The industry’s largest enterprise communications show, Enterprise Connect, was held earlier this month in Orlando, Fla. For decades, the show has followed the trend of voice and video shifting away from its own proprietary network to IP networks and, most recently, to the cloud.
The challenge for network managers is how to ensure the quality of network-based applications, particularly latency and jitter-sensitive ones—such as voice and video--remains high. Poor-performing apps can kill productivity because it makes it difficult to get any work.
On March 18, InfoVista announced some updates to its SD-WAN product, Ipanema, that ensure application quality remains high. SD-WANs are still relatively new but certainly on the rise. ZK Research data finds that less than 20 percent of companies have production SD-WAN deployments, but well over half of companies are now looking at the technology.
The value proposition most vendors take to market addresses a couple of big pain points: cost and application availability. All SD-WAN solutions use features such as multi-path to improve network uptime, and the use of broadband can lower costs. However, this doesn’t necessarily mean that applications will perform better. Availability and performance are both important, but they are two different things.
SD-WAN Enables Different Apps to Map to Various Networks
With a basic SD-WAN, different apps can be mapped to the various networks being used to theoretically improve performance. For example, voice and video can be directed over MPLS and best-effort traffic sent over a broadband connection. In this case, if the MPLS network gets congested but doesn’t go down, nothing will change, and users will suffer because video will be blocky and voice packets will drop.
In the event of a failure, traffic will be moved to the alternate connection; hence availability remains high, but performance suffers. InfoVista uses a number of machine-learning techniques to understand application delivery requirements and maintain a high quality of experience. Each flow and user session is analyzed to ensure business policies are enforced.
Most SD-WAN vendors will look at a network metrics such as packet loss, jitter, delay and roundtrip time, but it can be difficult to extrapolate application experience from network information. InfoVista combines the network data with a number of application metrics, including throughput, session information, server response time, transaction time and packet data, and then calculates an application quality score on a range of 0-10.
The process of automating and scoring application is nearly completely automated. Internal IT may have a good idea what on-premises application are running but likely have no idea what cloud services are crossing the network because of the rise of shadow IT.
Infovista's 'Massive Cloud Database'
As part of this release, InfoVista has built a massive cloud data base that can identify every cloud service being used. This data is then used to create the policies to prioritize applications, orchestrate services and ensure quality of experience. The company uses a number of complementary mechanisms, such as dynamic bandwidth allocation, class-based queuing, advanced compression, acceleration, prioritization and dynamic path selection and recalculates network capacity and traffic demands every second.
The cloud orchestrator is a great example of how on-premises infrastructure can work in conjunction with a cloud service to deliver a high-quality network experience.
It would be impossible for most companies to build this kind of capability in-house because it requires massive compute power and constant scanning of cloud apps to keep the database up to date in real time. I find there’s a religious debate regarding on-prem infrastructure and cloud-managed. Neither is better per se; the key is to use each to its strength.
As part of this announcement, InfoVista also updated its portal customers use to manage the network. The tool has always displayed QoE metrics for users, but now it has the ability to drill down into applications for troubleshooting purposes.
It also has historical data from retrospective analysis on a per-conversation basis. The new troubleshooting tools can reduce the time taken to find problems or at least understand it’s not a network problem. ZK Research data found that 90% of the time taken to solve a problem is identifying where it is and the drilldown capabilities can shorten that time dramatically.
Networks Connect Users to Services
In the digital era, businesses are network-centric, because it’s the thing that connects users to devices and to applications and the cloud. Networks can no longer exist in isolation from applications. The updates from InfoVista are a signal that the SD-WAN market is changing.
Historically, SD-WANs were about being cheaper than legacy networks with a bit of network resiliency thrown in. The new features tightly couple the network to applications to make them perform better, which is what business leaders care about.
For many organizations, the network is the business, and that requires a fundamental re-think of the WAN to be application centric instead of transport centric.
Zeus Kerravala is the founder and principal analyst with ZK Research. He spent 10 years at Yankee Group and prior to that held a number of corporate IT positions.