NetScaler appliances configured for global server load balancing (GSLB) provide for disaster recovery and ensure continuous availability of applications by protecting against points of failure in a wide area network (WAN). GSLB can balance the load across data centers by directing client requests to the closest or best performing data center, or to surviving data centers in case of an outage.
Following are some typical GSLB configurations:
- Active-active data center setup. Consists of multiple active data centers. Client requests are load balanced across active data centers.
- Active-standby data center setup. Consists of an active and a standby data center. When a failover occurs as a result of a disaster event, the standby data center becomes operational.
- Proximity setup. Directs client requests to the data center that is closest in geographical distance or network distance.
In a typical configuration, a local DNS server sends client requests to a GSLB virtual server, to which are bound GSLB services. A GSLB service identifies a load balancing or content switching virtual server, which can be at the local site or a remote site. If the GSLB virtual server selects a load balancing or content switching virtual server at a remote site, it sends the virtual server’s IP address to the DNS server, which sends it to the client. The client then resends the request to the new virtual server at the new IP.
The GSLB entities that you must configure are the GSLB sites, the GSLB services, the GSLB virtual servers, load balancing or content switching virtual servers, and authoritative DNS (ADNS) services. You must also configure MEP. You can also configure DNS views to expose different parts of your network to clients accessing the network from different locations.
Note: To take full advantage of the NetScaler GSLB features, you should use NetScaler appliances for load balancing or content switching at each data center, so that your GSLB configuration can use the proprietary Metric Exchange Protocol (MEP) to exchange site metrics.