What is the enterprise data center? What are its characteristics? Recent data from analysts like Dell’Oro suggest that 70% of the data center market is actually found within the enterprise. The remaining 30% is split equally between the big “single application – Web 2.0” organizations like like Google, Facebook, Salesforce, service provider data centers and cloud service providers like Amazon that provide hosting, co-location etc.
This market segmentation provides an excellent basis with which to model the characteristics of each customer type. They all have common goals like reduced OPEX and CAPEX as well as a need for maximized agility, automation and orchestration between network, storage and compute resources.
While all data centers are growing in terms of overall scale and level of services being provided, if we consider the adoption of cloud computing in the typical enterprise, enterprise data centers might actually shrink, stay at today´s size or grow at only a moderate level (from a purely physical footprint perspective). Data centers can cope with the growth of applications, compute and storage requirements through higher capacity infrastructure (network, compute and storage) and through a higher degree of utilization through virtualization. This leads to fewer physical servers, a higher density of services per server and smaller port count requirements with a higher bandwidth per port in the physical data center fabric.
In general, one can use the following attributes to distinguish the different types of data centers:
– degree of standardization
– number of applications supported
– focus on cost or focus on value to the business
– workload scale
– physical scale
So let’s take a look at the large “single application”, Web 2.0, SaaS data centers of Google, Salesforce and others:
– The degree of standardization is VERY HIGH with no exceptions allowed.
– The number of different applications is VERY LOW, sometimes just “one” that is spread across servers.
– The focus on cost is VERY HIGH.
– The focus must be on value to the business – usually the data center IS the business.
– The workload scale is VERY HIGH, supporting 10,000 or more workloads. Virtualization is often not required so this can mean 10,000 or more physical servers.
– The multitenancy rate is VERY HIGH as tens of thousands of customers need to be separated. This is often managed within the application .
– The physical scale is VERY LARGE, often consisting of multiple large data centers that occupy dedicated buildings or even clusters of buildings.
How does this compare to a service provider data center?
– The degree of standardization is MEDIUM, as individual customer requirements need to be accommodated.
– The number of different applications supported is VERY HIGH, based on the variety of applications.
– The focus on cost is HIGH, resource efficiency is important and the focus on value to the business is HIGH for differentiated services.
– The workload scale is VERY HIGH with 10,000 and more workloads. Virtualization is a key component but the number of servers still exceeds 10,000.
– The multitenancy rate is VERY HIGH as thousands to tens of thousands of customers need to be separated, both physically and/or logically.
– The physical scale is VERY LARGE, often consisting of multiple data centers of large size that can span an entire floor or multiple floors of a facility.
So how do these examples compare to an enterprise data center?
– The degree of standardization is rather LOW, driven by diverse business requirements and adaptation to their individual requirements.
– The number of applications is MEDIUM, supporting anywhere from a few hundred to a couple of thousand depending on the overall scale of the business.
– The focus on cost is MEDIUM, resource efficiency is important but the focus on value to the business is more important and thus VERY HIGH.
– The workload scale is LOW to MEDIUM, with only hundreds to a couple of thousand workloads. Virtualization is a key component and keeps the number of physical servers down to the hundreds or low thousands.
– The multitenancy rate is VERY LOW as only tens of different business unit customers might need to be separated. M&A activities still need to be supported, however.
– The physical scale is SMALL, usually with multiple data centers of average size that span a single room – less than a thousand square meters each.
Also, there are niche markets like high frequency trading (HFT) or research data centers for high performance computing (HPC are sometimes similar to Web 2.0 datacenters) that have only a few things in common. Unfortunately, the discussion in the market doesn’t often differentiate among all these diverse segments and tends to focus on the big data center implementations. However, it is important to make yourself aware of where you fit in this model, what your business and technical requirements are, and what your best solution fit is. The main focus at Enterasys and for our OneFabric architecture is the 70% of the market that requires an enterprise data center. This does not exclude the other markets, but it sets the stage for how we can provide the strongest offering to our customers.
I’ll be taking a closer look at the enterprise data center and its unique requirements in an upcoming webinar entitled “Software Defined Networks (SDN) – Where will it take networking?” on April 4th at 11:00 am EST. Please register today.