SD-WAN Featured Article

Enhancing Application Performance Using Intelligent Internet Routing

November 16, 2018

An attainable but often unrealized goal for SaaS (News - Alert) providers is sub-second application response time. This is an important goal because studies consistently show slow application response times will cause an interruption in users’ flow of thought, and further delay will result in customer churn and users abandoning the task all together. For example, Google reported that for 70 percent of the pages it analyzed, it took nearly seven seconds for the visual content above the fold to display on the screen, and it took more than 10 seconds to fully load all visual content above and below the fold. Its study found that as page load time goes from one second to seven seconds, the probability of a mobile site visitor leaving increased 113 percent.

Unfortunately, today’s inconsistent Internet performance makes response times of more than five seconds typical, particularly for end users who are located far from an application’s data center. New approaches are available, however, that enable SaaS providers to take control of Internet performance and improve delivery of their applications, resulting in better performance and ROI.

The Internet Backbone Problem

Nearly all web content flows through the Internet backbone. The Internet backbone isn’t under the control of a single provider; it’s composed of hundreds of providers that interconnect with each other. These networks charge each other to transport data, or enter into peering arrangements where they agree to pass a like amount of traffic to each other. Within the Internet backbone, all traffic is treated equally. Providers maximize profitability by minimizing the cost of sending traffic.

Least cost routing is the process of selecting a path for traffic along the Internet backbone based on the lowest cost, not on best performance. Least cost routing is enforced by the rules of the Border Gateway (News - Alert) Protocol (BGP), the routing protocol of the Internet backbone. BGP routes data from network providers’ own networks to other providers’ networks, and vice versa. When users visit a website, that data has commonly traversed many networks belonging to many different organizations. To ensure that data transmissions eventually get to their intended locations, Internet backbone routers keep a table of known and trusted routes. The BGP rules enable providers to prioritize traffic using cost-based weighting factors, putting customer experience for SaaS providers’ applications at the mercy of the network providers’ cost-cutting routing tables.

The fact that BGP’s rules dictate that traffic between two points will always take the same path regardless of network congestion further negatively impacts performance for SaaS providers. The combination of least cost routing and lack of congestion detection is the root of many performance problems with the Internet backbone, along with poor infrastructure, bad peering arrangements, and other inefficiencies.

Most SaaS applications use the Transmission Control Protocol (TCP) as their transport protocol. TCP provides the Internet with reliable, ordered and error-checked delivery of data between two systems. By design, TCP is optimized for accurate delivery rather than timely delivery. Thus, TCP sessions can incur long delays while waiting for out-of-order messages or retransmissions of lost messages. For this reason, real-time applications such as VoIP opt to use different protocols.

For SaaS providers this means that TCP data transfer algorithms are not efficient, particularly for larger files which require many round trips to transmit. TCP/IP requires each chunk of data to be acknowledged by the receiver before the sender sends the next batch of data. Since these data chunks are typically small and made up of thousands of bytes, transferring even 1 MB of data can require hundreds of separate trips through the Internet backbone.

A New Approach for Intelligent Network Routing

New advances in Internet overlay networks are emerging as effective solutions to address the data transport delays of the Internet backbone. These solutions use sensors and monitoring agents deployed into the Internet backbone for intelligent network routing to steer traffic around congestion and outages, and deliver better performance while reducing the cost and complexity of application, storage and networking infrastructure workarounds.

While many overlay networks are essentially very large private networks, some advanced Internet overlay networks leverage the infrastructure of major public cloud providers, and take advantage of the scale of public cloud investments in both location and peering relationships with local ISPs. Through this multicloud approach, they orchestrate optimal routes for each region and user connection. Advanced solutions take advantage of public cloud peering relationships with local ISPs to provide a fast “on-ramp” to the Internet backbone for end users anywhere in the world. Leveraging the public cloud, they offer a number of cloud best practices, including:

  • Multicloud: Deploying virtual overlay networks across many cloud providers and hundreds of cloud data centers.
  • Elastic scaling:  Auto scaling up and down based on real time demand for each region.
  • Resilient: Detection and dynamic recovery from a number of failure scenarios, including DNS, network and data center outages.

Advanced Internet overlay networks can act as self-healing networks that are resilient to a wide range of potential outages, including problems affecting entire data centers or regions, to ensure needed network performance for SaaS application delivery. They avoid problems including:

  • DNS provider failure – they leverage multiple Domain Name Server (DNS) providers. If one provider fails or slows significantly, advanced solutions will switch to a secondary DNS provider.
  • Network failure – they automatically reroute traffic in response to any network congestion, including data center or network failures.

Advanced solutions enable network administrators to easily view and analyze traffic flowing through the Internet overlay network. While the Internet backbone has traditionally been a “black box” to application providers, overlay networks can deliver dashboards showing performance and traffic states and analysis that provide in depth comparisons of traffic. Advanced dashboards can be customizable and include programmable widgets that display upload and download traffic by time period, region and by end-user domain. For each widget the dashboard can also show the performance improvement between the Internet overlay network and the standard Internet. Advanced analysis comparatively measures performance, latency and throughput, and by making the “middle mile” visible can effectively map end to end performance from user to app.

Advances in Internet overlay networks give SaaS providers a new approach for intelligent routing to overcome the inefficiencies of the legacy Internet backbone. These new technologies ensure end users experience SaaS app performance and dynamic web content in the manner they were designed. They also ensure that SaaS providers can eliminate a significant source of user frustration, keeping customers happy, and keeping them subscribed.

Pej will be joining the panel discussion, "The Promise of Application Performance," at the upcoming SD-WAN Expo. Join Pej and other industry experts as they illuminate the audience with what this maturing technology can do. Join us!

Edited by Maurice Nagle