Guest post byTodd Davis from TeloIP (originally posted on TeloIP blog on on 11/10/14 3:32 PM
To truly optimize WAN Performance we need to take a holistic approach
Optimize:an act, process or methodology of making something as fully functional or effective as possible
Business expectations for network performance have evolved rapidly in the past decade, and so have the tools available to help optimize your WAN. Despite this changing landscape the basic challenge remains the same for WAN Managers… how can you get best possible WAN Performance for your budget?
The traditional approach to “WAN Optimization” has narrowly focused on getting more throughput and better efficiency from a limited amount of bandwidth capacity. This has grown into a multi-billion dollar industry that’s nearing maturity.
It’s time to move beyond MPLS and take a fresh, holistic approach to WAN Optimization that brings network quality and application performance to the next level while reducing costs.
To achieve optimal network performance at the lowest possible price, you’ll need to deploy the right WAN solutions in the right combination at each WAN location. Part 1 of this article provides a quick roundup of available WAN solutions to explore the challenges they’re meant to resolve and their limitations. In Part 2 we’ll describe how Software Controlled Networking effectively brings it all together using Network Virtualization and a Cloud based management platform.
Table 1: Available WAN Solutions & Their Limitations
Private MPLS WAN is the traditional choice for WAN Connectivity. The private connection offers security and predictable bandwidth performance. MPLS functionality enables QoS traffic prioritization for real-time applications that need it. The problems here are cost and speed, with private MPLS links having a much higher bandwidth cost per Megabit compared to Public Broadband. This makes it challenging to meet bandwidth and budget requirements using Private MPLS links at all your WAN sites.
T1 Link Bonding was one of the earliest methods to alleviate WAN Bottlenecks by combining bandwidth from multiple circuits into one higher speed connection. This is a two ended solution, with one end at the central office switch where the bonded links terminate. The limitations here are that each link must be of the same type (T1) and from the same provider, and running at the same speed. Unfortunately, Bonded T1 solutions just aren’t fast enough these days for the cloud, and bonding any type of private links is a costly way to scale bandwidth.
WAN Monitoring tools collect data from network endpoints and routers, typically using SNMP, Netflow, or IPSLA type methods. The data is compiled into charts and reports that help IT Managers find problems, allocate resources, and plan future WAN capacity requirements. Having remote visibility into your network is key. However, it only helps you find problems so manual intervention is still required to resolve every issue. For cost efficient WAN Management you need more than monitoring, you need automated system responses and remote management tools that can quickly fix problems without truck rolls. Another limitation is that different networks and different carriers use different monitoring tools, making it difficult to get consistent performance metrics across a large multi-site and multi-carrier network.
IPSec VPN creates an encrypted connection between sites and provides WAN security over Public Broadband. This solution creates much needed alternatives for low cost high speed WAN connections. IPSec VPN is limited to providing security for the transport so it doesn’t address reliability and doesn’t provide QoS for application performance, and the on-site VPN support costs are driving many businesses toward Cloud Managed VPN solutions.
WAN Acceleration represents the cornerstone of traditional WAN Optimization; it uses a two ended approach that requires a network appliance at each WAN site. File compression is combined with advanced file management techniques to minimize bandwidth consumption between sites. WAN Acceleration helps where budget constraints or circuit availability prevent you from simply buying a faster connection. The limitation here is that you can only squeeze so much traffic through a given pipe, while some IP traffic can’t be manipulated for acceleration. Squeezing more traffic through a congested pipe tends to increase latency and may create packet loss. Bandwidth saturation can really hurt VoIP performance and increase lag time for Cloud based applications.
WAN Link Balancing, or load balancing provides another way to increase bandwidth by adding more connections. It’s a two ended network appliance solution that enables you to connect different types of links from different ISPs and steer site-to-site traffic across them based on predetermined rules. Low cost high speed Public Broadband links can be used to bring down bandwidth costs, and failover features help increase WAN Reliability. The main drawback of Link Balancing is that traffic only travels across one link or another, so your bandwidth is underutilized. File transfer speeds never burst above the speed of the single link that’s being used. Using preset routing rules means there’s no mechanism that reacts to changing network conditions. The failover here relies on DNS due to the IP address changes, and means that these users and applications may still get disconnected.
WAN Aggregation is another two ended solution that uses multiple WAN Links to increase bandwidth capacity. Aggregation goes beyond Link Balancing by physically distributing the packets over different links, combining bandwidth from all links to work together like one single larger pipe. This provides better bandwidth utilization and will typically include a fast WAN failover feature. The most advanced WAN Aggregation techniques include software that senses real time line conditions and reacts to avoid latency, jitter, and packet loss issues. This enables a virtualized form of QoS that improves application performance on Public Broadband and creates much needed MPLS Alternatives that can go faster and cost a lot less. There’s definitely a lot to like here. However, there are some key limitations which include distance and scale problems. Distributing the packets across different links is great for increasing WAN speed on the last mile, yet performance degrades when using WAN Aggregation for Long Haul traffic. It’s simply more difficult to keep the packets in sync when they travel great distances along different pathways. Long haul performance issues, plus some vendor limitations of hardware scalability have undermined the potential of WAN Aggregation for large multi-site and multi-carrier deployments.
Conclusion: Over the years all of these WAN Management tools have helped with efforts to optimize network performance, yet they each suffer from substantial limitations. To make matters worse, these techniques are often deployed in a piecemeal approach. This can lead to a tall stack of costly network appliances with substantial management and deployment complexities, and a significant opex cost for maintaining them all.