Now, besides the regular high availability feature that requires the use of two nodes, NetScaler also has some other high availability features such as clustering.
We will not dive deep into clustering, but we will be going through a basic overview. If you wish to read more about clustering, you should read the clustering guide from Citrix, which can be found at http://bit.ly/1NLPPIA.
Regular high availability operates with a two-node instance, where one is active and the other is passive. With clustering, we can scale from 2 up to 32 nodes, which are operating in an Active/Active state. This allows for a large amount of traffic throughput. In order to use this, we require an additional license. Also, there are some other requirements, as follows:
- To be on the same subnet
- To be of the same hardware type (for example, physical appliances)
- To be of the same platform type
- To have the same licenses (Standard, Enterprise, and Platinum)
- To be of the same software version and build
- To have their own feature license to be able to use clustering
Here, we can configure all NetScaler nodes from a single IP called the cluster IP, which in essence replaces the NSIP for management, so they act as one logical entity. The cluster IP is owned by the configuration coordinator, which is in essence the cluster coordinator role, which floats between the different nodes within a cluster.
Every VIP address that we create will automatically be available on all the nodes (called striped IP address) in a cluster; every Subnet IP (SNIP) address can also be distributed to all the nodes (striped IP address) or be available from just one node (spotted IP address). Citrix recommends using striped IP addresses for SNIPs.
Clustering can be set up either using Equal Cost Multiple Path (ECMP) or using cluster link aggregation.
ECMP is a routing protocol; it defines that a route has multiple paths to the destination with the same cost. This means that if I have multiple roads that I can travel to get to a destination and the distance is equally far, I just choose one of the paths. This way we can distribute traffic between the paths.
Different network vendors have different mechanisms to handle ECMP traffic. For example, Juniper uses a hash algorithm to determine if a packet should travel one path or another. Citrix has written an article on how to configure this using a Cisco Nexus 7000, which is available at http://support.citrix.com/proddocs/topic/ns-system-10-map/ns-cluster-traf-dist-ecmp-tsk.html.
Cluster link aggregation is an extension of link aggregation, which means we have one interface connected from all the nodes to a switch to create a virtual interface. So, instead of a regular link aggregation where we have multiple interfaces from the same appliance, we have one interface from many appliances.
Note
It is important to note that not all features running on a clustered environment are supported, for example, NetScaler Gateway. A list of supported services running on a clustered deployment can be viewed at http://support.citrix.com/proddocs/topic/ns-system-10-5-map/ns-cluster-feat-supp-ref.html.