You will always be squeezed, so consider using a comparable pseudo-website to
the one you intend to performance test against. If your company develops many websites there
may be one with comparable size and architecture against which you could at least test the loads.
As soon as the thread of some complete action is available test that. If nothing else it will help
accustom the tester to the tool.
???
Early alerting
. Design your environment such that you can get feedback to developers and
management asap. You will be time-boxed. Here are alerts of disaster:
??“ We can??™t get a separate environment.
??“ We can??™t get the tools.
??“ The tools don??™t work.
??“ We can??™t code the tools.
??“ The tools cause the system to crash.
??“ The tools give faulty readings.
??“ The data is unrealistic/unobtainable/unusable
??“ The tools slow the system unacceptably.
???
Capacity
: scaling the system is not a question of merely adding a server (changing the structure
will simply change the stress levels within the system) but rather of identifying:
??“ When the system will run out of capacity such as to affect latency, (try rehosting a copy of
your current system on a minimal hardware set and seeing where the bottlenecks are ??” never
use the backup system for testing unless you want the primary system to fail)
??“ How much warning you will get and how you will be warned?
??“ How many parts of the system from internet connection through web server cluster to legacy
system must be changed?
??“ How system elements (such as video) will require dedicated servers, and how this will affect
server pages and load balancing.
Pages:
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209