I Have Been Disillusioned by WAN Acceleration

My Expectations Were So High

DISCLAIMER: If you work for a WAN Acceleration vendor, please do not e-mail saying "our technology is different" or "we can make it work for you". Except for some different tweaks, in general, it's the same. Please don't bother.

When it came to WAN Acceleration, I drank the Kool-Aid. I was excited about the technology, it made sense as a way to fix typical problems on wide area networks, and there is good competition in the marketplace driving innovation. Thus, I pushed for the last several years to buy it. But, the soft ROI of "it will make us more productive" wasn't going to fly in the last two years given the economy. About 6 months ago though, because of business changes and data center consolidations, I found a way to produce a hard-dollar ROI. Essentially, we would take out our separate lab WAN and use the cost savings to buy WAN acceleration globally. It was a great business plan. I actually used what I learned in MBA Finance class to build the ROI model. Then we sold it to IT management and other business units. We were set to go. Next, we did a rather extensive pilot. We deployed devices from different WAN acceleration vendors throughout the network and started testing. And the results were....well.....nice, but not revolutionary. Yes, some web pages were snappier and Sharepoint was a bit faster, but nothing exciting. CIFS was amazingly faster, but honestly, our users are technology savvy and they just stopped using CIFS over the WAN years ago. Thus, it was a fix without a problem. More importantly, our smaller field sites just didn't see any benefit. These sites have two 3 or 6 Mbps circuits shared between 50-100 users. That's not a lot of bandwidth. Well then, let's drop some WAN acceleration in there and all will be great! Not. First, WAN acceleration really makes a difference at small bandwidth sites when its cache can serve up most user traffic. These are WARM transfers. If most data transfers are WARM, the users get a lot of benefit and the WAN is hardly used. In this case, you can keep the bandwidth low and invest in WAN acceleration. Unfortunately, given the dynamics of our field offices, a lot of traffic wasn't ending up WARM. Our field offices are not cookie-cutter implementations like bank branches or McDonald's Restaurants. Each office is really its own little business unit serving different customers and doing different work (including between the people inside each office). So, when the cache doesn't have the data, it's a COLD transfer. In this case, the WAN acceleration also provides some benefits: TCP acceleration, compression, QoS, error correction, etc. The problem is, with COLD transfers, the data has to still flow over the small WAN links. So, before WAN acceleration, a single user on these 6 Mbps circuits could max a TCP session at say 4 Mbps because of WAN delay. But, add WAN acceleration in and this 4 Mbps shoots right to 6 Mbps. At this point, the circuit can't transfer any more. But, the user doesn't really notice the difference because it's not that much more bandwidth AND now we have induced a WAN bottleneck. QoS kicks in and starts dropping packets which backs off TCP transfers of this user and the other 50 users at the site. Thus, the users don't notice the difference. And, because we hastened the bottleneck on the small WAN circuits, some users get angry. Despite all this negativity, I do feel there is a good role for WAN acceleration (at least in our network). That is at sites that do have plenty of WAN bandwidth. Currently, without WAN acceleration, because of WAN delay, users at sites with big bandwidth links can't utilize them. For example, a single TCP session from Amsterdam to California can max out at around 5.4 Mbps. But, we have nearly 90 Mbps of WAN bandwidth in Amsterdam into our MPLS cloud. So, we have the bandwidth, but not the ability to utilize it. But now bring in WAN Acceleration. Then we have the capability (the big bandwidth) to exploit all the benefits of WAN Acceleration (WARM and COLD transfers). If it's WARM, great. But, if it's COLD and still needs to be transferred over the WAN, we have the bandwidth to do it. Then you get all the other benefits of WAN acceleration, particularly TCP acceleration which can enhance application response time. Instead of 5.4 Mbps per TCP session, WAN Acceleration takes that COLD TCP session to maybe 20 Mbps. That's a big difference that users will notice. So, now all I need to do is to get our WAN providers to give us gobs more bandwidth and still find a couple million dollars for WAN acceleration gear. And for my next trick......

More >From the Field blog entries:

WAN Carriers Need Better Bandwidth Pricing Models for Enterprises

I Got an iPad

Is Enterprise IT Here to Provide Great Internet Access?

Latency in the Data Center Matters

Net Neutrality for Dummies

Cisco Nexus 5000's Poor Results in Data Center Switching Test

  Go to Cisco Subnet for more Cisco news, blogs, discussion forums, security alerts, book giveaways, and more.

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Related:

Copyright © 2010 IDG Communications, Inc.

SD-WAN buyers guide: Key questions to ask vendors (and yourself)