In-Depth
Application Acceleration in the Cloud Age
In the age of the cloud, applications and content are only as good as the network they ride on. We explore three application acceleration options.
By Amar Khan, Vice President of IP Services, Internap
The cloud will bring convenience to enterprise application users with the ability to access applications from anywhere at any time -- at least that is the goal. However, during a recent, extended trip throughout Europe and the Middle East, I was attempting to use my cloud-based CRM application and was reminded of how network challenges (such as latency) can quickly dampen this ideal. Although I was able to “access” the application, the performance was so poor that I abandoned my attempts within five minutes. That’s the spotty reality of the “anywhere anytime” cloud as it stands today.
My experience is not unique to a single application or single location -- in fact, it’s an all-too-common experience that business users and consumers all over the world face every day. This is a result of more applications moving from on-premise to the cloud. Although the cloud can yield powerful benefits, these are meaningless if we sacrifice reliable connectivity and performance. The challenge is that as applications and content move further away from the end user to centralized server farms, we are inevitably seeing greater latency and packet loss. In this new scenario, cloud applications are also falling victim to core Internet communications protocol TCP (Transmission Control Protocol) and its conservative approach to IP traffic delivery.
Because of its design (nearly 40 years old), TCP sacrifices performance to deliver reliability. It throttles back the delivery rate of IP traffic as the distance between the content or application source and the end user increases. Hence, the performance of my CRM application while I was traveling was directly affected because TCP slowed the communication between the server and me to a crawl.
As more enterprises rely on remotely-housed content and applications, the limitations of TCP are being widely felt. TCP works by slowly increasing the window size of data transmission -- think of this as the size of payload in each transaction. It conservatively increases this window size and then aggressively reduces it upon encountering high packet loss or latency variation. This results in an inconsistent delivery rate as TCP slowly speeds up and then slams on the brakes before speeding up again. Yet, TCP remains an underlying protocol of Internet traffic delivery today, so the question becomes: Is there anything we can do about this?
Fortunately, the answer is yes. Enterprises are taking steps to address TCP’s limitations today by employing WAN optimization and application acceleration. Although these are established technologies, they are being fast-tracked as priority IT investments because of their ability to dramatically improve Web performance. For example, at Internap we’re seeing our Accelerated IP (XIP) customers significantly reduce latency and improve application performance between geographically remote sites. Cloud platform customer Jolokia Networks increased performance by 50 percent for North American users and by 80 percent for European users, with those in more remote areas, such as South Africa and South America, enjoying even greater increases of up to 140 percent.
Let’s look at some of the application acceleration options available today:
Hardware-based acceleration solutions are proven to increase the speed of delivery two-fold or more in many cases. These solutions are typically designed for point-to-point architectures. This approach is advantageous when an enterprise has centralized an application at a specific hub and users are accessing it from known branch sites. This type of solution will also accelerate delivery when enterprises are communicating over the Internet between fixed branch locations. This method is not ideal for organizations with large amounts of mobile users or those that commonly access data from beyond branch sites, such as employees who travel nationally or globally; or, most important, those with customers accessing applications that reside in geographically remote locations.
Another approach that is gaining momentum is Acceleration-as-a-Service (AaaS), where no hardware is required and the acceleration is provided by the carrier or ISP. This allows the enterprise to set up their Internet service at the hub location where their application or content is hosted with a service provider that offers integrated TCP acceleration. In this case, hardware costs are eliminated and application and content delivery are accelerated, regardless of their geographical location. These service-based solutions are capable of speeding delivery of applications in the range of 1.5x to 4x, which can markedly improve the end-user experience.
A third solution, which will likely become the next frontier of acceleration technology, accelerates cloud application and content delivery in both directions. Organizations and their customers can not only receive information faster, but they can also deliver or upload it much more quickly.
A Final Word
In the age of the cloud, applications and content are only as good as the network they ride on. Acceleration technologies are likely to play a key role in the cloud landscape in the future, ensuring that users can quickly and reliably access critical information whenever needed -- not just from down the hall but from anywhere in the world.
Amar Khan is vice president of IP services at Internap where he is responsible for driving the value proposition and strategy for the company’s IP networking services, including Performance IP and Accelerated IP (XIP) . You can contact the author at akhan@internap.com