Why cloud still needs WAN optimization

BANGALORE, INDIA: The idea of cloud services has caught the attention of innovators, engineers, marketers, and pundits everywhere.

And it’s with good reason that it has — the potential for cloud services to cut the costs of IT infrastructure for businesses is tremendous. In fact, in a recent Merrill Lynch research report, the bank asserts that the migration to the cloud is a “multi-year tectonic and disruptive shift” that could create a $100 billion market by 2013.

However, like most technologies that are early in the hype cycle, there are plenty of truths, myths, and big problems to be sorted through and overcome. Cloud services is dealing with precisely these challenges right now.

What is the cloud? Is it new?

Determining a universally acceptable definition for “the cloud” is in itself a daunting task. Even casual discussions with technologists and industry analysts reveal dozens of different definitions, angles, and perspectives.

However, there is one thing that is clear: the cloud aims to shift IT resources and applications into a more scalable, cost-effective computing model that hides much of the complexity that enterprises deal with today.

In many aspects this transition is not new. For years enterprises have been centralizing their IT resources, in an attempt to make a private “cloud” that delivers greater scale and simplicity. Users are satisfied so long as they can complete their usual tasks with the same or greater productivity as they did in the past.

The cloud in today’s incarnation is new in that the goal is to push infrastructure into a shared services model, and potentially push it further out of the enterprise to the application provider.

New applications could be written specifically for the cloud; legacy applications could migrate to the cloud for more cost-effective management. In the ideal world of cloud services providers, the enterprise does not manage any data centres; only the cloud provider has data centres and those resources serve many enterprises.

In order for this vision to come true, cloud providers will have to overcome many challenges. Frequently mentioned are concerns about the regulation and security of sensitive data. Similarly, many have identified as a key issue the need for a business model that is both profitable for the provider and cheaper for the customer. In addition, but less frequently discussed, is the issue of application performance.

How does the cloud eliminate performance bottlenecks?

For many users, researches, and even engineers of cloud services, there is a misunderstanding that cloud eliminates performance bottlenecks. The belief is that if you use a cloud service, performance will be fine. In short, the cloud in itself does not eliminate performance bottlenecks.

{#PageBreak#}

We have seen this error before: enterprises who wanted to centralize thought that connectivity of their “network clouds” had improved so much that they could centralize without regard to performance — and they were proved wrong.

Applications running in cloud services are, just like other applications, subject to two basic limitations:

Capacity: A network — any network — has limitations in terms of how much data it can carry at any point in time. Whether the data is coming from an internal source or an external source, it will contend with all the other data that users are requesting.

Just because you have a cloud service, it doesn’t mean you will automatically get to put 10 pounds of data in a five pound bag.

Latency and Chattiness: As the distance between a user and his data increases, time to access that data increases significantly due to the combined effects of latency and application protocols inefficiencies, or chattiness.

While cloud providers have re-written some applications to use HTTP efficiently, tests shows that these applications are still suffering from the same issues that legacy enterprise applications face. And by de-emphasizing even what some enterprises currently have done to have services close to users, cloud services actually could make performance worse for enterprise users.

In a model where data simply lives “in the cloud” it is likely that data will be moved further away from the user in pursuit of more cost-effective data centers in the cloud.

As cloud services progress, there will be a natural evolution towards richer, more interactive applications. Today it is e-mail, CRM, and basic document creation; tomorrow it will be collaborative design, full-scale document management, and multi-layered production management applications.

As a result, we may see cloud services drive more data across the WAN across an ever-increasing set of distributed users. And as the cloud becomes more familiar and more flexible for enterprises, more organizations may push their applications out to the cloud to achieve cost benefits despite the fact that performance problems could increase in severity.

Combining cloud services with WAN optimization

As a new space with an age-old problem, it makes sense for cloud providers to look at other application architectures to see how they overcame the application performance problem in order to achieve more cost-effective deployment. That’s not to say the exact same solution will be applicable, but certainly can provide guiding experience.

Enterprises have adopted WAN optimization infrastructure as a key method of accelerating applications over the WAN and enabling the centralization of infrastructure to one or a few data centers. WAN optimization gives enterprises three key benefits:

{#PageBreak#}

Virtual capacity increases: WAN optimization makes a T1 feel like 3 or 5 T1’s by de-duplicating redundant traffic bytes across all key TCP-based applications.

Performance improvements: Applications perform 5, 10, or 15x better by eliminating the inefficiencies in application protocols, giving users the speed that they need to work effectively. These improvements apply to classic rich client applications like email, but also to virtual desktops and web-based applications.

Flexible design: By using a WAN optimization platform, enterprises no longer depend on their application or cloud provider to be a networking expert. In addition, the cloud provider doesn’t need to intimately understand the operations of all the legacy applications that may be pushed out of the enterprise.

Instead the cloud provider and enterprise have an additional, easy-to-manage layer of technology that helps to fix the performance issues of whatever applications the enterprise uses — whether they are rich client, thin client, or cloud-based.

Can WAN optimization apply to the cloud? Recent experience shows that, because WAN optimization has a history of accelerating Web-based (HTTP and HTTPS dependent) applications, they can effectively help in today’s cloud services.

One example of this is Bechtel, who is reorganizing their internal IT efforts into a cloud services model. When they started with a clean slate to build the best possible internal cloud for their needs, they included Riverbed Steelhead appliances.

This approach can also work with clouds that are external to an enterprise. For example, a major global payments, network, and travel company is currently using a system integrator to combine WAN optimization with external cloud services to provide the performance that they need.

Over time, as the cloud evolves, expect to see WAN optimization products develop an additional set of features that allow them to better support applications written for the cloud, as well as legacy applications that are migrated to the cloud.

In the future, customers will simply pick the architecture that works best for their organization — internal IT, internal cloud, cloud provider — and rely on their WAN optimization infrastructure to ensure the performance that their business requires.

The author is regional director, India/SAARC at Riverbed Technology.