Today, many companies are going through a digital transformation with a cloud-centric approach to enterprise application delivery. As core business applications move to the cloud, like Microsoft Office 365, the Internet becomes the business-critical network. Much of the traffic that was routed to an internal data center now goes to the cloud through the public Internet. SaaS acceleration is critical to a successful cloud app strategy.
The public Internet is a low-cost option, but public Internet performance is simply unreliable. The Internet consists of thousands of Autonomous Systems which pass traffic between each other based on least cost routing. The protocols they use using have no mechanisms to identify or avoid congested links. There is no such thing as an SLA for public Internet performance. End-to-end SaaS acceleration across the public Internet doesn’t exist. Companies are nervous about the “best effort but no promises” public Internet. Many are looking for a different solution. MPLS is a popular option for providing good site-to-site connectivity. However, it’s not a solution that accelerates delivery of cloud based apps to the user from the SaaS provider.
Options To Improve Internet Performance For SaaS
That leaves the following options:
- Keep sending traffic through the public Internet, without performance SLAs (and cross your fingers).
- Adopt an SD-WAN-based approach. Some SD-WAN providers have a service that optimizes traffic between their private regional PoPs. This service uses VPNs from the customer to the PoP.
- Get direct access from the public cloud provider that hosts the applications you use
- Use a private cloud on-ramp service that directly connects to multiple public cloud providers
- Use an Internet overlay network provider for SaaS acceleration
Each of these options has strengths and weaknesses. At the end of the day, what matters is the tangible performance improvements they bring. But how should that be measured? What does “good performance” mean? And how should customers judge potential vendors?
We will answer those questions by demonstrating the differences between a typical Internet connection and a high-performance network connection.
Reliable SaaS Acceleration Over Time
When you evaluate network solutions, remember the “tyranny of averages”. Wide variances in performance over time can be masked by a single decent looking number. A high-performance network should provide better average Internet performance. There should be very little difference between the best and worst speed of file transfers. The difference should be milliseconds and not seconds, every hour of every day. The smaller the variance, the better the guarantee that Internet traffic peak hours won’t affect the end user’s performance.
Below is an example of load time for a 5MB file. The load time is measured from Amsterdam to locations in North America and APAC. The grey line shows the performance of a broadband connection across the public Internet. This varies between 9-35 seconds. The green line shows the performance of that broadband connection routing across a high-performance overlay network. This shows consistent load times of ~1 second, with very little variance.
In this test, we use load time as our metric rather than time to first byte or full transaction time. Load time best isolates the network’s part in the transaction.
Reliable SaaS Acceleration Over All Geographies
Global performance is the second crucial metric. When you evaluate options, make sure you establish that application performance is consistent across all locations.
In general, SaaS applications that are closer to the user perform better. The laws of physics and complex longer routes are the enemies of Internet performance. They make it much harder to assure that more distant users get good performance levels.
Here, you can see a big difference in Internet performance between routing across the public Internet and routing to an overlay network. Across the public Internet, users in many regions take an average of over 15 seconds to load a 5MB file. In contrast, that time averages about a second with a high performance overlay network. Also note that the load times for the overlay network are very consistent across regions. Additionally, Mumbai is the only real outlier.
SaaS Acceleration Availability
The third and perhaps most important factor in Internet performance is network availability. This is the ability of a network to successfully transfer data from point A to point B at any given moment.
In the example below, the average availability of the broadband connection is 97.97%. In contrast, the overall availability of the high-performance overlay network is over 99.90%. The overlay network shows an increase in availability of 165 hours per year.
What Leads To These Internet Performance Differences?
Bandwidth describes how much data can travel through a ‘pipe’. In contrast, throughput is how much data actually travels through the ‘pipe’. To see this difference in action, run a speed test across your home broadband connection. You will find that upload speed is much lower than your advertised bandwidth. Having a 200MB (bandwidth) broadband connection is in no way means Internet performance throughput of 200MB.
Actual throughput may and often will vary wildly based on several factors. Among these are path congestion, peering, and the protocols in use.
TCP behavior under changing conditions-
TCP was made for reliable, ordered, and error-checked delivery of data between two systems. It is a very conservative protocol that aims for accurate delivery. It does this at the expense of high throughput.
When it sees congestion, it shrinks packet sizes and chops throughput to make sure packets can make it to their destination. This results in application performance slowing to a crawl. It also means angry and frustrated users.
Peering and transit-routing decisions-
These decisions are made by Border Gateway Protocol, the main routing protocol of the Internet. They are based on business relationships between Internet service providers. Route performance is not a factor. Traffic follows the lowest cost routes and those routes don’t change in response to congestion.
SaaS Acceleration Numbers Don’t Lie
The performance improvements that are possible with an overlay network like Teridion’s cloud based overlay are huge. When you’re selecting a network solution, use these numbers as performance guidelines. Above all, expect your vendors to match or exceed them. Whether your customers are internal or external, performance matters in assuring a good user experience.
These impressive numbers are an output of the continuous comparison testing that we do at Teridion. We continuously chart the performance of our Internet overlay network. Catchpoint, our monitoring partner, performs regular synthetic tests across Teridion’s network. It compares those with tests across the public Internet and charts the results. We do this for each of our customers, and we give them total visibility to the results. If your SaaS application is struggling with Internet performance issues leading to poor user satisfaction, try us out. It’s easy to get started, and the performance gains start right away.