Arthur Cole spoke with Dr. Cahit Akin, co-founder/CEO, Mushroom Networks.
When it comes to moving data across wide-area networks, there are two basic options: reduce the amount of data or provision more bandwidth. When applications or data do not lend themselves to compression or caching, however, only the latter option will suffice, although that usually ends up costing more. However, companies like Mushroom Networks are working with new WAN aggregation techniques like Broadband Bonding that don’t provide wider pathways, but more of them. The company’s Cahit Akin explains.
Cole: WAN optimization has long been considered the cure-all for remote office communications and cloud computing. But don't some types of traffic benefit from optimization more than others if, say, they are more amenable to advanced compression and cache techniques?
Akin: Wan optimization works exceptionally well in cases where the bits transferred over the wire can be cached and/or compressed. If that is not the case, minimizing the number of bits that are pumped through the IP pipe may not have much benefit. Encrypted and video traffic are good examples. In those cases, there is no other alternative than to add more bandwidth in order to move the traffic through quicker.
Cole: How do techniques like Broadband Bonding address this? Won't simply compiling more bandwidth on wide area networks simply drive up costs?
Akin: If compression is cramping more passengers into vehicles, then Broadband Bonding is adding more lanes to your highway — both are valuable for transmitting people between locations more effectively. In the case of Broadband Bonding, usually the IP tunnel that is created by combining several smaller IP connections has advantages compared to a single IP connection of equivalent specs. Usually, the option of such a single IP connection does not exist, but in cases where it does, it is very likely that it is cost prohibitive compared to creating an IP connection via Broadband Bonding. In addition, Broadband Bonding can diversify the connectivity over multiple ISPs. This creates “high 9s” availability across multiple ISPs that you can’t get with a single service provider.
Another advantage of having multiple smaller IP is the natural variations in the IP fabric will create diversity in the former setup. At any given time, the chances of having better latency metrics on one of the IP lines is more likely than a single IP line, so intelligent traffic grooming can enable low latency transmission, which may not be possible with a single IP connection. In other words, if you have a single lane highway and there is an accident, you can’t get around it, versus a multiple lane highway and the means to be able to change lanes in an agile manner. So overall, building an IP connection out of smaller WAN links would have performance, reliability and cost advantages compared to a single IP connection with equivalent specs.
Cole: Is there a danger that WAN availability could suffer as Big Data and mobile communications become more ubiquitous? Even with the most advanced optimization techniques, isn't there still a limited amount of bandwidth to go around?
Akin: Demand traditionally has always surpassed available capacity in terms of WAN resources. Innovation never stops and new applications with high network demands are always popping up — Big Data, real-time applications to name a few. This is a vicious cycle where the industry will keep innovating for efficiency and more bandwidth, and end users will demand more the next day. Even though temporary WAN availability issues are feasible, technology has always come to the rescue and I believe that the never-ending back-and-forth race is what pushes technology forward.