The Open Compute Project was a plan started by Facebook then joined by other computing leaders. These computing giants, who had datacenter hardware at the center of their own organizations, made their datacenter hardware architecture “seeable” so smaller firms could learn from the best how to be more efficient, and how to cost optimize their data centers. Since Facebook started the charge, other leaders in the industry, Apple, Microsoft, Google, and many other firms, including financial and telecom providers, have joined in.
One obvious thing that OCP (**not to be confused with our IT Xchange OCP, the Options Continuation Program**) is attempting to do for those who wish to evaluate the processes of the companies involved is to make things cost optimized for others that operate datacenters. Some however believe that no matter how well-intentioned many of these larger firms who are involved may be, the benefits may be limited to other large scale enterprises. Smaller companies that want to copy an OCP design in their own datacenter may find converting that information from the large scale enterprise to their own, smaller scale enterprise is challenging.
In addition to the price optimization piece, the larger companies hope that through collaboration they can drive innovation rather than waiting for hardware producers to innovate. They also want to lead smaller companies in an effort to remove tech that doesn’t serve a real purpose.
With the conversion of one size of structure to another in mind, Forbes evaluated OCP community to see what challenges they were facing. One challenge was that, even when the larger datacenters were used as a model, what worked for one did not necessarily work for another. Another issue within the same vein is an industry leader may have created their own, unique solution to an issue, then allowed others from the OCP market to look and learn. Some industry leaders believe this sort of “individualism” makes it challenging for smaller datacenters to learn or to ensure quality and savings when they imagine how to implement what they learn into their smaller systems.
Some in OCP would prefer to set standards (there is an OCP Foundation) to ensure that measures are set and met. Without these standards, they are concerned smaller companies may purchase hardware based upon what they see without realizing that it was a “one-off” special solution for a large company that could afford an investment in something that might fix a problem, or might not. A concern is that without these standards, an effort to save money might inadvertently cost more money for smaller companies as they waste money on things that work for Google, but might not work for them.
Another challenge is that knowing the software and hardware a larger company is using is great, but it does not fix the problem, at least not entirely. The software and hardware is an important part of the puzzle, but so is the design, and implementing that hardware and software in a way that makes the datacenter work efficiently. There is also the issue of being well enough staffed that if a smaller company tries different designs, someone can be dedicated to the trials and errors while doing all of the other work that is necessary. Know what hardware and software can be invaluable. Knowing the design can be invaluable. Understanding that what works for Apple may not remotely make sense for a smaller company might be even more invaluable.
Vendors seem to love putting a sticker on the equipment they sale (at least a hypothetical sticker). According to Forbes, most of these vendors do not have service and support capabilities that the vendors who sell to the big boys have. Those who deal with these companies may find that there are fewer field service reps and less warranty support than they would like.
IBM POWER9 Server
In part as a response to the OCP designed datacenters, Google announced the IBM Zaius Power9 server. The Zaius P9 is based on an IBM microprocessor and OCP’s Open server specs. The goal of this server is the develop datacenter architecture that will be able to keep up with changes and advances in technology such as memory, storage, and microprocessors. The project was a collaboration between Google and Rackspace and they have submitted their design to OCP to ensure that it meets the qualifications.
The hope by Google and by the industry is that this server will eventually be the starting point for more technology based on open data. Google hopes it will impact the architecture of how it stores data in the cloud in the future.
OCP Right Now
The OCP notion began in 2011, and since then they have done a good job helping others in the market understand the benefits of non-proprietary hardware design. They have made one focus to be encouraging others in the industry to remove anything in their datacenters from the system if it does not add value or serve a definable purpose. Not only do larger companies participating in OCP communicate to others, but the “others” are able to communicate and collaborate with each other. This collaboration “sets the pace” for innovation rather than relying on vendors to set the pace.
One definition of whether an entries works if the participation. The 2016 OCP Summit, a global event for even more collaboration, attracted 600 companies and 2400 people. Not bad growth for a venture that started in 2011.