For some time now containers have been a hot subject in the data center industry, but so far it has primarily been seen as a unit of scale for giants such as Microsoft/Google and as a fast way to implement extra capacity for companies who have run out of space or cooling in their existing data center space.
Recently we have seen an interesting turn though, as colocation providers have started adopting the concept. For instance multiple CRG West facilities have been expanded to offer housing of Verari Forest and HP POD containers, and i/o Data Centers have deployed SGI’s Ice Cube at their Phoenix ONE facility.
Are these early adapters of a future standard within the industry, or are they burning their hands on something that is doomed to fail? Although containers are not a match for all needs I definitely see potential in them, which I thought I would share with a post about some of the strengths and weaknesses I see in relation to colocation.
First of all the containers give a lot of flexibility, as it will be very easy to fit individual containers to cover specific needs in terms of power/cooling density, equipment type etc., which is harder to do in large server rooms. Furthermore it will be much easier for large clients with a lot of equipment at multiple locations, who might just configure the containers from a central location and then ship them around the world.
Adopting new technology by using containers will be easier than in traditional data center setups, as you can use containers with newer technology when expanding for more capacity. So as server depths, power consumption, cooling requirement etc. changes during the years, it is possible to buy new, or change existing, containers to fit those needs. Furthermore it will be easy to mix different containers optimized for specific vendors depending on what kind of equipment the clients are using.
One of the major advantages of shipping containers, compared to major server rooms at thousands of square feet, is the limited amount of people who will have access to it and shares the same infrastructure of the container.
That means fewer clients will be affected by an outage caused by infrastructure in the container as well as vandals or thieves getting in to a container.
Last but not least the scalability perspective is definitely also a big plus in relation to colocation, as it will allow providers with multiple data centers to move capacity around between the data centers. Generally the deployment time of expanding a facility should also be much shorter than with a traditional data center.
One of the biggest weaknesses is without any doubt that you will loose some of the advantages that you would normally see in economy of scale, as some infrastructure will be required per container that has to be divided on fewer cabinets than in for example a large server room with a lot of cabinets. Furthermore containers will typically require a lot of waste area to be able to move them around, so it will not be possible to use as much of the space as it would be in a typical facility.
Of course there are also more steps ahead for maturing the concept of containers. One of the challenges with the existing containers in a colocation environment is that even though they all have a common form factor, they do differ on a lot of points – where the doors are located, how the container is connected to water/power, what infrastructure is included in the container it self and what external infrastructure is required etc. Some standardization would probably be required on some of those areas.
With the concept maturing and becoming more of a mainstream product (driving costs down), who knows if we will see colocation data centers consisting primarily of containers in the future?