7.1. Server considerations

7.1.1. Processors

The processing power requirement on the server is low. A minimum of Pentium grade but a dual processor SMP box is suggested to handle multiple request seamlessly. We suggest a dual Pentium III 800MHz or above. Processing power of the server does not help to improve the performance or scalability of your Aptus system at all.

7.1.2. Memory

Memory requirement of the server is a big concern, if you have a lot of files and data with frequent access, it is suggest you have more memory on the server (512Mb or above). This will allow the server to make use of the memory to use as cache to reduce disk loads.

7.1.3. RAID disks

RAID is a minimum requirement of the Aptus server for some reliability. It is suggested the server to run RAID 10,000rpm SCSI drives. This will give you a high performance and reliable file system. It is suggested to use hot-plug SCSI drives so you don't have to experience system down time during a disk drive replacement.

7.1.4. SAN (Storage Area Network)

SAN may be a good choice if you consider disk drive capacity scalability problem in the future, but generally SAN is slower than direct attached drives.

7.1.5. Network adapters

It is suggested the server should at least use a 100Mbps Ethernet adapter and 1000Mbps are preferred to high loaded servers.

7.1.6. High availability clustering and fail over

Aptus is a stateless file server. It is possible to setup a fail over and redundancy. Since Aptus clients are sharing the same root file system with the server, a complete root sharing HA cluster setup is required. Other high availability solutions that only do partial fail-over will not work with Aptus . The only proven and workable HA solution is the ShaoLin HA-Cluster which provides a complete system fail-over solution please see ShaoLin HA-Cluster webpage for more details.

7.1.7. Power supply

UPS (Universal Power Supply) is suggested for the server to have the ability to overcome power failure.

Redundant power supply on the server will have the ability to overcome power supply failure.

7.1.8. NFS server tuning

You may want to increase the number of nfsd instances by modifying the nfs server init script at /etc/init.d/nfs. In a Redhat Linux distribution, it is controlled by a line RPCNFSDCOUNT=8. Try to modify this number to equal to the number of clients serving by this server to maximize the nfs server response. See Linux NFS HOW-TO for more information about Linux NFS.