System Sizing Guidelines for Integrity Virtual Machines Deployment -- Hardware Consolidation with Integrity Virtual Machines

Expected Utilization Example

After you identify the multipliers, you can generate the utilization information in a very

straightforward way. Consider a workload with the following utilization footprint:

CPU – 15%

Disk – 5%

Network – 2%

This workload currently resides on an HP 9000 Model A500 single-CPU server with one network

interface card (NIC). The server used for consolidation is an Integrity rx4640 (1.5GHz CPU, 4MB

cache Itanium 2 processor). The new server will be connected to the same SAN that the A500 is

currently using. Some enhancements will be made to the network configuration for the new server.

Expected CPU Utilization

From experience, the workload‟s performance is known to be very similar to the SPEC CINT2000

(SPECint) benchmark. The CINT2000 results for both platforms are available – 422 for the A500

and 1372 for the rx4640. Hence, calculate the multiplier as follows:

0.307

1372

422

CINT2000rx4640

CINT2000A500



As a result, the expected CPU utilization is:

15%  0.307 = 4.6%

That is, the expected utilization will be 4.6% of a single CPU on the rx4640.

Expected I/O Utilization

The mass storage configuration is roughly the same, so the multiplier used will be 1.0. The

expected utilization will be simply 5%, as it was before.

The workload is much more sensitive to network bandwidth than latency. The new network

configuration has 80% more bandwidth than the old network. The formula to generate the

multiplier is:

556.0

8.1



Hence, the expected network utilization is:

2%  0.556  1.1%

Hence, the expected utilization of one physical NIC on the rx4640 is 1.1%.

Extending the Utilization Calculations

The previous examples explain the basic arithmetic for calculating expected utilizations, but the

impact of workload consolidation also depends on the number of CPUs and I/O cards to which the

utilization applies. For simplicity, these examples use one of each, but a real scenario may have,

for example, four CPUs and two NICs on the existing server. In this case, multiply the utilizations

by the number of hardware units (for example, a four-CPU configuration results in a utilization of

4.6%).