Nov 16, 2016
Over the last 18 months one of the hottest topics in IT has been software defined WANs (SD-WANs). One of the advantages of SD-WANs is that they enable companies to leverage relatively inexpensive Internet bandwidth. There is no doubt that having additional bandwidth eliminates some performance problems. However, network organizations that are analyzing SD-WAN solutions need to answer a critical question: Does having additional Internet bandwidth negate the need for WAN optimization functionality?
The best way to answer the preceding question is to look at some key use cases.
Business Continuity (BC) and Disaster Recovery (DR) are closely related concepts that describe how a company prepares itself to continue functioning after a disaster such as a hurricane or a tornado. BC refers to the processes and procedures that an organization puts in place to ensure that at least a critical set of business functions will continue to function both during and after a disaster. This includes things such as ensuring that employees know where they are to report to work in case they are not able to access their normal work site.
Whereas BC is primarily the responsibility of the company’s business unit managers, DR is primarily the responsibility of the company’s IT organization. DR involves making sure that the IT infrastructure is in place to ensure that after a disaster that the company’s employees can still access the applications that support the company’s critical business functions. A key component of a DR plan is to have a mirrored site ready to support those business critical applications. This means that the data these applications need must be copied from the primary data center that houses those applications to the mirrored data center. This involves transmitting large volumes of information. In addition, since the data centers are typically far apart to ensure that a single disaster doesn’t take both sites out of production, this data has to be transmitted over long distances. It is tempting to use relatively inexpensive Internet bandwidth to support moving this data. However, because of how TCP’s congestion control works, transmitting large volumes of data over a lengthy WAN link can result in a phenomena called saw-toothing. Saw-toothing refers to the fact that a graph of the actual throughput would look like a saw as the data rates move up and down as the TCP widow size is dynamically adjusted. Saw-toothing results in the inability to fully utilize WAN bandwidth and hence can result in not being able to transmit all of the data needed to support the company’s DR plan. For many companies a better solution is to implement WAN optimization functionality referred to as de-duplication. De-duplication keeps the primary data center and the mirrored data center in synch while only sending a minimum amount of data over the WAN link.
There is no doubt that in some instances adding bandwidth is beneficial. One time when it isn’t is when the network has to support a chatty protocol such as the Common Internet File System (CIFS). CIFS is commonly used to enable users to share files and similar to other chatty protocols, it requires hundreds of round trips to complete a transaction. For the sake of example, assume that a transaction requires 200 round trips. If that transaction takes place over a WAN with 50 ms. round trip delay, the chatty nature of the transaction adds 10 seconds of delay which would cause users to complain bitterly. Since adding bandwidth doesn’t reduce the round trip delay nor the required number of round trips, adding bandwidth won’t reduce the amount of time that the transaction takes. WAN optimization functionality referred to as spoofing includes a number of techniques that overcome the impact of chatty protocols and hence this functionality does reduce the overall transaction time.
Supporting BC/DR and chatty protocols are just two reasons why WAN optimization functionality is still sometimes necessary even when relatively low cost Internet bandwidth is available. Another reason is the growing use of LTE as the primary WAN link into branch offices. Given that LTE-based services are billed on a usage sensitive basis, implementing compression functionality to minimize how much data transits the LTE link can be very beneficial.
While the use of WAN optimization functionality is not going away, the way in which that functionality is consumed is changing. Part of that change is that implementing WAN optimization functionality is becoming notably less complex in part because solutions are becoming more integrated. Another important change is how the functionality is priced. Whereas WAN optimization controllers used to be a capital expense, increasingly WAN optimization functionality will be acquired on a subscription basis where the amount that users pay will be based on factors such as the amount of bandwidth being optimized and the amount of time that the optimization functionality is used. While network organizations seldom implemented WAN optimization ubiquitously, over time these organizations have become and will continue to become even more focused on when and where it makes sense to implement WAN optimization functionality. For example, as network organizations adopt SD-WAN functionality they will increasingly assign applications to specific virtual network that transit the WAN. Based on the characteristics of those applications, network organizations will implement WAN optimization functionality on the appropriate virtual networks. Another change is that the use of WAN optimization functionality is expanding to include optimizing the use of network resources other than just the WAN links. For example, WAN optimization functionality that has the capability of breaking out and identifying traffic types can be used to determine which traffic doesn’t need to go through a firewall. Having this functionality extends the life of the firewall.