Balancing performance, integration and ease-of-use in virtualized environments
It’s clear that the move towards cloud computing is under way. Service providers and their vendors are rapidly developing products and services, as well as building out the infrastructure for various cloud offerings. Customers, large and small, are actively evaluating the relevance of cloud computing to their business, if not already participating and enjoying the benefits. How will network integrated services such as WAN optimization fit into the cloud model? The initial indications are that WAN optimization will play an important role as companies migrate their applications and services to the cloud — users will continue to be separated from compute and storage resources by bandwidth/latency constrained WANs. But what new challenges will integrating WAN optimization into cloud bring? I’ve outlined some thoughts below based on my experience with deploying WAN optimization in numerous environments and looking at what options are available in current cloud computing models.
- Platform Virtualization – Providing WAN optimization capabilities as a virtual machine image seems like an easy (and obvious) change versus the traditional model of delivering these services on dedicated hardware. While this has some obvious benefits on deployment simplicity, the impact on service scalability and performance shouldn’t be overlooked. For example, the leading Infrastructure as a Service (IaaS) providers offer generic x86 based hardware for virtual machine images to run on. This excludes the use of any custom hardware for acceleration of specific WAN optimization functions, such as SSL offload, etc. It also brings into question constraints around disk i/o performance, depending on the storage model(s) available from the cloud provider.
- Network Integration – The network integration challenges in a cloud environment look similar to the challenges faced in traditional network environments, although potentially more complex. Given the different layers of abstraction is cloud environments, it’s much less likely that you will have knowledge about, or access to, the underlying network infrastructure design details. Given this, how do you handle such issues as network interception of traffic? Perhaps directed mode is the preferred integration option versus transparent? Or perhaps using host-based WAN optimization solutions is the way go? Or perhaps some hybrid and other alternative that requires more service/feature development with the cloud provider itself? This area will need some work and potential collaboration with the cloud providers.
- Ease of Use – There are obvious of ease of use benefits, really the same that exist for virtualizing other resources in the cloud – economies of scale due to virtualization, simplistic provisioning and mobility. Even if the performance challenges discussed above negate the one-to-many compute virtualization benefits, the simplistic deployment and mobility benefits are there. One of first opportunities I can see for WAN optimization in the cloud is virtualizing the management platform instances. These systems only require IP connectivity to the WAN optimization devices handling user traffic, and generally have lower hardware performance requirements.
I’m interested in your opinions and feedback. What benefits and challenges do you foresee to virtualizing WAN optimization services in the cloud?
