As government agencies sit at roughly the midway point in their march to meeting the Federal Data Center Consolidation Initiative (FDCCI) target to consolidate at least 1,200 data centers by 2015, there is evidence of tangible results and a recognition that significant challenges lie ahead.
A recent MeriTalk report, “The FDCCI Big Squeeze,” finds that 60 percent of government IT managers are experiencing better use of IT staff as a result of the consolidating data centers, 57 percent are seeing reduced energy consumption and 47 percent indicate increased usage of more efficient computing platforms.
But as budget cuts and sequestration strengthen their grip over the next two years, the traditional “scale-up” approach to consolidating data center storage – which involves purchasing one or more very large monolithic storage systems to expand storage capacity – becomes difficult to execute. As a result, agencies are increasingly turning to a newer "scale-out" approach that allows for multiple storage systems to be aggregated together into one large storage cluster, offering similar or better performance and non-disruptive operations equivalent or better than monolithic scale-up systems.
While functional for past agency needs, the new environment challenges the value of a scale-up approach to consolidating data centers. The ability for agencies to expand storage more granularly through a scale-out approach represents a necessary alternative to having to budget for extremely large capital expenditures every few years to replace a monolithic scale-up system.
Time’s up for scale-up?
A scale-up approach to data storage is not a bad approach and, until recently, it was really the only mainstream choice in storage architecture, as building a strong scale-out storage architecture isn't easy. With dozens or hundreds of small, dispersed data centers and a relatively modest rate of data growth, many federal agencies were able to meet their requirements by purchasing a new scale-up storage array every 3 to 5 years.
However, this decade has brought new challenges. In this budget climate, making a large capital expenditure to replace a huge monolithic storage array with another huge monolithic storage array is not always feasible. Data center consolidation has reduced the number of data centers, but has also meant that agencies are trying to assemble multiple consolidated arrays into one central storage system, and in many cases, the requirements for that centralized storage system are more than any one scale-up storage array can provide. Finally, as referenced in my previous column, the growth in data scale with new federal big data applications and advances in video and analytics mean that the requirement to grow storage in a methodical manner year over year can conflict with the "replace a single big box" approach inherent to scale-up storage.
Benefits of scale-out approach
With regards to the two storage options, it isn’t solely a case of whether scale-up is meeting agency needs, but also about the expanded set of benefits that scale-out can provide.
From a fiscal standpoint, it's a lot easier for agencies to be able to expand in more cost-effective increments. Rather than having to make a large capital expenditure every four years to replace a huge monolithic array, being able to scale from 2 to 4 to 6 to 8 nodes of scale-out storage, and perhaps add/replace two nodes every year for a fraction of that cost is a lot easier to budget and plan for on an ongoing basis.
Scale-out storage also averts an approach that places all eggs in one basket. Parts of the scale-out architecture can be easily updated, and if there's ever an issue with a given storage node, it's easy to migrate data non-disruptively to another node to effect repairs without ever affecting critical agency business. Non-disruptive 24/7/365 access to data is necessary for critical governmental functions, and no agency can afford downtime with a storage system for maintenance – especially in a consolidated data center serving multiple functions.
Finally, as requirements change for application workloads over time, scale-out storage makes it easy to tier data and move individual volumes and Logical Unit Numbers (LUNs) from one type of storage to another, either on the same storage node or a different storage node.
Scale-out in action
While each agency has its own set of unique data requirements, an example of how an agency’s needs might evolve is as follows: an agency starts out with a modest data requirement, and works with a storage vendor to implement a new scale-out storage system at their new consolidated data center. As other data centers are consolidated, the agency is able to add additional nodes to the storage system on demand, non-disruptively, to grow performance and capacity as needed. In many cases storage brought in from other data centers can be joined into a scale-out storage solution, either natively if from the same vendor, or via a gateway functionality that virtualizes third-party storage and brings into a centrally managed scale-out storage offering.
As the agency grows over time, and some of that equipment starts to show its age, the agency might choose to purchase a modest set of new nodes to add into the scale-out storage cluster. Agency storage administrators are able to non-disruptively move multiple application workloads from one of the older nodes to the new nodes, and remove the old nodes without any downtime. As such, instead of a tech refresh that might have involved a significant capital expenditure and the work of several full-time employees to plan and execute the migration over several months, the agency is able to achieve the same result with a more modest incremental purchase and a few clicks of a mouse.
In order for agencies to leverage the full benefits of scale-out storage, it is important to understand the differences between some of the older scale-out architectures that imposed significant liabilities and limitations compared to traditional scale-up storage, relative to the new generation of scale-out architectures. The modern scale-out architectures are able to combine the best parts of scale-up architectures (storage efficiency, multi-protocol unified storage, integrated data protection, etc.) with the ability to implement them in a scale-out method. As long as agencies look for scale-out storage that still provides all the benefits of their traditional scale-up arrays, drawbacks are limited.
True general-purpose scale-out for shared infrastructure is still nascent within the government, but awareness is growing rapidly as agencies match budget realities to current and future storage requirements.