Guidelines for Configuring Virtual Partitions on Cellular Platforms

Abstract

The HP-UX virtual partitions (vPars) solution is one of the popular choices available to

customers who are considering workload consolidation on nPartitions (hard partitions on

cellular HP platforms). The vPars solution provides users with configuration flexibility and

negligible performance overhead. It is however important from a performance perspective to

understand and account for the localities of the resources assigned to virtual partitions

hosted on cellular platforms. The paper provides some guidelines for configuring virtual

partitions to achieve good workload performance on cellular platforms where an nPartition’s

resources span multiple cells. This paper also illustrates the influence of resource localities on

workload performance by using results from a focused study.

This paper assumes that the reader is familiar with the concepts of vPars and has a good

knowledge of the HP-UX operating system and its NUMA features (HP-UX support for the

non-uniform memory architecture on cellular platforms [2]).

Introduction

vPars allows a system administrator to create multiple soft partitions within an nPartition.

Each virtual partition runs its own instance of an HP-UX Operating Environment which can

be independently tuned. Each virtual partition is assigned a set of physical resources to

allow the HP-UX Operating Environment to boot and be operational. The vPars command

line interfaces offer a wide range of choices for assigning resources to the individual virtual

partitions. A thin trusted layer of software called the vPars Monitor resides between the

firmware layer and the HP-UX Operating Environments. The vPars Monitor controls the actual

assignment of resources to the individual virtual partitions based on a user specified partition

plan (vPars database). The vPars Monitor, along with the virtual-partition-aware firmware on

Integrity platforms, provides platform emulation services for the individual virtual partitions

hosted on the nPartition.

Customers who use various solutions for consolidating servers in their datacenters also have

expectations about the degree of performance impact these solutions may have on the

individual workloads. One of the key advantages of the vPars solution is that it allows direct,

non-virtualized access to the assigned subset of the nPartition’s physical resources. The

performance overhead due to the vPars software stack is therefore negligible. In fact, vPars

can help improve overall resource utilization.

On cellular platforms, memory access latencies vary depending on the type of access: local

access, remote access, or access to interleaved memory. This variation has an impact on the

performance of the workloads. The vPars software offers a high degree of flexibility for

configuring virtual partitions. Virtual partition resource assignments done without attention to

locality considerations can in some cases have a negative influence on the workload

performance.