At maincubes, TEAMWORK is written in capital letters and we proudly…
Did you know that earlier this year, one of the reasons why a huge Chinese colocation facility decided to adopt open source hardware had to do with carton boxes and plastic?
When this Chinese company did their proof of concepts, they discovered something amazing. Installing 1000 OCP based servers in their datacenter takes the exact amount of time as a traditional colocation facility needs to just … get rid of the carton and plastic packing material in which 1000 legacy servers and legacy switches arrive at their facility. Never mind the fact that all these now unpacked pieces of IT equipment need to be installed in racks as well. Their conclusion: installing an OCP server requires 250 (!) times less time than installing a traditional piece of IT equipment. Of course the reason for this huge benefit is the fact that OCP equipment arrives at the datacenter is fully populated racks instead of individual boxes that still need to be unpacked. Can you imagine the mountain of rubbish they have to get rid of afterward?
Choosing open source hardware often has unexpected advantages
More and more cloud providers, SaaS/IaaS companies, and enterprise IT departments look at implementing an open source hardware strategy. Many users choose a cloud-only approach that almost always means they are investigating the option of hosting their open source hardware in an outside facility. Many colocation datacenters are aware of this opportunity and try to position themselves as a facility that is able to host open source hardware.
But are they?
An open source hardware infrastructure is very different from a traditional or legacy approach. For example, a customer using open source hardware sends racks completely filled with IT equipment to their datacenters. These racks are higher than legacy cabinets and weigh a lot more than traditional racks. In a legacy environment, a customer sends empty racks to a datacenter and – separately – a handful of trucks full of boxes filled with servers, storage, and switches. A small army of technicians then unpacks these boxes and places all these individual pieces of equipment in the appropriate racks, and connect everything together. That takes forever and produces a big pile of rubbish. Not so when it comes to OCP racks because now a truck arrives at the datacenter, it unloads a number of racks full of already connected equipment – each weighing 1,5 ton. Once they are inside the datacenter facility, all they need to do is power up the racks and connect them to a network connection.
Most traditional datacenters are not able to host racks full of open source hardware. Because these racks are far too heavy for their raised floors. The ramps to enter the building are often too steep, so it is not possible to get the racks inside the building. But even if they somehow manage to get these racks inside, the elevators often are not able to handle 1500 kilo or more in equipment.
Scaling your infrastructure
But legacy facilities are facing other problems as well. An OCP rack produces around 6,5 kW in heat. Most legacy datacenters are not able to cool that down. And in many cases, it is not possible to get enough power to the rows of densely populated OCP racks.
So here is the first thing you need to consider when choosing your open source hardware datacenter: is the facility physically capable of hosting the very heavy OCP racks?
A second question that needs to be answered is this one. Is the datacenter able to scale your open source hardware infrastructure? In proof of concept, IT departments often use a small infrastructure that consists of maybe 6 to 8 racks. No doubt, they will, by default, choose such a small environment to test open source hardware as well. There might be legacy datacenters that are able to handle such a small proof of concept infrastructure. But are they still able to support you when you want to scale this infrastructure up? To – say – 48 or 64 racks or even more? Can you imagine the weight they need to be able to handle? And the cooling capacity they need to make available?
OCP Ready certification
To help companies understand which datacenters are capable of hosting open source hardware, the Open Compute Foundation initiated the OCP Ready certification program. A colo is allowed to use the label ‘OCP Ready’ when they can prove they meet a long list of conditions that impact the building (floors, ramps, elevators), the power infrastructure (is enough power available, no central UPS), the cooling capabilities (enough cooling capacity) and much more.
Many IT departments that decide to build an infrastructure based on open source principles are obviously keen on finding a collocation partner that is OCP Ready. In many data center districts such as Amsterdam and Frankfurt, more and more of these facilities will be available in the near future. So now what? How do you choose between 2 or 3 OCP Ready certified datacenter facilities?
Now it gets interesting. Because now it is not just about technology anymore. For example, if you are a European company or European government agency, you might – for obvious reasons – want to choose an OCP Ready datacenter that is based in Europe. So a facility that preferably is physically located within the European Union and is also owned by a European company. That narrows down your options quite a bit.
Next, you might want to look at their support infrastructure. For example, is your OCP Ready datacenter of choice able to support you when it comes to IT operations? That means that this datacenter will be able to help you out with any IT related tasks. What does that mean? Well, for example, you will not have to send your own personnel to this facility to perform all kinds of operational tasks. Instead, you simply send the required equipment to the facility and have their personnel perform whatever tasks are required. In terms of costs and planning, this is a huge advantage because the new infrastructure will be up and running in no time instead of having to wait weeks or more before equipment and the right people are available.
Location, location, location
You might want to consider one more thing. Open source datacenters outperform their legacy competitors across the board. But when two or more OCP Ready datacenters compete with each other in terms of speed, availability, etc., we have to look at other criteria as well. Like location.
Here I think you should look at more than just location relative to a major Internet Exchange. Depending on the services you offer to your own customers, you might want to consider the location of your OCP Ready facility relative to other datacenters in your infrastructure as well. For example, you might have to support applications that require mirrored operations in 2 datacenters. So you are able to offer your customers 100% uptime. In that case, your OCP Ready facility needs to be geographically located very close to your other (legacy or not) datacenters. So often, your best option is to choose an OCP Ready datacenter that is based right in the middle of a major datacenter hub like Amsterdam or Frankfurt.