ICP GDT6618RD, Addendum, Page: 12 / 111

12
A. Introduction
In today's information technology world, High Availability Systems are gaining more and
more importance. The demands on these systems are not only the continuous availability
of all important data, i.e., "365 days * 24 hours", but also efficient resource sharing of the
relatively expensive components. The secret behind the popular word "Clustering" is noth-
ing more than the intelligent combination of several computer systems with the purpose to
simultaneously increase the security and computing power of the whole system. This "Clus-
ter" appears to the user just like one single system with extraordinary characteristics. A
special Clustering software executes the controlling and management of the Cluster and its
members.
There are several strategies to design Clustering systems. In a first model, the various
members (also called nodes) could only be connected with each other through the normal
User LAN. In another step, the nodes could additionally share an I/O bus (e.g., SCSI, Fibre
Channel). To avoid "Single Points of Failure" (these are vital components which cause a
general failure if they fail) in such systems, host based RAID controllers are ideally suitable
for cluster systems. The choice of host based RAID controllers for the nodes of a cluster
system implements the appropriate redundancy and security in the mass storage part of
the system right from the beginning. In contrast to this, external RAID controllers carry the
risk of becoming a single point of failure. If they fail the whole cluster system goes down.
Apart from that, these kinds of solutions are normally pretty inhomogeneous: The host
adapters come from vendor A, the external RAID controller from vendor B. Who guarantees
the flawless and interefence-free operation of both components ? Who supports you when
the system causes problems ?
Since 1990 ICP vortex has been intensively engaged in the research and development of
host based RAID products for highest performance and security requirements. Not only be-
cause of our products’ outstanding performance, our expertise and continuity in develop-
ment, but also due to our excellent customer support, ICP Controllers are accepted and
known as top leading-edge products all over the world. All ICP Controller products are
pure-bred
hardware solutions . All functionality required for the very complex tasks is hard-
ware-implemented on the controller. Thus, RAID is fully independent of the computer sys-
tem (the host) and the operating system. Software RAID solutions are normally not usable
for cluster systems. Thanks to the wide operating system support and easy-to-use installa-
tion and maintenance utilities, setting up and using high performance and fault-tolerant
mass-storage subsystems for almost every purpose is child’s play.
For a successful installation of a Clustering System it is very important that the person who
carries out the installation has detailed knowledge and experience in the areas Microsoft®
Windows NT/2000 Server, Microsoft® Cluster Server, Novell/Orion, Linux/Watchdog and ICP
Disk Array Controllers.
This Addendum contains important information, particular regarding the operation of ICP
RAID Controllers in a Clustering System.
Information on the Operating System, the functioning of the Cluster and other system de-
tails may be found in the corresponding system manuals.
For information on the installation of the ICP Disk Array Controller and the configuration of
Disk Arrays, please refer to the "User's Manual" or "Quick Installation Guide" of your ICP
Controller. These documents may also be found on the ICP System CDROM.
We would like to thank you for purchasing an ICP Controller.
ICP - Intelligent Computer Peripherals
®