Will your users be willing
to accept poor performance for a long period of time?
Figure 7-3. An active/active cluster configuration
CHAPTER 7 n CLUSTERING 185
THE N+1 PRINCIPLE
The primary goal of clustering is to reduce the risk and impact of any single node failing. An
active/active configuration is an effective means of maximizing hardware investments, but it doesn??™t
necessarily significantly reduce the impact of a failed node. While it does reduce the risk imposed by
having a system completely unavailable, there is the possibility for significant performance degradation,
which in turn can be as disruptive as complete unavailability.
The standard recommendation in any high-availability/scalability scenario is to work from a simple
N+1 formula, where N represents the number of servers required for satisfactory performance, and 1
represents an unused, standby node. If you follow this recommended formula, losing a single server
should have no direct impact on performance. Should scalability become an issue, you can configure
the standby server to offload some of the work from the other nodes.
Pages:
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360