High Availability Clustering and Data Recovery Options
Different options are available to provide for high availability during routine maintenance or unexpected outages and for service recovery in the event of a disaster. These options are selected during server setup on the New Server Configuration checklist. The Systems Support group will work with you to determine the best options based on your availability needs and goals.
Using the 3-Tier Server Farm network and other high availability software, following are options that can help achieve your availability goals.
Servers in Different Buildings
Since the Server Farm network spans two buildings, servers can be installed on the same network in different buildings to allow service IPs, applications, storage, etc. to float from one node to the next. This option uses the high-availability cluster software, SUN clustering for the Solaris operating system and VMware for the Windows operating system. In a disaster in which Rhodes Hall is lost, the Server Farm networks in CCC will be re-routed by hand through another router. IPs, storage, applications will remain the same.
Multiple Connections per Server
Sun/Solaris and VMware servers can be connected to the same network through two different NICs (network interface cards) and two different tiered switches to have redundant paths to the network. During routine switch maintenance and upgrades, the Solaris OS uses IPMP (IP network multipathing) to switch connections between tier 1 and tier 2 to provide high availability. IPMP is usually used with the Sun cluster software product, but can be used on stand-alone Sun servers for the same purpose. VMware uses active/active load balancing between the to NIC's for all the vlans it supports. Each virtual server automatically has access to this redundant network path.
Load Balancer and Several Back-end Servers
A load balancer can be used to direct traffic to more than one server behind the load balancer. The primary load balancer is on a tier 1 network. The backup load balancer is on a tier 2 network. If some of the back-end servers are on a tier 1 network and some are on a tier 2 network, the service remains available during switch maintenance. To provide for disaster recovery, place back-end servers in both data centers as well.
Fault Tolerant Applications
Fault tolerant applications are designed to continue working despite network outages. Server names for the service are configured into the application on the user's workstation. The application attempts to connect to each server in turn until successful. If your applications are designed for fault tolerance, to provide high availability to your users, distribute the servers on both tier 1 and tier 2 network switches. To provide for disaster recovery, place servers in both data centers as well.
Critical Services
For critical services like DNS, servers are spread between both buildings, using different networks, routers, and even off-campus locations.
Higher Availability Options
If none of the above options is the right choice for your service requirements, send e-mail to systems-support@cornell.edu. The Systems Support group can discuss other options with you.