Jan 12, 2012
Stick around this industry long enough and you’re bound to hear the “same old-same old” posited as if it’s truly novel. Such is the case with this recent post by Lori MacVittie on the “epic failure” of standalone WAN optimization. It advises that to really improve application performance one needs to fix the applications themselves using an application delivery solution – the application equivalent of an “Extreme Makeover.”
Fixing the network using WAN optimization, it asserts, has little or no impact anymore.
The problem with this analysis is that it 1) is based on an incomplete understanding of WAN optimization, 2) only addresses a subset of enterprise applications, and 3) takes a naïve view of IT operations.
On the first point, all too often organizations believe that bandwidth is the sole root of all evil when it comes to poor application performance. Sometimes this is true; often times it is not. Very often the latency from communicating over long distances and packet loss resulting from operating over shared WANs like the Internet or MPLS networks conspire together to undermine applications throughput. For example, an application running across a 150 Mbps link, coast-to-coast link (50 ms latency and 0.5 percent packet loss) ends up with a peak throughput of just over 3 Mbps. (Check it out yourself here or see an independent source here.)
A truly robust WAN optimization solution fixes bandwidth, latency, and loss issues, making it much more intrinsic to application delivery than a bandwidth-only scenario described in the “epic failure” post. In other words, there is much more to WAN optimization than just “squishing data,” as the writer suggests.
On the second point, while many applications can actually be sped up using application acceleration, others cannot. You can’t accelerate VoIP, video conferencing or virtual desktops, for example, as it’s impossible for someone to speak or type faster. What you CAN do is improve the quality of these applications by fixing latency and loss issues that cause phone calls to be dropped, video calls to be pixilated, and VDI to suffer from slow screen refreshes. These are all network, not application problems, and therefore require a WAN optimization solution.
Similarly, when doing data replication or data migration between data centers, the performance bottleneck is often not at the application level. Data leaves the source at the maximum available rate, but poor network conditions impact data throughput. THIS is why Recovery Point Objectives (RPOs) are missed, not because the application isn’t running fast enough. So, once again, this problem must be solved by fixing the network, not the application.
On the third point, I refer to the classic 80/20 rule. Fixing WAN performance is relatively easy (20 percent of the effort), yet yields significant performance improvements. Yes, maybe 20 percent more gain can come by fixing individual applications, but is this realistic when the average enterprise has over 50 mission critical applications going over the WAN? And how many versions of the same application are out there?
Focusing on fixing the application, as the author writes, “with its unique transport and application layer behaviors,” is asking IT to individually address the dozens if not hundreds of enterprise applications traversing the WAN. That’s just plain unrealistic. Just ask IT leaders at Google, Microsoft, Expedia, Toshiba, ARM, and hundreds more companies who have all decided the best route is to fix the network using WAN optimization.
Solving application performance problem with some vendors’ “holistic application delivery systems” is a lot like buying new clothes, equipment, and books to lose weight when in reality walking around the block for 30 minutes a day and reducing fat would probably solve 80 percent of our problems.
WAN optimization is like that simple routine and diet. It won’t solve all of your applications problems, but it will solve many of them and do so across all the applications in your WAN at one time. And while joining a health club might make a dent in your wallet and then force you to wait weeks to see the results of your fitness goals, not so with WAN optimization. Now you can find out if WAN optimization will solve your application delivery problems in only a few hours and at little to no cost to you.
So what are you waiting for?