The age of public cloud computing is now: 97% of companies already make use of at least one cloud service in the form of Software as a Service (SaaS). The ubiquity of SaaS brings with it a correspondingly increased dependence on the Internet for access to (often) critical enterprise applications. And, as with the centralization of applications into data centers, shifting to SaaS likewise increases reliance on the enterprise WAN. This is because half of all companies backhaul all Internet traffic across the WAN, and 31% backhaul at least some of it. That means, for many or all branches, Internet traffic reaches the branch office via the WAN. The perceived cloud computing performance of SaaS applications is ultimately a composite of the apps' performance plus the Internet's plus the WAN's.
Improving Internet performance
In order to make the SaaS experience better for end users, network managers and engineers can enhance the performance across the Internet and the WAN.
On the Internet connection, "one-armed" or asymmetrical acceleration technology, like application delivery optimization (ADO) appliances and packet shapers, can improve performance. ADO devices can prioritize conversations between desktops inside the company network and known or sanctioned SaaS providers, possibly guaranteeing them a minimum amount of bandwidth, and can limit the speeds or rates of competing network resources to make sure there is room for SaaS services. Some ADO devices can even prioritize specific streams within the conversations (i.e., financial system transactions over accounting queries, queries over report printing, etc.). Caching of commonly used Web pages and other content can also improve performance.
Improving WAN performance
Looking to the WAN, ADO appliances can not only prioritize but actively optimize and accelerate the traffic as well. They can compress content to reduce transmission latencies, for example, and spoof protocols to make applications appear more responsive to end users. They also can alter traffic at both the HTTP (since it's mostly Web traffic) and TCP levels to improve WAN performance by multiplexing requests or connections to turn a myriad of round-trips into one. For example, if the elements used to draw the homepage of a SaaS application normally require 10 HTTP GET requests to retrieve, the page can't be drawn completely until all 10 have finished their round-trips of request and response. By multiplexing the requests, the ADO device can, on behalf of the client device, send a single request, get a single response from the server and feed the 10 separate responses back to the client at LAN -- rather than WAN -- speeds.
Because the WAN level works below the TCP level, Quality of Service (QoS) tagging can be used to prioritize SaaS application packet delivery across the WAN; it can also be used to prioritize lossless delivery, which is perhaps better than prioritization because the delays caused by retransmission can be very damaging to perceived performance.
Improving cloud computing performance
Another option for improving SaaS performance is to take some of the Internet out of the loop by establishing a private line connection to a provider's point of presence on the Internet, removing some uncontrolled and uncontrollable hops from the equation. Enterprise IT can do this with or without a service provider's approval. For example, if they use AT&T's big Massachusetts point of presence, they can get AT&T to connect into the same facility so that enterprise traffic goes only over the network inside that building. Likewise, one can remove the WAN from the equation by introducing Internet connectivity directly to the branch and restrict traffic to and from sanctioned providers to improve security and protect performance.
More on WAN and cloud computing performance
Integrate your WAN with Internet cloud services
Cloud computing services you should consider for your WAN
How your WAN intersects with cloud services
Cloud computing tutorial for WAN managers
How vendors have improved SaaS performance
Of course, with SaaS performance mostly out of the hands of IT, it's hard to be able to speak sensibly about it without some specialized application performance management (APM) tools to provide insight into real behaviors inside the corporate WAN or LAN network.
Even so, without a chance to see inside the provider's infrastructure up to its Internet links, APM tools are at a distinct disadvantage in trying to provide cloud performance data. Some SaaS providers are trying to fill the gap themselves, as it is in their best interest to be able to say definitively when a cloud performance problem is not in their infrastructure. They are deploying client-based tools (from vendors such as ThousandEyes) that users download as a browser extension and use to map performance across all the hops between themselves and the provider's infrastructure.
The bottom line is that even in the age of SaaS, IT has ways to help cement good application performance for critical tools. However IT decides to improve public cloud computing performance -- whether down low with QoS tagging or up high with protocol optimization or from the side by changing the way users get to SaaS services -- it can take an active role in ensuring a smooth shift to this new enterprise technology paradigm.