CPU Instant Capacity

The first Utility Pricing Solution that HP delivered was CPU Instant Capacity. This provides the ability to acquire a system with extra CPU capacity for future upgrades. The CPUs are physically present in the system but are inactive so they don't participate in process scheduling by the operating system.

This section focuses primarily on the permanent activation of Instant Capacity CPUs. We will cover temporary activation in the next section.

CPU Instant Capacity Use Model

The use model for Instant Capacity is really quite simple. The goal of Instant Capacity is allow you to defer the cost of an upgrade you know you will need in the future. This upgrade could be needed because:

• You are expecting an increase in the number of users for the application the server is being purchased for

• The server is being purchased for a consolidation environment and the applications will be migrated to the new server gradually over the course of a year or more

• You expect to be deploying new applications on the server in the future

As part of the acquisition process for the server, you will work with your HP Sales technical consultant or authorized partner to ensure that there is sufficient spare capacity on the system to meet the needs of expected upgrades. The norm in the past was to pay the full cost of these resources from the beginning. Using Instant Capacity in this environment is an improvement because:

• The bulk of the cost of the extra resources needed for the new workloads is deferred.

• Capacity planning becomes easier because even if you overestimate the need for resources, you will have extra resources to apply to other workloads. Essentially, the cost of overestimating is lower because you only pay a fraction of the cost for the unused capacity.

One thing to note is that the CPUs can be activated while the system is on line. To take advantage of this you will need to ensure that the extra CPUs are physically located in the nPars where they will be activated. Once you have determined how many Instant Capacity CPUs you want, you purchase what is called a right-to-access (RTA) license for the CPUs. When you need to upgrade the system, you simply send a purchase order to HP for the balance of the cost of those CPUs. HP will respond with information about how to access the Utility Pricing portal to get a right-to-use (RTU) codeword that can be used to license the additional CPUs. Once you have applied the codeword to any HP-UX partition in the system, you can then use the icod_modify command to activate processors in the relevant partition.

CPU Instant Capacity Requirements

The Instant Capacity program is a combination of a financial services agreement and the technology to implement it. The general principles of the program include the following:

• Customer purchases the RTA license for the Instant Capacity processors when the system is purchased.

• Customer must maintain the Instant Capacity software running on all the HP-UX partitions on the system or HP will assume that all of the resources assigned to the partition are active.

• Customer must agree to migrate to newer versions of the Instant Capacity software when they are released.

• The customer can assign a system contact that will receive e-mails from HP when there is a change in the status of a CPU or a compliance exception has occurred.

• If the system is moved to another location, then you need to notify HP ahead of time. Generally speaking, HP won't have any issues as long as the system isn't being moved to a different country.

Some other benefits of the Instant Capacity program include:

• There is no need to purchase HP support or software (OE) licensing for the additional CPU until it is activated.

• Instant Capacity CPUs can be factory or field installed.

• There is no premium for purchasing the CPU via Instant Capacity. The RTA cost paid up front is applied to the cost of the CPU when it is activated. In addition, if the cost of the CPU decreases between the time the system is acquired and when you activate the CPU, you pay the reduced cost.

• Instant Capacity CPUs are used as part of the Dynamic Processor Resilience feature of HP systems. If the system detects intermittent errors on a CPU, it will take it off line to ensure that it doesn't cause the system or partition to crash. If there are available Instant Capacity CPUs on the system, it will activate one of those CPUs at no cost to you to keep the system at full capacity.

Using Instant Capacity CPUs to Do Processor Load-Balancing between nPars

An interesting feature of Instant Capacity is that the number of inactive CPUs is measured across the entire complex in a cell-based server. In other words, if you have a complex that has been split up into multiple hard partitions, it doesn't matter which partition the inactive CPUs are in. This is easier to explain using an example. Figure 8-4 shows a server complex with four cells, split into two nPartitions with two cells each and a total of four inactive Instant Capacity processors.

Figure 8-4. A Cell-Based Server with Two nPartitions and Instant Capacity Processors

In this picture you see that the system has been partitioned into two electrically isolated nPartitions. Each partition has two cells with a total of eight physical CPUs. However, there are two inactive CPUs in each partition, so these are really six CPU partitions, with a total of twelve active CPUs in the complex.

One other very important feature here is that these Instant Capacity CPUs can be activated and deactivated without requiring a reboot of any of the partitions. So if the workload in partition nPar1 experiences a spike in load, you can deactivate CPUs in nPar2 and activate the other two CPUs that are already physically present in nPar1. This is depicted in Figure 8-5.

Figure 8-5. Using Instant Capacity Processors to Increase CPU Capacity in nPar1

As you can see in this picture, there are still twelve active processors on the system. However, we now have eight active CPUs in nPar1 and only four active CPUs in nPar2. This makes it possible to satisfy the spike in load on the workloads running in nPar1 and the extra CPUs do not sit idle in nPar1 when they are not needed.

Of course, this can go the other way also. Suppose the workload in nPar1 is no longer busy and the workload in nPar2 experiences a spike. Figure 8-6 shows nPar2 with eight active CPUs.

Figure 8-6. Using Instant Capacity Processors to Increase CPU Capacity in nPar2

This makes it possible to satisfy the performance requirements in nPar2 with the same number of active CPUs on the system.

If you think about this, Instant Capacity allows you to:

• Scale each of these nPars from four CPUs to eight CPUs

• One CPU at a time

• While both partitions remain up and running

• While maintaining the electrical isolation between the partitions

• Without incurring any cost for activating the Instant Capacity CPUs

If you are purchasing a cell-based system and plan to do consolidation by carving the system up into multiple nPartitions, you really should consider purchasing the system with some Instant Capacity resources so you can get this flexibility. This effectively makes it possible to share the spare capacity between the partitions rather than having it sit idle in one partition when it could be used in another. If you are considering getting permanent licenses for all the CPUs, this will reduce the cost of the system and give you additional flexibility for allocating resources to the workloads that need them.

Several important considerations:

• This does require the system to have inactive Instant Capacity CPUs.

• The partition you want to activate the CPU in must have an available inactive CPU physically present.

The key to this last point is that you should ensure that each of the nPars has enough physical CPUs in the partition to satisfy the maximum amount of capacity that might be required for the workloads running there. Of course, if you guess wrong, you can always reconfigure your nPartitions to add additional cells to the partition, but this will require a reboot, so it is preferable to have some extra capacity there in the form of Instant Capacity CPUs.

Using Instant Capacity CPUs to Do Processor Load-Balancing between Separate Servers

A new capability provided on top of Instant Capacity is the ability to migrate the licenses between systems. This is effectively the same type of functionality as described above, but you now have the ability to do this across physically separate systems. This provides an additional capability to ensure that you don't have idle processors on one system when there are workloads on another that can use them. A few key points about this new capability:

• The Instant Capacity Share Group is a new concept from HP. This is a group of systems over which you want to be able to migrate Instant Capacity licenses. There will be restrictions on the members of each group.

• A management software component is responsible for managing the migration of licenses between servers. This software runs on any system local to the servers in the group and can manage multiple groups.

• The mechanism for deactivating and activating processors is the same as when moving between nPars.

• This capability is provided as a separate product at an additional cost.

Although there is an additional cost for this capability, this will probably be more than offset by the savings you will realize because you won't have to purchase permanent licenses for capacity that is only needed for short periods of time.