Advertisment

Veritas Cluster Server

author-image
PCQ Bureau
New Update

VCS (Veritas Cluster Server) can support from two to 32 nodes on Windows NT/2000 and up to 128 nodes on Unix , Solaris or other enterprise-class server

OSs.

Advertisment

You can configure your cluster on any platform. All the servers in that cluster can be configured from one machine using VCS’ GUI. The GUI is loosely based on Windows Explorer. For each application configured, it shows you a tree-like structure with all the dependencies. So, if the application fails, you can see at what point the problem lies.

You can configure it on active-active servers (where each server is running independent applications serving different users and can also take over the load if a server in the cluster fails), or active-passive ones (where one or more servers in the cluster is there only to backup the primary server or servers, and is not running any other applications). In the configuration, you can specify what VCS should do when an application fails–how many times it should try to run it on the same server again, what server it should ‘failover’ to, what sort of failures to alert you about, how to alert you (e-mail, SMS, or pager), and so on.

VCS uses its proprietary LLT (Low Latency Time) protocol (which is claimed to have significantly less overheads vis-à-vis TCP/IP) for server-to-server communication, whereby each server sends ‘heartbeats’ to all the rest in that cluster giving its health status.

Advertisment

You can specify in the configuration after how much duration these heartbeats should be sent (every 30 sec is the default); whether servers should use the NIC-to-NIC connection, the shared hard disks, or your LAN to send these broadcasts; and what priority is to be assigned to these. For example, you could specify that the packets be sent through the direct connection first, then through the shared hard disks if the former is not available, and then through your LAN. In this case, if the direct connection between the servers is broken, VCS will not report a failure immediately–it’ll monitor whether the server is healthy through the other connections.

In the case of an application failure, for example, your database fails because something’s wrong with the NIC, VCS shuts it down on the server, and then copies the full thing, with all the dependencies on to the designated server. Users will not know of the back-end process but may get a waiting time if they try to access the database when the ‘failover’ is in progress.

Other features are built-in load balancing, detailed logs of all activity, and remote configuration of clusters. In the latter, you can create a cluster over your WAN link comprising servers residing in different geographic locations. You can configure, monitor, and control all of these from one location. You can also create independent clusters of servers in different locations–say a Solaris cluster running Oracle in your Delhi office and an NT cluster running Microsoft Exchange in Mumbai–and manage all of these from one location. Price: Approximately Rs 200,000 per node.

Contact: Veritas Software Solutions. Tel: 022-8303025/28 Fax: 8303043.

Unit 217, ‘B’ Wing, Bonanza, Sahar Plaza Complex, Andheri (E), Mumbai 400059.

E-mail: savio.monteiro@veritas.com

Pragya Madan in Mumbai

Advertisment