The 24-hour profile consists of 24 values that correspond to each hour of the day (0 to 23). Each value is calculated by taking the peak value for that specific hour over the number of monitoring days.
Consider a system monitored for 3 days. The following chart presents the data set for its Processor workload, and the calculated 24-hour profile:
Table B-4 Processor Workload Data Set
|
Hour |
Day 1 |
Day 2 |
Day 3 |
24-Hour Profile |
|---|---|---|---|---|
|
0 |
200 |
300 |
500 |
500 |
|
1 |
300 |
300 |
250 |
300 |
|
2 |
300 |
400 |
450 |
450 |
|
3 |
700 |
250 |
350 |
750 |
|
4 |
500 |
500 |
500 |
500 |
|
5 |
600 |
400 |
500 |
600 |
|
6 |
500 |
550 |
750 |
750 |
|
7 |
400 |
600 |
800 |
800 |
|
8 |
500 |
300 |
250 |
500 |
|
9 |
450 |
350 |
640 |
640 |
|
10 |
200 |
240 |
340 |
340 |
|
11 |
300 |
600 |
+60 |
650 |
|
12 |
300 |
- |
- |
300 |
|
13 |
550 |
340 |
850 |
850 |
|
14 |
440 |
500 |
600 |
600 |
|
15 |
800 |
540 |
600 |
800 |
|
16 |
750 |
260 |
430 |
750 |
|
17 |
500 |
640 |
450 |
640 |
|
18 |
600 |
200 |
350 |
600 |
|
19 |
300 |
250 |
650 |
650 |
|
20 |
200 |
300 |
200 |
300 |
|
21 |
250 |
350 |
400 |
400 |
|
22 |
600 |
250 |
350 |
600 |
|
23 |
500 |
650 |
750 |
750 |
PlateSpin Recon considers and treats dual-core or multi-core machine as two-processor or x-processor machines.
The effect of hyper-threading is controlled by two configuration parameters: Aggregate Instance Function (SUM, AVG) and Hyper-threading Normalization Factor (1 – 2). The default value for aggregate instance function is AVG, and the normalization factor is set to 1.25. These values can be configured in the advanced settings under .
PlateSpin Recon collects each individual processor instance’s utilization values at a given time and then calculates the processor utilization based on the aggregate function. If the aggregate function is SUM, then the physical processor utilization is the sum of all logical processor instances’ utilization. If the aggregate function is AVG, then the physical processor utilization would be average of all logical processor instances’ utilization.
Consider a system with single-core processor with hyper-threading enabled at p MHz speed. For a given period of time, PlateSpin Recon calculates the percentile utilization values for 24H profile values of instance X and Y as follows.
{X}= {X1, X2, Xi…..X24}, Xi is the utilization of the processor instance X at hour i.
{Y}= {Y1, Y2, Yi…..Y24}, Yi is the utilization of the processor instance Y at hour i.
Table B-5 Aggregate Function Calculations
|
(Aggregate Function, Normalization Factor) |
( SUM, f ) |
( AVG, f ) |
|---|---|---|
|
Reports: Peak % |
Max { X + Y } |
Max { (X + Y) / 2 } |
|
Reports: Peak (MHz) |
Max { X + Y } * (1/100) * p * f |
Max { (X+Y)/2 } * (1/100) * p * f |
|
CPM: Workload calculation |
{X+Y} * (1/100) * p * f |
{(X+Y)/2} * (1/100) * p * f |
PlateSpin Recon takes both number of cores and hyper-threading into account when calculating the effective speed of a machine.
Server template specification: 2 x 3000 GHz, dual core, and hyper-threading enabled
Number of processors: P = 2 * 2 = 4
Processor capacity: C = 3000 GHz
Hyper-threading normalization factor: n = 1.25 (default value)
Scaling factor: A(p) = P / { 1 + ä (P-1) } , where ä = 0.05
A(p) = 4 / { 1 + 0.05 * 3} = 3.4782
Effective speed = A(p) * C * n = 3.4782 * 3000 * 1.25 = 13043 MHz
Energy Use (kWh/yr) = Power Usage (W) x Server Derating Factor x Number of hours in a year (8760) / 1000
Heat Dissipation (BTU/yr) = Energy Use (kWh/yr) x 3.41
Annual Energy Cost = Energy Cost per kWh x Energy Use (kWh/yr)
Energy Use (kWh/yr) = Power Usage (W) x Server Derating Factor x Number of hours in a year (8760) / 1000
CO2 Emission (1b/yr) = Energy Use (kWh/yr) x 1.54
The constant factor of 1.54 has been considered as per the standards.