When we started with our cloud architectures some years ago, LATAM’s market was just looking to move traditional systems into high density virtual private servers. Now, the amount of applications distributed among lower density instances is starting to increase to a such point that the density of energy per rack has increased three times.
New application’s architectures requires several layers of components: app servers, web servers, databases, static data (or objects), balancers, CDNs, Anti-DDoS and so on… And to be more efficient, you’ll need to distribute these layers in small instances. These smalls instances helps you to scale out with the demand and be full automated.
Commodity hardware is being the best option to lower prices, get flexibility and provide agility to respond to the customers’ demand. Storage and network are also composed by commodity x86 boxes thanks to new trends like “Software Defined” technologies. Every rack unit requires now until 20KW to feed and cool, and high density zones start to come up among the datacenter’s clusters.
New options are required for cooling and keep PUE (Power Usage Efficiencies) at lower values. Options such close coupled or direct cooling solutions at very rack where the equipment is installed – an example of it at OptiCool site -. Investment would be higher and probably could change the Datacenter’s building design, but the operational costs are much lower – you could cool 10Kwatts equipment consuming no more than10 watts of energy –
Well, my finger is killing me and need to stop writing
See you later!