).
HA is concerned with circumventing all unplanned outages and consists of both FT and RR solutions. HA generally means RR o
r clustering; the terms are used interchangeably.
The strategy with fault tolerance is to prevent the mission-critical system from coming down at all costs. While HA systems are primarily software solutions with supporting hardware, FT solutions are the opposite: hardware solutions with supporting software. Tandem Computers, which is the market leader in FT systems, uses a shared-nothing system with dual processors carrying out every task simultaneously. When a fault is noted on one processor, it is reported to the other processor, where the task continues with no downtime.
Unlike FT systems, RR systems do not have dual processors running simultaneously. An RR system must switch over to a newly started second software process and recover data as quickly as possible.
John Oltsik, a server specialist at Forrester Research, thinks that RR technology has become good enough for the majority of mission-critical applications. "RR stops short of providing the zero-fault, 100 percent uptime of the FT
system," says Oltsik. "But it is much more affordable for midrange to smaller companies and can offer uptime in the 99.999 percent range, with guaranteed data integrity."
Continuous operations is a peer branch with HA. It uses the strategy of avoiding any planned downtime, such as hardware maintenance, disk backup and replacement, CPU replacement, and so forth.
The star at the top of the tree is called continuous availability, the sum of HA and continuous operations.
illustration_link (15 Kbytes)
Clustering and redundant arrays of systems are two approaches to HA and FT.
Defining Terms
