Clustering (Multiple Servers, One Operating System)

Clustering technologies enable multiple servers to work in unison to present the appearance of a single computing environment. Technically, each server is running its own operating system, but they work together as if they were one.

Examples of clustering technologies include:

Serviceguard Clusters
Extended Campus / Extended Distance Clusters
Metropolitan Clusters
Continental Clusters

Serviceguard

A Serviceguard cluster is a networked group of HP Integrity or HP 9000 servers (known to Serviceguard as nodes) having sufficient redundancy of software and hardware that a single point of failure will not significantly disrupt service. The ability to continue operating in spite of a software or hardware failure makes Serviceguard clusters highly available.

Technology Summary

When you configure a Serviceguard cluster, you set up as many redundancies as possible, both in hardware and software. To provide the greatest degree of availability, Serviceguard typically works in conjunction with other high availability products such as:

MirrorDisk/UX
VERITAS Volume Manager (VxVM)
Disk arrays using various RAID levels
HP Powertrust power supplies (Uninterruptable Power Supplies)

With Serviceguard, you configure packages (collections of HP-UX and application processes) on the nodes of your cluster. There are three types of Serviceguard packages:

Failover: The most common type of Serviceguard package is the Failover Package. A Failover Package is installed on multiple nodes in a Serviceguard cluster, but runs only on one node at a time. The node running the primary copy of a Failover package is called the primary node. Should the primary node fail, a copy of the package on another node (known as an adoptive node) takes over with little or no interruption in service until the primary node can be restored to service. Multiple adoptive nodes can be defined for extra redundancy.
Multi-node: Multi-node Packages run (simultaneously) on one or more nodes in a Serviceguard cluster. A multi-node package can be configured to remain running as long as at least one copy of the package remains running. Multi-node packages do not fail over. HP supports them only for specific applications.
System Multi-node: System Multi-node Packages run (simultaneously) on all nodes in Serviceguard cluster. They remain running as long as all copies of the package contents remain running. All copies of System Multi-node packages halt if even one of the copies in the cluster halts. System Multi-node packages do not fail over. HP supports them only for specific applications.

Serviceguard clusters use TCP/IP networking services for reliable communication among nodes in the cluster. This includes heartbeat messages (signals from each functioning node that are central to the operation of the cluster).

Tools to Administer/Configure

Serviceguard clusters can be configured and managed using Serviceguard Manager or using Serviceguard commands, including:

cmviewcl: View information about a high availability cluster
cmrunnode: Run a node in a high availability cluster
cmhaltnode: Halt a node in a high availability cluster
cmruncl: Run a high availability cluster
cmhaltcl: Halt a high availability cluster

How to Obtain

Serviceguard is included with the Mission Critical Operating Environment and with various specialized mission-critical and virtualization bundles. Consult your HP Sales Representative or an HP authorized reseller for more information.

For Further Information

The following Serviceguard document contains detailed information on configuring Serviceguard clusters:

Managing Serviceguard, Thirteenth Edition, B3936-90105

Other Cluster Types

Serviceguard clusters can be located in a single room or single building; additionally, they can participate in clusters spread over vast distances. The distance between nodes largely determines which form of clustering you are using.

More information about the following cluster types is available in the document Designing Disaster Tolerant High Availability Clusters (HP Part Number: B7660-90013), available on docs.hp.com.

Extended Distance Clusters

An extended distance cluster contains alternate nodes located in different data centers separated by some distance. Extended distance clusters are connected using a high speed cable that guarantees network access between the nodes as long as all guidelines for disaster tolerant architecture are followed. The maximum distance between nodes in an extended distance cluster is set by the limits of the data replication and networking technology used.

Sometimes referred to as an extended campus cluster, this cluster type services an area approximately the size of a college or corporate campus.

Metropolitan Clusters

A metropolitan cluster is geographically dispersed within the confines of a metropolitan area requiring right-of-way to lay cable for redundant network and data replication components.

Sometimes referred to by the shorter term metrocluster, this cluster type services an area approximately the size of a town or city, or several closely located cities.

Continental Clusters

A continental cluster is a group of clusters that use routed networks or common carrier networks for data replication and cluster communication to support package failover between separate clusters in different data centers. Continental clusters are often located in different cities or different countries, and can span hundreds or even thousands of miles.