Avoiding Network Problems

I l @ ve RuBoard

According to a 1998 survey by the MERIT Project, sponsored by Computer Associates, networking was identified as the area most in need of management automation to reduce outages. Network downtime can be prevented by watching for trends in performance that could lead to overloads, and by providing redundancy, so that single network component failures are not noticed by users.

Network outages are often due to network bandwidth problems. By using tools such as NetMetrix and PerfView, increases in network usage can be identified and addressed before they become problems. Users can reduce network bandwidth requirements by using tools such as compress on files before transferring them over the network. The previous section discusses ways to avoid network performance problems.

You can't keep a networking component from failing, but you can set up your system so that users are not impacted by the failure. To prevent the downtime associated with a NIC failure, high availability software and redundant LANs can be used. MC/ServiceGuard can detect failed LAN cards and automatically configure a backup LAN card to take over the work. MC/ServiceGuard also supports the ability to automatically move TCP/IP addresses on Ethernet, FDDI, Token Ring, 100VG, and ATM LANs. High availability networking products that provide additional protection are available from HP, such as HA NFS and HA ATM.

You may also want to use MC/ServiceGuard to protect your network application. It can detect application failures and restart the application locally or move it to another system if the problem is persistent.

You should take advantage of the support for redundancy in various network services. The Domain Name Service, for example, enables you to configure primary and secondary servers. Name resolution can also be set up to rely on a configuration file or an alternate name service if the DNS server is unavailable. By setting up this redundancy, you can avoid the connectivity problems that would result from a failed name server.

High availability software was mentioned as a way to mask network failures. However, a fault needs to be detected and corrected before any additional failures cause unplanned downtime. EMS can be used to report the failure of a NIC. Use any planned downtime as an opportunity to replace the LAN card.

I l @ ve RuBoard