2010 Prediction: Dave Demlow, Double-Take Software

By Dave Demlow
published: Thursday, December 10 2009

2010: Looking Into the Clouds

In 2010 we will see increasing interest and activity around cloud provided infrastructure as a service (IaaS) providing on-demand, pay as you go access to scalable compute and storage infrastructure. Clearly IaaS will not be the only area where services will be provided by the cloud, but it’s getting so much attention because it is a model that can be easily understood by most people, can evolve from and work with existing enterprise applications and architectures and can be quickly and easily trialed in low risk areas such as test or development and expanded incrementally. In addition, there are already a surprising number of providers to choose from with similar but differentiated service offerings and pricing models that can meet a variety of market needs.

I say IaaS is easily understood because when it comes right down to it, it’s all about virtual machines which by now are very well understood. Package up your operating system and application stack into a nice portable container, and run that container on a shared infrastructure but isolated from other workloads sharing the same infrastructure. The next logical question is whether to own the all the infrastructure to run those containers, which depends on the nature and longevity of the workloads, the utilization of existing infrastructure, geographic constraints (sometimes), compliance, legal issues, management preference, and of course, cost. I expect for the foreseeable future that most companies will want to have a mix of internal and external infrastructure to meet the changing needs of various workloads.

As a software developer, and particularly as the developer of software commonly used to replicate, failover and recover these “workload containers” (whether they are already virtualized or not) from one geographic location to another, we have been very early cloud watchers and users ourselves.

In addition to the test and development benefits that most software companies can receive by utilizing cloud infrastructure resources to run virtual machines in the cloud at least for peak demand times or special ad-hoc projects, being able to test things like cross country or trans-Atlantic data replication using virtual machines distributed across multiple clouds, purchased just for the time we are using them is fantastic. Of course, it also makes you start to think about many other possibilities.

Within a datacenter or corporate WAN, there are tools like ours available that can continuously backup and failover workloads on physical or virtual servers into virtual machines (P2V and V2V failover) in different locations for disaster recovery and high availability. If IaaS can allow me to run virtual machines and standard operating systems and applications, can I use it as a recovery site and actually failover workloads from my datacenter to the cloud?  Sure, with the right networking and technologies, it’s certainly possible (and available). Coincidentally this mirrors the same thought processes and customer adoption that we saw early on when server virtualization itself was first limited to test environments, then moved to disaster recovery which gave people the confidence to start moving production workloads into their virtualized infrastructure.

As enterprises begin to consider any IaaS provider, there are several key things they should consider. The first set is around selecting the cloud provider that best fits the needs of their organization and application. Some aspects are obvious such as the provider’s cost structure, service level agreements (SLAs) and the specific nuances of their offering. Very often the details vary significantly from one IaaS vendor to another in areas like Operating System Support (do they support the versions of Windows and Linux you intend to use), VM provisioning options (can you create a VM with any size boot drives, upload your own VM’s or use only their existing templates), and networking options (i.e.: VPN access, load balancers, public and private IP addresses). Some vendors can be very flexible and responsive to custom requirements, while other offerings are based on achieving maximum economies of scale through standardized cookie cutter configurations. Although we often think of the cloud as being “out there somewhere,” for some workloads, sometimes location itself is important whether it is for legal, compliance, audit reasons or network latency / throughput issues. And yes, some workloads might also justify a thick pipe directly to a regional cloud datacenter to get reasonable bandwidth and low latency at a reasonable cost.

Given these differences and the changing dynamics of this new market, I think it is especially important for cloud users to avoid “vendor lock-in” wherever possible and consider their exit plan – whether it is how they could move their cloud workloads back on premise, move workloads between cloud providers or for the most critical workloads even use multiple cloud providers to provide redundancy for a given workload. This is where technologies that can protect, move, and recover these full workload containers across long distances and where different virtual, physical and cloud platforms can play a critical role.