Shedding light on broadband performance for an optimized SD-WAN

A decade ago, if I had asked a network manager to consider using broadband to connect branch offices, he would have looked at me as if I had a second head. Times and opinions change, though, and broadband is now as viable as MPLS or any other network service.

The catalyst for this has been the rise of software-defined WANs (SD-WANs). Their multipath capabilities and network optimization make it possible to use broadband for business connectivity, and organizations of all sizes have been jumping aboard this trend. One of the challenges that remain, though, is that not all connections or types of broadband are created equal.

Not all broadband is created equal

A network manager at a business with branches located nationwide recently told me that his company’s preferred broadband type has been cable, but he has been surprised by the variability in the resulting bandwidth by location and time of day. For example, in one large metro area, he is purchasing 50MB of bandwidth, more than adequate for most branches. During the mornings he often sees throughput of 100MB or more. Later in the day, however, when kids get home from school and the Xboxes and Netflix subscriptions start coming online, he has seen his bandwidth drop to as low as 8MB. 

He’s not alone in this experience. Broadband variability is a big problem for companies looking to shift to SD-WAN, and network professionals might wish they had a “broadband scorecard” to help them decide which type of broadband to use. 

Comparing apples to apples 

One company that can shed some light on broadband performance is ThousandEyes. (Disclosure: ThousandEyes is a client of ZK Research.) ThousandEyes differentiates itself from other network performance companies by providing visibility and performance information of the internet. In actuality, there is no single internet, but rather a collection of interconnected networks, and ThousandEyes has agents deployed in a number of different nodes to let businesses that rely on the internet understand how those various networks are performing. This can help alleviate congestion and improve the performance of business applications.

The company recently expanded its network intelligence coverage to include broadband ISPs in the six major metro areas: Washington, D.C. (Ashburn, Va.), Chicago, Dallas, Los Angeles, San Jose and Seattle. ISPs covered are AT&T, CenturyLink, Charter, Comcast, Cox and Verizon, which together account for most of the broadband options available. As part of the launch, ThousandEyes has also increased the IPv6-enabled coverage by over 150%. The company now has about 250 of its cloud agents located across 152 cities globally, giving its customers exceptional visibility into how their cloud and internet-facing apps are performing.

Visibility can deliver a higher ROI to SD-WAN 

For network managers, ThousandEyes provides critical information to migrate to an SD-WAN, since it lets them know what they normally wouldn’t know until it’s too late. For example, if a company is planning to put broadband into a branch location, the broadband performance information can help understand the performance prior to implementation and make an educated go/no-go decision. Without it, the network team could make the decision to shift to an SD-WAN for cost-savings purposes only to have the help desk get crushed with calls because the performance is bad, effectively wiping out the return on investment.

Also, the data from ThousandEyes can help plan for contingencies. One of the great things about SD-WANs is that traffic can be redirected to alternate paths, almost on demand. If the visibility tool shows that the performance of a certain broadband connection has congestion issues at a certain time of day, the network team could plan to move critical apps over to a backup link for that time period.

The last benefit is that network operations teams can use this data to continually test and validate the network and be proactive with network evolution. The higher-fidelity information enables trending to be done so network professionals can better predict when problems are likely to occur. Action can be taken to increase bandwidth, switch providers or change the architecture before users start calling.

One of the interesting data points from my research is that almost three-quarters of application performance issues are reported by the end user putting IT in continual firefighting mode. ThousandEyes can help reverse this.