Temporary Instant Capacity


It was a bit of a challenge to separate this section from the previous one because the benefits are tightly coupled. Temporary Instant Capacity relies on the fact that Instant Capacity is available. One of the reasons it makes sense to separate them is the fact that there may be some benefits of TiCAP that would lead you to want Instant Capacity in the first place.

Key Benefits

Given that Temporary Instant Capacity requires CPU Instant Capacity, we will focus here on the additional benefits you get by purchasing TiCAP licenses for your Instant Capacity CPUs. These include:

  • Instant turn-on capability for Instant Capacity processors

  • The ability to activate capacity for short periods of time

  • It can be used on systems that were purchased as capital equipment rather than leased

In order to remove the need for e-mail to be sent to HP from the system, Instant Capacity now requires a codeword before you can permanently turn on a CPU. This means that Temporary Instant Capacity is required to instantly activate a processor. To resolve this, HP is now providing a balance of Temporary Instant Capacity with each Instant Capacity CPU purchased. This makes it possible to temporarily activate the CPU and acquire the codeword to permanently activate it later.

Temporary Instant Capacity gives users the ability to react to short-term spikes in load on applications. These spikes can be expected or unexpected. They can be the result of more users on an application or a failover. The important thing is that rather than having to purchase the spare capacity up front that is needed for these spikes, you can use TiCAP to react to them when they are needed. This reduces the up-front cost of the system, increases the average utilization of the system, and still provides the spare capacity required for these spikes in load.

Temporary Instant Capacity is a utility pricing alternative to pay per use that supports systems that have been purchased as capital equipment rather than leased. It also provides more control over your costs because you decide when to purchase capacity and how much.

Key Tradeoffs

This will be a really short section. There were a couple tradeoffs in this section, but they have all been addressed by version 8 of Instant Capacity. Version 8 eliminates the need for e-mail from the server and the new Multi-Complex Instant Capacity solution allows a shared pool of Temporary Instant Capacity.

Temporary Instant Capacity Sweet Spots

There are several situations where Temporary Instant Capacity is really useful: reacting to short-term spikes in load and managing capacity on a failover system.

Variations in load are often significantthe load during a peak time on a weekday, or during overnight batch processing can be an order of magnitude larger than the load over the weekend. The average utilization of a system can't possibly get over 30%40% unless a workload management tool is used to share spare capacity or some solution is implemented to activate capacity when the spikes occur. This is the sweet spot of TiCAP: you can comfortably run a system at a high level of average utilization because you know that additional CPUs can be activated instantly when one of those spikes occurs.

Sweet Spot #1: Use Temporary Instant Capacity to React to Short-Term Spikes in Load

Temporary Instant Capacity gives you the ability to instantly react to expected or unexpected spikes in load. This allows you to run your system at a higher level of average utilization while still maintaining a comfortable surplus of capacity.


The second sweet spot for Temporary Instant Capacity is that it provides a low-cost failover server. It is common for companies to set up clusters so that production applications will failover onto a server that is running a lower-priority workloadsuch as batch, development, or production testing. Often those lower-priority workloads don't need anywhere near the capacity required for the production workload. The combination of Instant Capacity and Temporary Instant Capacity makes it possible to purchase a system with sufficient capacity for the production workload but have permanently active capacity that is sufficient for the lower-priority workload. The rest is inactive Instant Capacity. Then, when a failover occurs you can turn on whatever capacity is required for the production workload at the time of the failover. If the load is light when the failover occurs, you may need to activate only a small number of processors, but if the load is heavy, you will still have the capacity required. This is the perfect opportunity to let HP to pay for most of your spare capacity. In this case, you will rarely use that capacityat least that is the hope.

Sweet Spot #2: Inexpensive Failover

Temporary Instant Capacity makes it possible to lower the cost of a failover system. Simply activate the capacity required only when the failover occurs and deactivate it when the workload migrates back to its primary server.


One other sweet spot for Temporary Instant Capacity is when a customer would really love to get a PPU utility solution but for whatever reason can't lease a system. The combination of Instant Capacity and TiCAP provides a very similar flexible cost structure, but the bulk of the system is purchased and can be depreciated as a standard capital asset. This has the added advantage that you have full control over the utility costs. You choose when to purchase Temporary Capacity, how much to purchase, and when to consume it.

Sweet Spot #3: Utility Pricing without Having to Lease the System

Temporary Instant Capacity makes it possible to pay for only the capacity you use. This provides much of the benefit of PPU, but it can be implemented on a capital asset.


Deciding When to Permanently Activate an Instant Capacity CPU

This discussion is in the Temporary Instant Capacity section because we are specifically addressing the issue of how to determine when to do a permanent activation based on how much TiCAP you are consuming. This question isn't as simple as it might seem. At first glance, it would make sense that if you are consuming 30 CPU days of TiCAP in less than 30 days, you should activate a CPU permanently. The reason this isn't the right answer is that TiCAP gives you the ability to activate any number of CPUs, whereas the permanent activation only gives you one. Therefore, if every time you activate a CPU, you activate 10 of them (maybe because this is being used for failover only), then activating a single permanent CPU will reduce your consumption of Temporary Instant Capacity by only 10%. This meager savings probably wouldn't be offset by the cost of the permanent activation for quite some time.

So, the key issue is return on investment. How long will it take for the Temporary Instant Capacity savings to recoup the cost of the permanent activation? You will need three data points for this:

  • The percentage of time in each day there was at least one Temporary CPU active over the last month or more. The key is that you ignore CPUs 2 through Nall you care about is the first one. This is because this is the only CPU you will stop paying for when you permanently activate one.

  • The cost of permanently activating a CPUall you care about here is what has not yet been paid. The initial cost for the RTA license is a sunk cost and therefore should not be considered for this decision. Also consider that the remaining cost will probably go down as the cost of the CPU goes down over time. You will need to get the current cost from your HP sales representative or authorized partner.

  • The cost of a 30-CPU-day TiCAP licensedivided by 30 so you can have daily costs.

This is the formula:

Permanent activation cost / (Daily TiCAP cost * percentage used by first CPU)

This formula yields the number of days it will take for the savings in Temporary Instant Capacity to pay for the permanent activation. As an example using fictitious numbers, let us assume that the permanent activation cost is $1,000 and the TiCAP cost for 30 days is $300 (which means $10 per day). We could use the percent utilization to determine how long it will take to recoup our investment in a permanent license. If the average utilization of the first CPU is 100%, which means that the CPU is always on, it would take 1000/(10 * 100%) = 100 days. So the investment would be recouped in roughly 3 months. Now, if the average utilization of the first CPU was 10%, it would take 1,000/(10 * 10%) = 1,000 days. This is almost 3 years.

Armed with this information, you can decide what level of return on investment is sufficient for your organization and determine when to activate a permanent CPU based on levels of utilization.



The HP Virtual Server Environment. Making the Adaptive Enterprise Vision a Reality in Your Datacenter
The HP Virtual Server Environment: Making the Adaptive Enterprise Vision a Reality in Your Datacenter
ISBN: 0131855220
EAN: 2147483647
Year: 2003
Pages: 197

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net