Cisco Cisco Evolved Programmable Network Manager 2.0 Guide De Montage
27
6
High Availability
•
•
•
•
•
High Availability Deployment Models
Cisco EPN Manager HA feature supports the following three HA deployment models.
Consider the following factors when deciding whether to use the Local, Campus, or Remote HA deployment model:
•
Exposure to disaster—Remote HA deployments are least likely to be affected by natural disaster, resulting in a less complex
and costly business continuity model. While the same is true for Campus HA deployments, this can complicate business
continuity planning and increase disaster recovery. Local HA deployments are most vulnerable to disaster because of server
co-location, which can result in increased insurance costs.
and costly business continuity model. While the same is true for Campus HA deployments, this can complicate business
continuity planning and increase disaster recovery. Local HA deployments are most vulnerable to disaster because of server
co-location, which can result in increased insurance costs.
•
Whether you can use a virtual IP address—Only Local HA deployments can use virtual IP addresses. A virtual IP address is
a single IP address that will always point to the active server, even after a failover and fallback. It also allows both the
primary and secondary servers to share a common management IP address.
a single IP address that will always point to the active server, even after a failover and fallback. It also allows both the
primary and secondary servers to share a common management IP address.
•
Bandwidth/latency—Bandwidth/latency is best for Local HA deployments because the primary and secondary servers are
connected by short network links that have high bandwidth and latency. Campus HA deployments may have lower
bandwidth/latency than Local HA deployments. Remote HA deployments have the least bandwidth.
connected by short network links that have high bandwidth and latency. Campus HA deployments may have lower
bandwidth/latency than Local HA deployments. Remote HA deployments have the least bandwidth.
•
Administration—HA administration simplest for Local HA deployments, with increasing complexity for Campus and
Remote HA deployments. Remote HA deployments will require administrative remedying.
Remote HA deployments. Remote HA deployments will require administrative remedying.
•
Configuration of device event forwarding—Configuring event forwarding can be simplest with Local HA deployments
because you can use a virtual IP address, and then configure your devices to forward events to that single virtual IP address.
Without a virtual IP address, you must configure your devices to forward events to both the primary and secondary servers.
because you can use a virtual IP address, and then configure your devices to forward events to that single virtual IP address.
Without a virtual IP address, you must configure your devices to forward events to both the primary and secondary servers.
High Availability Installation Considerations
The Cisco EPN Manager HA framework is subject to the following limiting factors (this applies to all HA deployment models):
•
The framework requires at least 1 Gb per second of network bandwidth is required to handle HA operations. These
operations include (but are not restricted to) HA registration, database and file synchronization, and triggering failback.
Because Cisco EPN Manager uses a single physical port for all its networking needs, there can be occurrences of insufficient
bandwidth which in turn will affect HA performance.
operations include (but are not restricted to) HA registration, database and file synchronization, and triggering failback.
Because Cisco EPN Manager uses a single physical port for all its networking needs, there can be occurrences of insufficient
bandwidth which in turn will affect HA performance.
•
The framework requires net latency across network links between the primary and secondary servers. Irrespective of the
physical proximity of these two servers, high latency on these links can affect how Cisco EPN Manager maintains sessions
between the primary and secondary servers.
physical proximity of these two servers, high latency on these links can affect how Cisco EPN Manager maintains sessions
between the primary and secondary servers.
HA performance is always sensitive to the net throughput delivered by the network that connects the primary and secondary
servers. This restriction applies (to at least some degree) to all of the deployment models. For example, in a geographically
dispersed deployment, a Remote HA deployment is more likely to have problems due to low bandwidth and high latency. But
if Local and Campus HA deployments are not properly configured, they are highly susceptible to problems with latency that
result from bandwidth limitations on high-usage networks.
servers. This restriction applies (to at least some degree) to all of the deployment models. For example, in a geographically
dispersed deployment, a Remote HA deployment is more likely to have problems due to low bandwidth and high latency. But
if Local and Campus HA deployments are not properly configured, they are highly susceptible to problems with latency that
result from bandwidth limitations on high-usage networks.
HA Deployment Model
In this deployment, both servers are located:
Example:
Local
On same subnet (Layer 2 proximity)
Servers located in same data center
Campus
Different subnets connected via LAN
Servers located in same campus, city, state, or province
Remote
Different subnets connected via WAN
Servers are geographically dispersed