I am going to talk about SSP in my first post in the blog, this is a very interesting and large subject so I don't mean to tell everything about it. For that , I recommend to read the chapters about SSP and processor virtualization in the RedBooks IBM PowerVM Virtualization: Introduction and Configuration and IBM PowerVM Virtualization: Managing and Monitoring.
I have tried to summarize what I have read about this technology and put together the most important elements to understand them better.
The first thing to take into account is that you will need PowerVM standard edition to take advantage of this.
When you create a LPAR, you can choose using dedicated or shared processors. This post will be focused on using shared processor and its advantages.
Shared processors are physical processors whose
processing capacity is shared among multiple lpars and shared processor pools is a capability that allows processor cores to be moved between lpars based on workload demands.
LPARs which use shared processor can be grouped into shared processor pools. By default, there is one Pool with ID 0 whose values cannot be changed.You can modify the rest of shared pools or create new ones and configure its values.
Two reasons for using shared processor pools are reduce software license charges. For instance, you can create a shared pool with a maximum number of processors and add lpars to that pool, you can set a limit of 3 processors and 5 lpars inside that pool. The lpars would share those processors depending of its workload and the configuration that you make and you would only have to pay license for 3 processors and maximize physical processor usage.
Some definitions of the key elements to understand SPP:
entitlement is the number of processing units that are currently allocated to a partition
virtual processor is the representation of the physical processor(s) that are seen by the operating system. The processing capacity is distributed among the virtual processors.
For example, if a micro-partition has 1.60 processing units and two virtual processors, each virtual processor will have the capacity of 0.80 processing units.
capped shared partitions cannot use more processing capacity (entitlement) than assigned.
uncapped shared partitions can use more processing capacity (entitlement) than assigned, it is limited by the number of virtual
processors assigned or the maximum processing
unit allowed by the shared processor pool that the logical partition belongs to.
uncapped weight is a number in the range of 0 through 255 and is only used when there are more virtual processors
ready to consume unused resources than there are physical processors
in the shared processor pool. All logical partitions compete
for the same unused physical processor, even
though they belong to different shared processor pools.
If the LPAR is uncapped, a good practice is set the desired processing units to cover a major portion of the workload and desired virtual processors to match the peak workload, virtual processor desired should be between 50 to 100 % greater than processing units
The unused processor capacity available in the Shared-Processor-Pool is distributed among all runnable uncapped micro-partitions, based on the uncapped weight proportion of each
The minimum and maximum settings in the HMC have
nothing to do with resource allocation during normal operation. Minimums
and maximums are limits applied only when making a dynamic change to
processing units or virtual processors using the HMC.
When you create a SPP you can set two attributes:
Maximum processing units: defines the upper boundary of the processor capacity that can be utilized by the set of lpars in the Pool. The Maximum Pool Capacity must be equal to or greater than the Entitled Pool Capacity.
Reserved processing units: The system administrator can assign an entitled capacity to a Shared-Processor Pool for the purpose of reserving processor capacity from the physical processor for the use of lpars in the Pool. It is distributed among uncapped lpars in the Pool according to their uncapped weighting.
The Reserved Pool Capacity ensures that there will always be some capacity to be allocated above the entitled capacity of the individual micro-partitions.
There is another attribute which cannot be configured, entitled pool capacity and it is the sum of the entitlement capacities of the micro-partitions in the pool plus the Reserved Pool Capacity.
The diffrerence between the maximum pool and the entitled pool will be the additional processor that lpars will be able to receive from other pools
If you have a large number of lpars on your system, it is a good practice to start your critical lpars first, from the biggest to the smallest. This helps you to get a better affinity configuration for these lpars because it makes it more possible for the POWER hypervisor to find resources for optimal placement.
Monitoring
The nmon tool has a metric that provides the number of “unfolded VPs.”
You can use nmonanalyser to see this information on a time-based graph.
You can also use # mpstat –s 1 9999 to check the number of folded virtual processors. You’ll see utilization at 0.00 percent for the folded virtual processors.
If your LPAR is constantly running with processors folded away, consider reducing the number of assigned virtual processors since some of them are never being used. If an LPAR’s virtual processors are rarely or never folded, consider increasing the number of virtual processors to allow the LPAR to complete more work.
Another useful nmon metric is virtual processor utilization, which is listed as VP%. If you constantly find your VP% to be low, consider reducing the number of virtual processors assigned to your LPAR.
No comments:
Post a Comment