Mixed architecture support drives private cloud adoption in Asia

By Claudio Caballero

March 15, 2013, 3:56 PM PDT


Takeaway: Private cloud projects are being deployed all over Asia. Some of the largest deployments are effectively supporting both enterprise application architectures and web-scale application architectures.

Discussions of cloud computing tend to focus on public Information as a Service (IaaS) cloud providers, especially Amazon Web Services (AWS), and their high-profile web-scale clients such as Dropbox and Foursquare, which use AWS IaaS and PaaS services to deliver their Software as a Service (SaaS) applications to millions of users on the Internet.

Although AWS (Singapore and Japan) and Rackspace (Hong Kong) are already in Asia and expanding and there are surely those planning to compete with them for IaaS public cloud customers, private cloud projects are being quietly deployed all over the region, with installations of at least 10,000 servers routine and more than 30,000 not uncommon.

A key consideration for many of these private cloud deployments is being able to effectively support both traditional enterprise application architectures and the newer web-scale application architectures of the kind we often read about being deployed on the public IaaS clouds (such as Netflix’s well documented use of AWS to power its streaming video service). These web-scale application architectures are conceived to thrive with the commodity hardware approach of public IaaS providers, where compute node failure is not a disaster to be recovered from but an expected glitch to be accommodated, like the occasional power surge from a lightning strike. When a compute node fails, the web-scale application architecture will compensate for it seamlessly, using APIs provided by the public cloud providers.

Enterprise applications, however, are designed to be run on traditional high-availability (HA) hardware, with redundancy built-in at every layer. If you are running a traditional ERP application, it does not expect the servers to fail except rarely, as part of traditional HA clusters.

The results are that, unless you are a startup or otherwise have no applications that rely on traditional enterprise architectures, if you want to reap the benefits of the cloud approach to your IT infrastructure (self-service, metering, elasticity, etc.) to, for example, reduce the time it takes to provision new computing resources (even fully virtualized shops in traditionally managed data centers can sometimes take days to complete a provisioning ticket), then your private cloud needs to support both architecture types.

In Asia, some of the largest private cloud deployments take just this approach, with notable cases being telecommunications carriers. They have critical systems, such as service provisioning, billing, CRM, and ERP, that run on enterprise-style architectures. They also need to deploy new SaaS services to their customers that must be built on web-scale architectures. A private cloud that efficiently handles both lets telcos avoid having to take a hybrid approach.

A hybrid approach, where customer data from enterprise architecture applications leaves the telco’s data center to be used in a web-scale application running on a public cloud, could run afoul of various regulatory and technical hurdles in places like India and China (with its “great firewall”). There is also the issue of paying for the bandwidth to connect to the public cloud providers. A private cloud solution that handles both types of architectures solves all of these problems.

At least one prominent cloud vendor finds this capability to be a critical selling point in Asia. Jason Jackson, Director of Cloud Services, APAC, for Citrix, said to me: “CloudPlatform’s ability to efficiently accommodate both enterprise and cloud-era workloads inside of a data center has been instrumental for many of our clients’ success.”

Another factor driving these private cloud deployments in Asia is that the public cloud advantage of eliminating capital expenses in favor of pay-as-you-go operating expenses works great for shoestring startups, but is not a factor for large public companies and may even be a disadvantage if they have financial incentives to spend cash on tangible assets. This is especially true of nominally public companies that have some government ownership.