… where response time and speed are concerned, many businesses automatically assume Google.com- and Amazon.com-levels of performance from services such as Google App Engine and Amazon EC2, but this can be a mistake.

-- ESJ, “Q&A: Managing Performance of Cloud-Based Applications and Services”

A big mistake, indeed. While the underlying systems may be optimized and faster than fast, that doesn’t mean that applications won’t suffer poor performance. There are many other factors that determine how an application will perform, and most of them are variable. They can change from day to day, hour to hour, and from user to user. It certainly won’t hurt to optimize the network and, for providers at least, it’s about the only thing they can do to help customers whose applications may be suffering from performance problems.

Network optimization is a very different game from application and application delivery optimization. The former focuses on well, the network and doesn’t take into consideration anything about the application like protocols and connections and repetitive data delivery. The ability to accelerate an application (because that’s really what we’re talking about) is not something that can be achieved solely by optimizing the network. Optimizing the network does not impact the application’s ability to process requests and return responses, nor does it improve the capacity of the application, nor does it reduce the chattiness of an application protocol. It doesn’t do anything for the application because it’s is, appropriately, focused on the network.

Application performance issues can be caused by poorly performing networks, yes. But a provider only has access over one leg of that network – the one they control. Poor performance caused by networks outside the organizational boundary of control cannot be addressed by localized network optimization.

Always-on connectivity is now a life-style and it certainly doesn’t appear to be slowing down. Users connect to applications at all hours of the day and night and from disparate places from day to day. They use a wide variety of devices that connect over multiple networks. And while 80% of all successful cloud computing providers may soon implement network optimization as predicted by Gartner, they can only optimize what they control – and what they control is their network, not the networks over which users connect.

An application deployed in the cloud may perform perfectly fine over a corporate LAN connection, and perhaps even over the corporate wireless network. But once a user moves to their own wireless network, or a cellular network, or the Internet cafe down the street’s network, all guarantees regarding performance are off.

What’s necessary to solve the problems of application performance in the cloud is not simply faster networks but smarter networks. Smarter networks that are capable of adapting in real-time to changing conditions that impact application performance; conditions that cannot be addressed by network optimizations that are constrained by organizational control boundaries.

Our focus thus far regarding the new network has been on operational concerns: the ability of infrastructure components to integrate and collaborate with each other in new ways to ensure rapid adaptation to changing operational events within the data center. But the other side of the new network is the ability to handle with alacrity the dynamism inherent in the users; in the users desire and ability to move seamlessly from desktop to laptop to mobile device, from corporate LAN to remote office to Internet cafe. Each of these endpoints and networks has different performance characteristics and capabilities and thus may require specific network and application delivery network policies to address specific problem areas that hinder acceptable application performance.

Complicating the matter is the fact that different user + network combinations can have dramatically different impacts on the application servers. Devices with lower bandwidth capabilities consume application resources for longer periods of time because it takes them longer to send and receive data. More mobile users thus increases the number of concurrent users a single application instance can support given desired performance characteristics which may result in the need to scale sooner than if all users were accessing the application from a desktop over a fat LAN connection.

Without the ability to adapt automatically, on-demand, to the many permutations possible for users, devices, networks, and applications there will continue to be problems with application performance – both in the cloud and out. This is not a problem peculiar to cloud and, in fact, is a problem for just about any organization with applications that may be accessed in a variety of ways and over multiple types of networks. It is not reasonable to codify policies that cover all the possible scenarios because the landscape is constantly changing. The decision regarding which policy to apply is a highly complex formula with many variables that need to be calculated every time a request is made. And that’s what application delivery infrastructure does: the dynamic application of delivery and security policies based on real-time conditions in the internal network, the external network, the client, the server, and in between.

Application delivery may have once been a “nice to have” or only required for large-scale sites, but those days are over. Application delivery is rapidly becoming a must have for more and more organizations to efficiently manage the myriad variables that impact application performance.

At some point in the future (soon one would hope) network optimization will become table-stakes for cloud computing providers.  Smart shoppers will require that performance points such as “intra-network latency”, “average network utilization”, and “average time to provision” be disclosed in order for customers to balance performance against price and with other mitigating acquisition factors. Network optimization is a good start, but it’s just that – only a start.

Before we can sufficiently address application performance (which is really application delivery performance) we will need a more mature cloud that provides not only network optimization but application delivery optimization services. Because it’s the latter that can assist with application performance issues on both sides of the equation.