18.4 LPAR Measurements

The following chart shows measurements taken on a partitioned 12-way system with the system’s CPU utilized at 70 percent capacity. The system was at the V4R4M0 release level.

Note that the standalone 12-way CPW value of 4700 in our measurement is higher than the published V4R3M0 CPW value of 4550. This is because there was a contention point that existed in the CPW workload when the workload was run on large systems. This contention point was relieved in V4R4M0 and this allowed the CPW value to be improved and be more representative of a customer workload when the workload is run on large systems.

Table 18.1 12-way system measurements

LPAR

Stand alone

Total

CPW

 

LPAR CPW

 

Average LPAR

12-way

LPAR

 

 

 

 

 

 

 

 

 

 

Configuration

Increase

Primary

 

Secondary

 

Secondary

Overhead

CPW

CPW

 

 

 

 

 

 

 

 

 

 

 

8-way, 4-way

4700

5020

7%

3330

 

1690

 

n/a

10 %

(2)

6-ways

4700

5140

9%

2605

 

2535

 

n/a

9 %

(3)

4-ways

4700

5290

13%

1770

 

1770

 

1750

9 %

While we saw performance improvements on a 12-way system as shown above, part of those improvements may have come about because of a reduction in contention within the CPW workload itself. That is, the measurement of the standalone 12-way system required a larger number of users to drive the system’s CPU to 70 percent than what is required on a 4-way system. The larger number of users may have increased the CPW workload’s internal contention.

With a lower number of users required to drive the system’s CPU to 70 percent on a standalone 4-way system., there is less opportunity for the workload’s internal contention to be a factor in the measurements.

The following chart shows our 4-way measurements.

Table 18.2 4-way system measurements

LPAR

Stand alone

Total

CPW

LPAR CPW

Average LPAR Overhead

4-way

LPAR

 

 

 

 

Configuration

Increase

Primary

Secondary

 

CPW

CPW

 

 

 

 

(2) 2-ways

1950

2044

5%

1025

1019

3 %

The following chart shows the overhead on n-ways of running a single LPAR partition alone vs. running with other partitions. The differing values for managing partitions is due to the size of the memory nest and the number of processors to manage (n-way size).

 

Table 18.3

LPAR overhead per partition

 

 

 

 

Processors

 

Measured

Projected

 

 

 

2

 

-

1.5 %

 

 

 

 

4

 

3.0 %

-

 

 

 

 

8

 

-

6.0 %

 

 

 

 

12

 

9.0 %

-

 

 

 

IBM i 6.1 Performance Capabilities Reference - January/April/October 2008

 

© Copyright IBM Corp. 2008

Chapter 18 - LPAR

299

Page 299
Image 299
Intel 7xx Servers, 170 Servers, AS/400 RISC Server manual Lpar Measurements