With the rising popularity of SD-WAN, there is a growing debate that WAN optimization is becoming obsolete. SD-WAN is gaining acceptance and for good reason. It creates an intelligent overlay of multiple transports on your WAN to efficiently and automatically route traffic over the most optimal path. In the process, customers often experience a substantial boost to overall usable capacity which relaxes the need for aggressive WAN optimization techniques. Although the benefits of SD-WAN are many, it does not completely cover all of the benefits provided by WAN optimization.
The Pros and Cons of WAN Optimization
The rise of WAN optimization began around 2004 and addressed three primary networking issues determining the end user experience when accessing data from across the WAN: bandwidth, latency, and packet loss.
Bandwidth limitations are addressed by minimizing the amount of data passed across the network. Typically this is done through compression and deduplication algorithms. To ensure applications don’t “hog” the capacity of a connection, WAN optimization appliances will also prioritize application traffic. This way applications that immediate access to the wire, such as voice call, for example, are guaranteed access even during heavy usage.
As the distance between end-users and their data grows, bandwidth gives away to latency and packet loss as the primary determinants of session capacity. Network delay, or latency, defines how long packets take to travel from one designated point to another. Latency is often measured to the destination and back what’s called the “round trip time” (RTT). Caching techniques and protocol-specific optimizations minimize latency by reducing the number of application-layer exchanges that are necessary across the network.
Packet loss occurs when network congestion or problems in the physical infrastructure cause packets to be lost during transmission. It’s expressed as a percentage of packets. As a rule of thumb, Internet connections frequently experience 1 percent packet loss. Packet loss will be addressed by some WAN optimization appliances using forward error correction (FEC) that allows receiving stations to automatically regenerate lost packets without requiring transmission.
(For a deeper explanation of the impact latency and loss have on capacity, and the benefits of Cato Network optimization download our ebook “MPLS vs Internet vs SD-WAN Cloud Network”.)
Packet retransmission is a major source of latency for TCP sessions. TCP was designed to adapt to the underlying network conditions. At a high-level, TCP operation sends a group of packets (a “window”) before waiting for an acknowledgment of receipt from the destination. TCP gradually increases its window size to a maximum size. Packet loss causes the window sizes to be reduced, repeating the process all over. As a result, packet loss or significant delays in receiving acknowledgments will impair TCP performance.
SD-WAN and the “3 Cs”
Fast forward to the present to SD-WAN. Whereas WAN optimization improves the performance of an individual connection, SD-WAN improves the overall network. More specifically, SD-WAN addresses issues impairing MPLS networks:
- Cost: SD-WAN uses lower cost Internet circuits instead of expensive MPLS circuits allowing organizations much more flexibility in how they design their networks.
- Capacity and Performance: With the Internet, SD-WAN provides far more capacity at locations. By load-balancing connections, businesses can also easily to aggregate multiple low-cost Internet connections for even more capacity. SD-WAN routes traffic across the optimum connection based on application requirements and real-time loss and latency conditions.
- Cloud access: MPLS is optimized for branch-to-datacenter traffic and not for branch-to-cloud. WAN optimization, a dual-sided technology, is limited in its ability to optimize cloud access where installing WAN optimization appliances is complex (for cloud datacenters) or technically not possible (for cloud applications).
SD-WAN on its own brings cost and performance benefits, but SD-WAN alone doesn’t solve all the inherent issues with WAN traffic. As Alec Pinkham notes, “SD-WAN has no ability to affect traffic once it leaves the endpoint location. Once the traffic is on the WAN, it will follow the rules of the WAN as defined by the providers. SD-WAN technology puts the packets on the currently best-performing WAN (or combination of WANs), but it does nothing to actually make those WANs work better.”
SD-WAN and WAN Optimization Working Together
To take advantage of the benefits of both SD-WAN and WAN optimization, look for solutions such as Cato Networks Secure, Global SD-WAN as a Service. This solution provides a global, SLA-backed backbone with built-in network security, delivered as a cloud service. Cato’s multi-segment optimization optimizes WAN and cloud traffic in three segments: the last mile from the source location, the middle mile that connects all locations and the cloud, and the last mile to destination.
Read about SD-WAN vs. MPLS