WAN Optimization is an umbrella IT term that has no precise definition. It provokes an impression, but each vendor and expert has their own nuanced view of the term. I am going to define it as the use of hardware, software and process to achieve the cost-effective delivery of data throughout an enterprise. For data delivery to be cost-effective, an enterprise must understand the criticality and value of the data, and end-user performance expectations. This avoids the issues caused by a per-bit cost assessment.
"An effective WAN Optimization strategy needs to incorporate awareness of network conditions and application priorities into its management function"
Our company’s WAN was designed to provide enterprise-level, converged voice and data services. It was deployed in the late 1990’s when T1s were the most common high-speed WAN access method. Enterprise clients demanded consistent and predictable voice quality. To meet this demand, we implemented early WAN optimization technologies such as QOS, Mean Opinion Score monitoring and MPLS traffic engineering in our core.
In 2016, clients now have an array of high-bandwidth connectivity options. However, the migration to the cloud, globalization, mobile work force, BYOD, video and the Internet of Things are driving a projected fivefold annual growth rate in bandwidth demand. A WAN optimization strategy is critical to meeting the hunger for bandwidth without drastically increasing costs and network complexity for your company.
The biggest issue we have in developing a WAN optimization strategy is that many clients do not understand their data traffic patterns, or the impact of delay, jitter and packet loss on application performance. As a best practice before embarking on a WAN Optimization Strategy, the following baseline information must be gathered:
1. Existing sources of bandwidth consumption
2. Priority and criticality of each source of bandwidth consumption
3. How changes in your enterprise and application environment will impact bandwidth demand
4. End-user expectations for application performance
5. Application performance dependencies on network quality
6. Accurate network topology
7. Performance of the current network in relation to packet loss, jitter and delay
The primary function of WAN Optimization is to overcome issues caused by delay, packet loss and jitter. Network delays can have a drastic impact on the actual throughput of TCP-based applications. The actual throughput of a gigabit circuit can be reduced by over 50 percent based upon TCP window size and delay. We once had a client with a network to China that had a 350ms round trip delay and was only able to use a tenth of their bandwidth for TCP applications. Packet loss and jitter exacerbate these issues and cause performance issues on their own.
WAN Optimization in a traditional private WAN using MPLS and other SLA backed circuits is very straight forward. Standard practice is to implement QOS and traffic shaping for instances of network delay over 30ms in conjunction with a WAN optimization appliance. With end-to-end control of the circuit you are able to implement TCP compression, data redundancy elimination TCP flow control and application acceleration to improve performance. The information from network data analysis can be used to configure and implement a WAN Optimization solution on a private network. Their private WAN is backed up by Dynamic Multipoint Virtual Private Network (DMVPN) tunnels traversing the open internet.
The movement to the cloud and the demand for disaster recovery has complicated WAN Optimization. End users expect the same response times, predictability and reliability they received from locally-based applications. The use of the open internet to access key applications such as WebEx, Saleforce.com or Office 365, and the flooding of private WAN links, has made this reliability and predictability more difficult to achieve. The open internet is especially challenging because it is constantly changing and an enterprise loses control over packet behavior as soon as it leaves their local router.
WAN Optimization needs to include new functionality to deal with new networks. An effective WAN Optimization strategy needs to incorporate awareness of network conditions and application priorities into its management function. Here is an example of how this works:
A client uses our cloud-based hosted voice system to collocate several critical servers in our data center. They have several key public cloud-based applications that are used to generate revenue and require consistent performance. The client has three sites. At each site they implemented a 10mbs MPLS connection to our cloud and a bulk 1 gigabit internet circuit. The MPLS circuit has internet access through the data center. They wanted to use the MPLS for voice, both circuits for their server data and the bulk circuit for accessing their cloud-based applications. They wanted to back the voice up over the bulk circuit in case of an MPLS outage, and they wanted to back up the internet circuit over the MPLS. However, only specific cloud applications would be backed up and the failover would be initiated when packet loss, jitter and delay increased so that the end-user response deteriorated below an acceptable level. This avoided flooding the backup circuit with low-priority traffic. The MPLS and internet circuits were provisioned from diverse carriers.
The benefit to the client was that they were able to use low-cost internet bandwidth and reduce their expensive SLA backed bandwidth. It also mitigated the problems associated with relying on a single internet carrier for critical cloud applications. They were able to supplement their MPLS bandwidth with bulk internet bandwidth and reroute critical applications based on internet network conditions. The end result lowered costs while creating a more reliable and predictable end-user experience.
This would not have been possible if a WAN Optimization strategy had not been developed based on the information gathered in step one through seven above. The development of an effective WAN optimization strategy requires a detailed knowledge of your existing network, and plans for future expansion. This knowledge must be implemented into an effective optimization platform that is network and application aware. With constant managing and monitoring, an enterprise will be able to implement the most cost-effective method of delivering data.