Red Hat rolls out high availability server

IT managers can better manage traffic and ensure uptime for their Web servers with this month’s release by Red Hat Inc., the Linux vendor claims.

Red Hat High Availability Server 1.0 lets network professionals cluster two Linux, Unix or Windows NT servers for failover or load-balancing purposes. The idea is to avoid downtime or traffic bottlenecks on servers that provide IP-based functions, such as Web or FTP servers, mail gateways and VPN support.

“I am looking to use clustering technologies for the Unix and Windows NT hosts on my network,” says Gary Mansell, senior technical analyst for Ricardo Consulting Engineers in West Sussex, England. “I want to failover and load balance two server hosts so that during normal working conditions the load is spread between two servers but when one crashes, the other takes over its work as well as its own.”

Mansell manages 300 servers and estimates that server downtime costs him lost productivity of more than US$150 per user per hour.

During configuration of the Red Hat cluster, the administrator decides whether the software will be used in failover mode or as a virtual server. In Failover Service mode, two servers are connected to each other over the network. One server is inactive and waits for the other server to fail. A signal called a heartbeat is continuously sent from the primary to the standby server that indicates the primary server’s working condition. When the primary server has to be taken down for maintenance or fails (signaled by the heartbeat disappearing), the standby server takes over.

In Linux Virtual Server (LVS) mode, two servers front-end groups of Web, FTP or other servers to balance traffic between them. LVS can front-end heterogeneous groups of NT, Solaris or Linux servers and supports a variety of user-configurable routing methods, including round robin, least connections and IP masquerading.

In round-robin routing, the first request goes to the first server, the second request to the second server, and so on. Least-connection routing allocates traffic to the server with the least number of connections or traffic. IP masquerading is a way for servers to access the ‘Net as if they are sharing the same ‘Net address. The technique provides a secure way to hide the internal network from outsiders and saves on Internet IPv4 address space.

International Data Corp. (IDC), a market research firm in Framingham, Mass., estimates that Linux grew faster than any server operating system during the past two years and accounts for about 25 per cent of the market. NetCraft, an Internet consultancy in Bath, England, reported 36 per cent of all Web sites run on Linux servers.

Red Hat High Availability Server runs on Intel platform servers of any configuration. Software for each two-node cluster is US$2,000. It is available immediately.

Red Hat is at