We tend to resist change. We may also need some time to learn to use new solutions.
I thought of it by driving my new car. As quite a lot of modern cars do, it has a “stop and start” device allowing the engine to adjust power in response to a red or green light. The idea is to save energy and to decrease pollution. I remember a friend who used this device two or three years ago. I considered the inconvenience (noise and vibrations during the restart) coupled with little potential savings were killers for this approach.
Nevertheless, this technology progressed well with sensors and the intelligence to calculate the right moments to adjust power (for example the device is off as long as the oil is not warm enough). The net result today is a saving of about 10 percent on my gas consumption!
Today for me “stop and start” is similar to the unused lamps in my house. You only have to pay for what you use!
If we apply this established fact to computing, few people imagine being able to cut unused resources and activate them on request. Nevertheless this concept is essential in areas surrounding cloud and Smarter Computing to bring some flexibility and reductions of cost, just as in the automobile industry or with electricity.
A few years ago the CFO and CIO did not look at this aspect which has become priority for them connected to the current crisis. Today the use of the Capacity On Demand (COD) is a real differentiator and presented in typical initiatives like “Fit for Purpose” (F4P) developed by IBM teams.
IBM developed various solutions for the resources on demand arena. I often use these approaches with high-end Power architectures.
The decision-makers appreciate these solutions issued from mainframes and large systems. Naturally the financial calculations depend on various sales offers and every customer is unique.
I would like to discuss three innovative and interesting approaches illustrating this COD concept:
Permanent Capacity on Demand
Permanent Capacity on Demand is the first stage of the path to flexibility. You buy and install a number of processors planned for your target architecture in three to five years. You will “activate permanently” on the fly, following your rhythm, with processors necessary for your growth.
Compare a classic order of setup to activation through a software key. This reduction of set up time and easy management of unforeseen increases in load are appreciated by the clients.
Temporary Capacity on Demand
Pay by processor day
This solution goes further than the previous one. The activation is “temporary.” This is very similar to the “stop and start” device in the cars I mentioned earlier. You pay only for what you use on average and adjust the temporary peaks such as the end of the month or end of the week.
A meticulous calculation must be made according to the applicable tariffs and the use of the machine. With several of my clients, we demonstrated a savings of roughly 20 to 30 percent on CPU costs compared to a continuous activation of the underutilized processors. The solution allowed a prompt response and adapted for temporary peaks. This approach also improves the cluster usage.
With a smart infrastructure approach, the time of using two identical machines for an active or passive cluster at 100 percent is over. By definition the machine on standby is not used in more than 99 percent of the cases and is expensive.
It is then judicious to activate a minimum of resources on this passive machine. We will use the temporary COD to activate the whole of the machine identical to the defective one only when the switch is happening. The process of inflating the number of processors is automated through the software of high availability (such as PowerHA). Still, a substantial economy of about 30 to 40 percent of CPU costs reduced is often observed.
Capacity Backup (CBU)
Capacity Backup for various Power models help in the event of Disaster Recovery Plan (DRP). The behavior described above for local clusters can be extended to the whole of the production sites and DRP.
Whole or part of these “on demand” devices also exists for memory. It should be noted that the mechanisms are different between CPU and memory in terms of timing and usage. A more detailed technical study is often necessary to choose between various memory solutions. In a future post I will address one of these interesting mechanisms.
So don’t wait to use the “stop and start” mechanisms to start your journey with IBM Smarter Infrastructure. Don’t think that commodity and cheap devices are the solution. Be smarter and improve your infrastructure with smarter solutions such as COD.
Philippe Lamarche is currently an IBM Infrastructure Architect after a long career in the hardware division (STG), working with French industry customers and System Integrators. He has spent over 30 years at IBM in different technical positions. He is a Certified IT Specialist and Infrastructure Architect at expert level. He is involved in current key French cloud or PureFlex projects, especially in the industry area. Philippe runs Smarter Computing workshops in France and has participated in two Power AIX-related Redbooks projects. You can reach him on Twitter: @philip7787.
To effectively compete in today’s changing world, it is essential that companies leverage innovative technology to differentiate from competitors. Learn how you can do that and more in the Smarter Computing Analyst Paper from Hurwitz and Associates.