Cloud computing is becoming ubiquitous with more players coming out with public cloud offerings (IBM Public Cloud, Korean Telecom, Tata Communications) and a number of cloud platforms have matured in the last one year including cloud.com (acquired by citrix), Cloupia and OpenStack. The latter, derived from NASA’s OpenNebula project and supported by RackSpace, is fast becoming adopted as open source cloud platform. While there is movement towards normalization between IaaS and PaaS cloud models with solutions like Cloud Foundry and OpenShift, recent forecasts place the IaaS market at $4 billion by the year 2015 (Cloud Computing and Managed Services Spending Forecast).
As the different cloud providers and platforms compete for users and enterprises to adopt their platforms, some of them have also identified and are addressing the user needs to network and federate across the distributed infrastructures. Such capital investments in large enterprises are geographically distributed and they are planned for peak load. However, in reality their loads may vary over geography and time zones resulting in significant inefficiencies in resource utilization. For example, a New York data center for an enterprise may be operating at peak with plans in place to expand the data center, while the enterprise’s France data center may be operating at 20% utilization. A cloud platform covering both the data centers will enable workloads to be distributed across the data centers. However, this creates a homogeneous cloud computing environment that leads to vendor lock-in and thus delaying crucial decisions to move to the cloud.
Forrester reports the next big leap in both technology and business models around elastic compute resources to be a market place in the form of bidding for compute resources at auction or acquiring them through a broker http://gigaom.com/cloud/future-of-cloud-computing-more-clouds-seriously/. Broker models for companies to buy access to compute infrastructure as a service on demand is realized by, for example, the clearing house solution by SpotCloud based on their Enomaly cloud platform. However, such models suffer from homogeneity of the solution and vendor lock-in. In the future, Forrester expects a “full broker” to emerge that automatically spans all clouds and acts as a “fixer” for companies seeking to deploy new apps and handle busy times of the year http://gigaom.com/cloud/future-of-cloud-computing-more-clouds-seriously/
Driving efficiencies in a federated cloud ecosystem requires consumers to have the ability to (1) search and discover best platforms and providers to host their workloads on demand, and (2) dynamically monitor and reconfigure their federated cloud based on current and predicted utilization.
Cloud computing consumers and enterprises have variety of options to choose from for their movement to the cloud. However, each workload is different and a single cloud provider may not be the best option for the workloads of an enterprise. For example, a music service can use Amazon EC2 and S3 for their website, while using a different infrastructure and service for music delivery. Certain workloads need to be executed where the data is located, while others are efficient when run closer to its users. Finding the best place to run a workload depends on a number of factors. ComputeNext’s federation solution provides consumers to find the best place to run their workloads and for providers to focus on their cloud platforms and value added services that differentiates them in a federated cloud eco-system.
This article was written by Sundar Kannan, Founder and CEO of ComputeNext Inc. ComputeNext, based out of Bellevue WA, is developing technologies towards a market place for compute resources including the ability to search and federate workloads across heterogeneous cloud platforms.