That is obviously a very important question. But at the same time the answer to that question might not be so simple.
Adoption of open source hardware is increasing rapidly and more and more companies and government agencies have started to implement open source strategies in their datacenters. At conferences and in the media we start to hear a lot of success stories, but so far the well-known consultancy firms – think Gartner, Forrester and others – haven’t published a lot of studies on the financial impact of using open source hardware. No doubt we will start seeing these studies in the near future. But so far most studies have focused on the growth of the open source hardware market.
Facts and figures
Having said that we might be able to come up with some interesting facts and figures once we start looking at the data we see coming from OCP related commercial projects that have been launched during the last few years. With ‘commercial’ I mean: non-hyperscaler users. Enterprise environments are much smaller and have a different design than the infrastructures the hyperscalers need. So it is much better to look at the numbers and percentages that have been communicated by enterprise class datacenters.
So let’s have a look at some numbers:
First of all it is clear that OCP optimised datacenter facilities are now being build at 75% lower CAPX than traditional Tier III enterprise datacenter facilities. Let’s assume a traditional CAPX for a Tier III enterprise datacenter at 8 million euro per megawatt of IT load. In most traditional cases that would translate into 200 5 kW racks. In other words: a datacenter facility overhead per rack of 40.000 euro. That means that an OCP optimized facility have a per rack overhead that is 30.000 euro lower than a traditional installation. So if you are using OCP hardware that means you can build the datacenter and also fill it with OCP hardware at approximately the same price as building a traditional Tier III enterprise datacenter infrastructure facility. What does that mean for your datacenter?
As we saw before in this blog series the telcos were one of the first group of users to recognize the benefits of OCP gear. So what have they seen in term of financial benefits? Let’s look at AT&T. John Donovan of AT&T in a presentation at Open Network Summit told his audience that AT&T has experienced a 70% reduction in CAPX and more than 50% reduction in OPEX because they moved from proprietary hardware and software to a combination of open source telco software (CORD) and OCP hardware. Comcast, another US telco, has seen a 90% (!) reduction in equipment count by moving from traditional gear to an OCP based solution.
A little closer to home: Deutsche Bank. Here again we see a massive reduction in the number of servers. When they did an open source hardware project they started out with 200.000 traditional servers. Once the migration was finished the adoption of open source gear enabled them to run 80% of their applications on just 20% of their original server count.
Last example: SK Telecom. This Korean telco was an early adopter of OCP. At the OCP Summit 2017 – this is a yearly gathering of the OCP community – they did a presentation on a number of trials they had done. Interestingly enough they did not only look at the reduction in hardware count, but also at power consumption per server, which is obviously a hot topic these days. They came up with some interesting numbers that give you a good indication of the possible savings when it comes to energy costs:
- SK Telecom compared power consumption of OCP and legacy systems under different room temperatures and workloads.
- At 20% workloads OCP gear uses between 18 and 22% less energy, at a 50% workload the difference varies from 15 to 23%.
- As the inlet temperature rises, the efficiency also rises and the benefit of using OCP gear increases even further.
- In idle state OCP servers better legacy servers under all the room temperatures.
- With extreme workloads the power consumption difference drops slightly.
Already in 2017 SK Telecom came to the conclusion that OCP systems are about 20% more power efficient than 19-inch legacy systems. Most power efficiency can be gained by using full rack configurations. In other words: fill all 43 positions in an OCP rack with servers. With half-rack configurations – using 21 servers per rack – the power efficiency decreases by 10 – 40%. And since a lot of servers are never fully utilized in terms of IT workload the fact that OCP gear uses 20 to 40% less energy at workloads of 0 to 20% is an interesting fact to keep in mind as well.
These are just a few examples. There are much more figures available. Like the number of operations personnel you need when managing a traditional datacenter versus an OCP based facility. If you are interested in those numbers (and believe me: they are as impressive as the figures mentioned above), please let me know, so we can discuss in more detail.
So what can we learn from all these facts and figures?
First of all: no, so far the well-known analysts have not published any full-blown studies into the financial impact of moving to open source hardware. But the numbers we have seen from the companies that have made the switch already are very impressive. Reductions of 70 or 80% when it comes to server count or 20 to 40% less energy usage per server – that sounds very promising, to say the least.
Experience Center can help
But no two datacenter infrastructures are the same. The architecture you have developed has a major impact on the financial benefits you might be able to achieve. So calculating your financial benefits requires a detailed study of your specific infrastructure. That is also one of the reasons why maincubes, Rittal and OCP solution provider Circle B have set up a European OCP Experience Center. Because we feel that moving from a legacy environment to an open source infrastructure requires a 2-step approach:
- First we have to work on the technical details: how to design a datacenter infrastructure that is based on open source principles? The Experience Center in Amsterdam offers you all the facilities you might need to figure out the technical details and to test drive your new datacenter solutions. We have the software, hardware and the housing available to help you do that.
- The second step is looking at the financial consequences of such a migration. No doubt the examples I have mentioned above give you an interesting indication of the financial benefits that moving to an open source approach might have.
Now let’s do these calculations based on your infrastructure, your applications and your server count. No doubt already you have done a few quick calculations yourself while reading this blog. Are you ready to have a much more detailed look at the financial benefits you might be able to achieve? Let me know. Our team is ready!
Antje Tauchmann, Head of Marketing at maincubes