or C2Node2 as a home node and attempts to sign in to C1Node1, C1Node2, or C1Node3. C1Node1, C1Node2, or C1Node3 automatically redirects the Cisco Unified Personal Communicator client to its home node.

Note


For more information about establishing intercluster peer relationships and syncing users, see IM and Presence Service Intercluster Deployment Configuration.


Related Topic

AXL/​SOAP Interface

Clustering Over WAN

The IM and Presence Service supports Clustering over WAN deployments.

Multinode Deployment Models

You need to consider how you are going to deploy the multinode feature in your network. You configure your desired multinode deployment model in system topology management GUI in Cisco Unified CM IM and Presence Administration. Select System > Cluster Topologyin Cisco Unified Communications Manager IM and Presence Administration to access system topology management GUI.

This module provides an overview of the deployment model options for the multinode feature, and provides examples of these deployments on system topology management GUI.

You only use system topology management GUI to configure your local IM and Presence cluster. See the intercluster peer module for information about configuring intercluster peer relationships with remote IM and Presence clusters.

Balanced User Assignment Redundant High Availability Deployment

You can achieve a balanced mode High Availability deployment by evenly balancing users across all nodes in the subcluster, but only using up to 35% of the CPU of each IM and Presence Service node.

The balanced mode High Availability deployment option in a redundant mode supports up to 45,000 users per cluster. For example, if you have six IM and Presence Service nodes in your deployment, and 15,000 users, you assign 2,500 users to each IM and Presence Service node.

When you use the balanced mode High Availability deployment option in a redundant mode, as compared to a non-redundant mode, only half the number of users are assigned to each node. However, if one node fails, the other node will handle the full load of the additional 50% of users in the subcluster, even at peak traffic. In order to support this failover protection, you must turn on High Availability in each of the subclusters in your deployment.

See the following figure for an example of this deployment model on system topology management GUI. In this example, there are 15,000 users in total, so 2,500 users are evenly balanced across the six nodes.

Figure 2. Balanced User Assignment Non-Redundant High Availability Deployment

Active/Standby User Assignment Redundant High Availability Deployment

For this deployment model, assign all your users to the primary IM and Presence Service node, and none to the backup node. When you turn on High Availability in the subcluster, the backup node can handle all traffic from the primary node if the primary node fails.

See the following figure for an example configuration for this deployment model on system topology management GUI. In this example, there are 15,000 users in total, so 5000 users are assigned to the first node of each subcluster.

Figure 3. Active/Standby User Assignment High Availability Deployment

Scalability Options for Deployment

IM and Presence Service clusters can support up to six nodes. If you originally installed less than six nodes, then you can install additional nodes at any time. If you want to scale your IM and Presence Service deployment to support more users, you must consider the multinode deployment model you have configured. The following table describes the scalability options for each multinode deployment model.

Table 1 Multinode Scalability Options

Deployment Mode

Scalability Option

Add a New Node to an Existing Subcluster

Add a New Node to a New Subcluster

Balanced Non-Redundant High Availability Deployment

If you add a new node to an existing subcluster, the new node can support the same number of users as the existing node; the subcluster can now support twice the number of users. It also provides balanced High Availability for the users on the existing node and the new node in that subcluster.

If you add a new node to a new subcluster, you can support more users in your deployment.

This does not provide balanced High Availability for the users in the subcluster. To provide balanced High Availability, you must add a second node to the subcluster.

Balanced Redundant High Availability Deployment

If you add a new node to an existing subcluster, the new node can support the same users as the existing node. For example, if the existing node supports 5000 users, the new node supports the same 5000 users. It also provides balanced redundant High Availability for the users on the existing node and the new node in that subcluster.

Note   

You may have to reassign your users within the subcluster, depending how many users were on the existing node.

If you add a new node to a new subcluster, you can support more users in your deployment.

This does not provide balanced High Availability for the users in the subcluster. To provide balanced High Availability, you must add a second node to the subcluster.

Active/Standby Redundant High Availability Deployment

If you add a new node to an existing subcluster, you provide High Availability for the users in the existing node in the subcluster. This provides a High Availability enhancement only; it does not increase the number of users you can support in your deployment.

If you add a new node in a new subcluster, you can support more users in your deployment.

This does not provide High Availability for the users in the subcluster. To provide High Availability, you must add a second node to the subcluster.

Related Tasks
Related References

Cluster-Wide DNS SRV

For DNS configuration, you can define a cluster-wide IM and Presence Service address.

The SIP Publish Trunk on Cisco Unified Communications Manager uses the cluster-wide IM and Presence Service address to load-balance SIP PUBLISH messages from Cisco Unified Communications Manager to all nodes in the cluster. Notably this configuration ensures that the initial SIP PUBLISH messages are load-balanced across all nodes in the cluster. This configuration also provides a High Availability deployment as, in the event of a node failing, Cisco Unified Communications Manager will route the SIP PUBLISH messages to the remaining nodes.

The cluster-wide DNS configuration is not a required configuration. Cisco recommends this configuration as a method to load-balance the initial SIP PUBLISH messages across all nodes in the cluster. IM and Presence Service sends subsequent SIP PUBLISH messages for each device to the node where the user is homed on IM and Presence Service.

Multinode Hardware Recommendations

When configuring the multinode feature, consider the following:

  • Cisco recommends turning on High Availability in your deployment.
  • Minimize your hardware, for example, instead of using six MCS 7825 servers that support a total of six thousand users, choose two MCS 7835 servers that can support a total of five thousand users.
  • Use the same generation of server hardware.
  • Use similar hardware for all nodes in your deployment. If you must mix generations of similar hardware, put the same generations of older hardware together in a subcluster and put fewer users on this subcluster than on the more powerful subclusters. Note that we do not recommend this deployment practice.

Warning


For multinode deployments using mixed hardware (for example, UCS, MCS, or VMware), it is highly recommended that the IM and Presence Service subscriber and database publisher nodes in the same subcluster have similar database size. If a significant difference in database size exists between the two nodes, you will receive an error during installation of the subscriber node.


  • Use the following disk drives for the multinode feature:
    • MCS 7816: minimum one 160GB drive (a 250GB drive can also be used)
    • MCS 7825: minimum two 160GB drives (two 250GB drives can also be used, upgrade required from smaller 80GB drives)
    • MCS 7835: minimum two 146GB drives (upgrade required from smaller 72GB drives)
    • MCS 7845: minimum four 72GB drives (upgrade recommended to four 146GB drives)

      Note


      The MCS 7845 with four 72GB drives can run the scalability feature, but this hardware with four 146GB drives is preferred.


  • If you have older-generation hardware, follow the disk drive upgrade recommendations above. You must meet the minimal disk capacity on each server in the cluster in order to achieve scale.

Note


Upgrading drives will allow you to use older hardware in a multinode cluster. However, Cisco recommends that you use the latest hardware available for the multinode feature because this hardware has more powerful CPU, more memory and faster input/output processing.


For a list of the supported hardware for the multinode feature, and hardware user assignment guidelines for the multinode feature, see the IM and Presence Service compatibility matrices at this URL:

http:/​/​www.cisco.com/​en/​US/​products/​sw/​voicesw/​ps556/​products_​device_​support_​tables_​list.html

Clustering Over WAN for Intracluster and Intercluster Deployments

IM and Presence Service supports Clustering over WAN for intracluster and intercluster deployments.

WAN Bandwidth Requirements

At a minimum, you must dedicate 5 Mbps of bandwidth for each IM and Presence Service subcluster, with no more than an 80 millisecond round-trip latency. These bandwidth recommendations apply to both intracluster and intercluster WAN deployments. Any bandwidth less than this recommendation can adversely impact performance.


Note


Each IM and Presence Service subcluster that you add to your Clustering over WAN deployment requires an additional (dedicated) 5 Mbps of bandwidth.


Intracluster Deployments Over WAN

IM and Presence Service supports intracluster deployments over WAN, using the bandwidth recommendations provided in this module. IM and Presence Service supports a single subcluster geographically split over WAN, where one node in the subcluster is in one geographic site and the second node in the subcluster is in another geographic location.

This model can provide geographical redundancy and remote failover, for example failover to a backup IM and Presence Service node on a remote site. With this model, the IM and Presence Service node does not need to be co-located with the Cisco Unified Communications Manager database publisher node. The Cisco Unified Personal Communicator client can be either local or remote to the IM and Presence Service node.

This model also supports High Availability for the Cisco Unified Personal Communicator Release clients, where the clients fail over to the remote peer IM and Presence Service node if the services or hardware fails on the home IM and Presence Service node. When the failed node comes online again, the clients automatically reconnect to the home IM and Presence Service node.

When you deploy IM and Presence Service over WAN with remote failover, note the following restrictions:

  • This model only supports High Availability at the system level. Certain IM and Presence Service components may still have a single point of failure. These components are the Cisco Sync Agent, Cisco Intercluster Sync Agent, and Cisco Unified CM IM and Presence Administration interface.
  • This model supports High Availability for the Cisco Unified Personal Communicator Releases 8.5 and 8.6.

IM and Presence Service also supports multiple subclusters in a Clustering over WAN deployment. For information about scale for a Clustering over WAN deployment, see the IM and Presence Service SRND.

Related Topic

IM and Presence Solution Reference Network Design (SRND):

http:/​/​www.cisco.com/​en/​US/​docs/​voice_ip_comm/​cucm/​srnd/​7x/​uc7_​0.html

Local Failover

You can also deploy IM and Presence Service over WAN where one subcluster is located in one geographic site, and a second subcluster is located in another geographic site. The subcluster can contain a single node, or a dual node for High Availability between the local nodes. This model provides no failover between geographic sites.

Subcluster failure detection

The IM and Presence Service supports a failure detection mechanism for a subcluster. Each node in the subcluster monitors the status, or heartbeat, of the peer node. You can configure the heartbeat connection and heartbeat intervals on IM and Presence by selecting Cisco Unified CM IM and Presence Administration > System > Service Parameters > Cisco Config Agent (service). In the section General Cisco Config Agent Parameters (Clusterwide), configure the following parameters:

  • Heart Beat Interval: This parameter specifies how often in seconds the Cisco Config Agent sends a heartbeat message to the peer Cisco Config Agent in the same subcluster. The heartbeat is used to determine network availability. The default value is 60 seconds.
  • Connect Timeout: This parameter specifies how long in seconds the Cisco Config Agent waits to receive a response from a connection request to the peer Cisco Config Agent. The default value is 30 seconds.

Note


We recommend that you configure these parameters with the default values.


Method Event Routing

When you deploy IM and Presence Service over WAN we recommend that you configure TCP method event routing on IM and Presence Service. Choose Cisco Unified CM IM and Presence Administration > Presence > Routing > Method/Event Routing to configure method event routes.

Multinode Configuration for Deployment Over WAN

When you configure the IM and Presence Service multinode feature for an intracluster deployment over WAN, configure the IM and Presence Service subcluster, nodes and user assignment as described in the multinode section, but note the following recommendations:

  • For optimum performance, Cisco recommends that you assign the majority of your users to the home IM and Presence Service node. This deployment model decreases the volume of messages sent to the remote IM and Presence Service node over WAN, however the failover time to the secondary node depends on the number of users failing over.
  • If you wish to configure a High Availability deployment model over WAN, you can configure a subcluster-wide DNS SRV address. In this case, IM and Presence Service sends the initial PUBLISH request message to the node specified by DNS SRV and the response message indicates the host node for the user. IM and Presence Service then sends all subsequent PUBLISH messages for that user to the host node. Before configuring this High Availability deployment model, you must consider if you have sufficient bandwidth for the potential volume of messages that may be sent over the WAN.

WAN Bandwidth Considerations

When you calculate the bandwidth requirements for your Clustering over WAN deployment, consider the following:

  • In your bandwidth considerations, you must include the normal bandwidth consumption of a Cisco Unified Communications Manager cluster. If you configure multiple nodes, Cisco Unified Communications Manager uses a round-robin mechanism to load balance SIP/SIMPLE messages, which consumes more bandwidth. To improve performance and decrease traffic, you could provision a single dedicated Cisco Unified Communications Manager node for all SIP/SIMPLE messages sent between the IM and Presence Service and Cisco Unified Communications Manager.
  • In your bandwidth considerations, we also recommend that you consider the number of contacts in the contact list for a Cisco Unified Personal Communicator user, and the size of user profiles on IM and Presence Service. See the IM and Presence Service SRND for recommendations regarding the size of a contact list when you deploy IM and Presence over WAN. Note also that the maximum contact list size on IM and Presence Service is 200, so you need to factor this in to your bandwidth considerations for systems with large numbers of users.

For additional information, see the IM and Presence Service Solution Reference Network Design (SRND):

http:/​/​www.cisco.com/​en/​US/​docs/​voice_ip_comm/​cucm/​srnd/​7x/​uc7_​0.html

External Database Recommendations

If you configure external database servers in your Clustering over WAN deployment, Cisco recommends that you co-locate the external database servers with the IM and Presence Service nodes that will use the external database servers.

For more information about external database servers and IM and Presence Service, see Database Setup Guide for IM and Presence Service on Cisco Unified Communications Manager.

Intercluster Deployments Over WAN

IM and Presence Service supports intercluster deployments over WAN, using the bandwidth recommendations provided in this module.