"Roll your own" architecture to avoid pricey vendor lock-in

The term "roll your own" used to refer to the use of rolling paper to create your own cigarettes (or other, more relaxing, brand of smoking instruments). The goal of rolling your own was to save money by buying tobacco in bulk and doing the labor yourself. This practice is still in use today, and it's growing again thanks to the now insanely expensive price of pre-rolled cigarettes. Much the same phenomenon is playing out among those who purchase reference architecture. I use the analogy here to highlight how the data center has come full circle in the way it builds out infrastructure.

Back in the day (say, last year), companies bought hardware from whomever they thought would give them the best performance for their dollar. The IT folks would then "roll their own" reference architectures for each application and provide services and support for those applications based on what they created.

An example is a database server running on Unix variants connected to the network over gigabit Ethernet or a storage area network running over Fibre Channel and backed up over a separate network through a Unix-based backup server to a Fibre Channel-based tape drive. The IT folks involved in creating these reference architectures might have included administrators for the network, database, server, storage, backup and data center elements, as well as  whoever got picked to be responsible for disaster recovery (DR). A security team might have also been involved to mitigate risk.

The architecture and all the procedures for provisioning and procuring servers, storage, network and everything else was documented and adhered to as the strategic platform of choice within the organization, and everything was monitored and adjusted as time went on to assure the architecture met new business requirements and scaled over time from a capacity planning viewpoint.

All the work required to roll your own data center infrastructure was time consuming, complex and expensive. The new approach, which is fast becoming standard, is to purchase a reference architecture directly from a vendor or partner for in-house or cloud-based operation. This new way to roll out data center infrastructure negates the need for input from all the folks who used to be involved, with the added benefit of reducing costs and complexity and improving support. The problem is some vendors are trying to lock clients into THEIR solutions, which reduces competition, drives up the overall cost again, and locks out other vendors who might offer additional benefits if included within the reference architecture.

Once locked into a solution, it's hard to get out. This issue is one of the main roadblocks to widespread cloud adoption and a major reason why many organizations are getting their feet wet in the cloud by implementing a private cloud first. Until there are standards and safeguards in place, both technically and contractually, to easily and securely migrate data between cloud providers, it may be better to continue to roll your own reference architecture for an internal private cloud.

Christopher Poelker is the author of Storage Area Networks for Dummies, the vice president of enterprise solutions at FalconStor Software, and deputy commissioner of the TechAmerica Foundation Commission on the Leadership Opportunity in U.S. Deployment of the Cloud (CLOUD²).