(HA) provides redundancy through failover from a faulty primary Cisco WebEx
Meetings Server (CWMS) system to a backup CWMS HA system in the same physical
CWMS Multi-data center (MDC) deploys multiple data centers, and
then joins them into a single CWMS system. Failover is similar to a HA system,
except that MDC system data centers are peers both serving users and they are
not geographically limited. Indeed, deploying multiple data centers
geographically close to users improves network performance. A CWMS system
cannot support both HA and MDC.
The conditions for redundancy are:
The HA virtual machines must be co-located in the same data center as the primary virtual machines. All these virtual machines must be on the same VLAN or subnet. The speed and latency requirements for connectivity between the primary and HA components are the same as defined previously for the primary virtual machines. Splitting the primary and HA components of the system between data centers is not supported.
The MDC virtual machines are not required to be co-located in the same data center.
Connectivity between all the internal virtual machines must be fully redundant, so that the failure of a switch or network link does not sever the connectivity between the primary and HA or MDC components. To achieve this redundancy, each host server should have redundant connections to multiple Ethernet switches.
The primary and HA Internet Reverse Proxy (IRP) virtual machines must be on a common VLAN or subnet (typically not the same subnet as the internal virtual machines). Connectivity between the Internet Reverse Proxy virtual machines should also be redundant, in the same manner as the internal virtual machines.
After joining data centers in an MDC system, IRP can be enabled or disabled on the data centers. The IRP configuration for all data center in the CWMS MDC system must match; there cannot be a mismatch.
The addition of an
or MDCsystem does not increase the total system capacity.
Whether you deploy an 800 user system with or without HA, the total system
capacity remains the same; the maximum number of simultaneous audio connections
The HA or MDCsystem comprises redundant virtual machines for each virtual machine type in your deployment. (For a description of each type of virtual machine, see Virtual Machines In Your System.) For example:
A 50 user system consists of an Admin virtual machine and optionally an Internet Reverse Proxy (IRP) virtual machine for public access. If you add a HA (MDC is not available) system, the combined 50 user system consists of two Admin virtual machines and two IRP virtual machines.
A primary 250 or 800 user system consists of an Admin virtual machine, a Media virtual machine, and optionally an IRP virtual machine. If you add a HA or MDCsystem, the combined 250 or 800 user system comprises two Admin virtual machines, two Media virtual machines, and two IRP virtual machines.
A primary 2000 user system consists of an Admin virtual machine, three Media virtual machines, two Web virtual machines, and optionally an IRP virtual machine. If you add a HA or MDCsystem, the combined 2000 user system comprises two Admin virtual machines, four (three plus one redundant) Media virtual machines, three (two plus one redundant) Web virtual machines, and two IRP virtual machines.
In an HA or MDCsystem, the public VIP address and private VIP address are shared with the primary system. (The public VIP address and the private VIP address are different and are not shared.) When one virtual machine is down, the other virtual machine uses the same VIP address. Because of this behavior, a virtual machine failure is transparent to users as meetings continue without placing unusual demands on the DNS infrastructure. However, a shared VIP address can only be implemented on a single network segment or VLAN; splitting a VLAN across two data centers creates multiple problems.
connectivity between the primary and HA internal virtual machines to be within
the same data center, greatly reducing the problem of distinguishing between a
virtual machine failure and a network failure. Allowing a split network can
result in split meeting connections and conflicting database updates. It is
more practical to construct a true HA network segment within a single data
center than between multiple data centers.
In an MDC system, the data is replicated across data centers (except for the License Manager). Therefore if a data center goes down or network connectivity is lost, the surviving data center continues to serve users independent of geographic location.
The best way to build a fault tolerant system is when most system components operate as “all active.” However, certain key components, notably the database service, are “active/standby." (Web servers and media components in the HA system are dependent on the primary system components.) Any latency or interruption on the connections results in delays for users, particularly when joining meetings. Latency between media service components increases audio and video latency for some users during meetings. For Cisco WebEx Meetings Server, 4 ms of network latency is acceptable between the internal virtual machines.