Guidelines for Configuring Virtual Partitions on Cellular Platforms

15
The percentage of base versus floating memory in the virtual partition (as long as the virtual
partition had the prescribed minimum amount of base memory [4]) did not seem to make
any noticeable impact on the performance of these two workloads. Other workloads may
exhibit different sensitivities to the choice of the memory granule size and the percentages of
base versus floating memory in a virtual partition.
Guidelines for Virtual Partition Configurations
Given the wide range of requirements for customer workloads and the various ways
customers could configure their nPartitions, the vPars Monitor does not have the intelligence
and higher level understanding of the intended usage of the vPars within an nPartition. It
cannot make independent decisions regarding localities of resource assignments to each
virtual partition. Instead the virtual partition user is given full control for assigning the virtual
partition resources via a rich set of vPars command options.
The following are a few guidelines that might help a typical virtual partition customer. Please
note that these guidelines are quite generic in nature, and are not meant to be interpreted as
a direct outcome of the results from the focused study discussed in this paper.
1. Resource spanning. Customers should create nPartitions with the least possible
number of cells and try to pick cells that are closest to one another (that is, cells that have
zero or the least number of crossbar hops between them).
2. Resource locality. While configuring/modifying nPartitions with large
3
number of cells
it is important to take into consideration the layout of the cells, and the number of crossbar
hops. A great majority of customer workloads perform better on a cellular platform when
attention is paid to resource localities. Users are highly encouraged to configure sufficient
4
amount of CLM for each cell in the nPartition.
3. Amount of resources in each virtual partition. Tools like HP Integrity Essentials
Capacity Advisor can be used to collect the resource usage profiles over a period of time
and analyze this data across the workloads that are being consolidated on a given target
platform. Users are encouraged to use tools like these to get an estimate of the amount of
resources needed for each of the workloads and, hence, each of the virtual partitions being
co-hosted on an nPartition.
4. Memory granule sizes. The memory granule size chosen determines the minimum
granularity of memory assignment to the virtual partitions, and also has an impact on the
amount of time it takes to do dynamic memory deletion/addition. The HP-UX 11i v3 kernel
limits the large page size to the size of the granule. Hence, if the chosen memory granule
size is too small it can lead to a loss in performance especially if the virtual partition contains
applications that benefit from large memory page sizes.
5
Users are encouraged to do some
baseline performance studies for their workloads and then configure the size of the memory
granule.
3
The definition of large depends on the size of the server. nPartitions with 4 or more cells are considered large.
4
The definition of sufficient depends upon the nature of the workload and the size of the nPartition. A few workloads have shown good results in
nPartition configurations having 3/4 or 7/8 of the memory in each cell configured as CLM.
5
The sensitivity to memory granule size is low as was observed in the study discussed earlier in the paper.