Performance and capacity planning
Hi there
I've been reading through the MS documentation on Performance and capacity planning and I'm now pretty confused
http://technet.microsoft.com/en-us/library/cc261795(office.12).aspx
The gist of it seems to be:
They created a test environment - several farm configurations were used for testing, ranging from a stand-alone computer to eight Web servers with single and clustered computers running Microsoft SQL Server 2005. Testing was performed with eight client computers
simulating from 32 through 256 user connections
So how do they get these test results:
Farm configuration
RPS
Total number of user connections
Light usage
Typical usage
Heavy usage
Extreme usage
Mix
Read
Mix
Read
Mix
Read
Mix
Read
Mix
Read
1 by 1
50
100
90,000
180,000
50,000
100,000
30,000
60,000
15,000
30,000
2 by 1
99
185
178,200
333,000
99,000
185,000
59,400
111,000
29,700
55,500
3 by 1
115
265
207,000
477,000
115,000
265,000
69,000
159,000
34,500
79,500
4 by 1
120
275
216,000
495,000
120,000
275,000
72,000
165,000
36,000
82,500
5 by 1
136
280
244,800
504,000
136,000
280,000
81,600
168,000
40,800
84,000
6 by 1
130
280
234,000
504,000
130,000
280,000
78,000
168,000
39,000
84,000
7 by 1
134
290
241,200
522,000
134,000
290,000
80,400
174,000
40,200
87,000
8 by 1
130
280
234,000
504,000
130,000
280,000
78,000
168,000
39,000
84,000
Where do they get the "total number of user connections" up in the 100,000s when they simulated only 256 user connections?
How is RPS calculated?
What is measured and what is derived?
I understand that they've used different farm configurations but what did they run against each config and how are the results calculated?
If anyone can shed some light on this it would be great
Cornelius had a go but his SPCAP is now nowhere to be found - SharePoint Capacity Planner is being discontinued as a result of Microsoft System Center Capacity Planner (SCCP) formally being discontinued
His useful article is here
http://www.cjvandyk.com/blog/Lists/Posts/Post.aspx?ID=100 he tries also to make sense of the MS documentation but ultimately I still can't figure out how the formulas are supposed to work - especially with SQL clusters and high numbers of users (over
100,000)
Anyone got any advice? Does anyone out there use some modelling for capacity planning or is everyone doing what we've been doing which is just keep throwing hardware at the environment until the customer stops complaining
Cheers
Jonj
May 27th, 2011 9:36am