How Cloud Networking Can Drive Application Performance
Not very long ago, I could have started this blog post by declaring that “enterprises and service providers are moving to cloud”. But now, the transition to cloud that has been foretold, prognosticated about, and tracked feverishly by the industry is… well, it’s done. A recent study found that 96 percent of enterprises had adopted public cloud computing, and not in a half-hearted way either. On average, respondents had deployed applications in about three different clouds and were experimenting in two more, and public cloud is carrying 40% of their compute workload.
Naturally, this vast workload represents a large and dynamically expanding market ($411B by 2020 is a lot of money!). Established public cloud providers continue to expand their infrastructure and reach, and the market is flooding with newer entrants who are also gaining traction, with IBM Softlayer and Oracle notably making big gains.
Many of the most common applications found on user’s desktops have moved to a cloud or hybrid cloud delivery model and a subscription-based consumption model, with Microsoft Office (and Exchange) and Adobe Creative Cloud being two obvious examples. And while the transition of the old guard is a good indicator of the state of the market, even more important is that when the next killer enterprise app appears, you can be assured that it will be SaaS-delivered.
Mobile users now have the expectation that they will have access to- and the capability to interact with- all their data no matter where they are and no matter what device they have in hand. Along with that comes the presumption that application performance, speed, and response time will be consistent with “desktop” levels.
Today, in practice, the Internet is expected to do the same kind of work and deliver the same kind of application performance that used to be the exclusive province of the LAN. But- is the Internet ready to be the new LAN? And if the answer is no, then is there some jiu-jitsu that can use the weight of the massive infrastructure of public cloud to create a faster, smarter Internet?
First, let’s look a little deeper at the three main constituents: enterprises, public cloud providers, and users.
Enterprises Value Their Time To Market Over Managing Infrastructure
A significant reason for the wholesale transition to cloud is the importance enterprises have on time to market and time to value. Enterprises would rather focus on the development and delivery of their services than worry about setting up and managing monolithic infrastructure. Cost and depreciation of infrastructure technology is a huge reason to not go through purchasing and refreshing on-premise infrastructure every few years, not to mention the tedium and pain of continual patching, resolution of version conflicts, troubleshooting and regular software upgrades. Another big reason for this move is the agility cloud networking platforms provide, with virtually infinite scalability and a pay-as-you-use consumption model.
It’s easy to limit our vision of cloud elasticity to the compute and storage resources consumed by the enterprise. But if we expand that view to include the huge network capacity that is represented by the aggregation of public cloud providers, it begins to look like an underutilized resource that could drive a better Internet if it could be harnessed in the service of cloud networking.
Public Cloud Providers And Economy Of Scale
Since Amazon Web Services became publicly available in 2006, it has had 62 price cuts (as of the latter part of 2017). Google, Amazon, and Microsoft all drop price points regularly in a hyper-competitive market. A large weapon in this ongoing price war is the price-performance-value relationship expressed by Moore’s Law. The computing power per dollar has increased by an order of magnitude about every four years for the last quarter-century. The price of 1GB of storage was $7.70 in 2000. Today its 2 cents, and that in itself represents over a 50% drop from just two years ago. In addition to the continual drop in infrastructure pricing, the economies of scale are vast- public cloud infrastructure is a reflection of intentional cloud optimization to meet customer performance needs.
User Productivity Through The Cloud
The excellent usability and performance of consumer devices and consumer-targeted apps has steadily increased the pressure on enterprise applications to keep pace with user expectations. If you’re a SaaS provider or an enterprise, keeping up with these expectations isn’t just catering to the whims of finicky users. It has a direct, measurable impact on their engagement and productivity.
Ilya Grigorik, in his excellent book High Performance Browser Networking, points to studies that consistently report the same human reactions to interaction delay across all application types and devices. A response time of at least 200ms will be reported by a user as “lag”. Once the 300ms response time threshold is reached, it feels “sluggish”. At the 1 second threshold the user has checked out mentally and context switched, and at 10 seconds they have given up on the task altogether.
It’s to the clear benefit of both the SaaS provider and their enterprise customer to deliver great cloud performance all the time every time to a user, no matter where they are or what device they’re using. The enterprise keeps their users busy, and the productivity gains and absence of user complaints about performance keep the enterprise happy with with the SaaS provider.
Driving Application Performance In The “Everything Cloud” Age
So what does all this mean? In short, we are going to see ever increasing traffic across the public Internet. This will be not only static content like loading a web page. The quantity of dynamic, bidirectional traffic will continue to skyrocket as well, and it will be accompanied by, if anything, more demanding user expectations. Ask any user of cloud based office suites like Office 365 or Google G-Suite about the frustrations of poor internet performance. Nothing drags productivity to crawl like not being able to do basic tasks like updating a spreadsheet, slide deck, or document, or needing to send large files fast and waiting for what seems an eternity for the upload to complete. The fact is that when we access services in the public cloud, rarely do we go through a single provider. Going from the east to west coast of the United States, you may go through 30 to-40 different hops on multiple underlying ISPs (Internet Service Providers). Anywhere along that path, congestion and packet loss can lead to cloud apps slowing to a crawl or freezing altogether.
While there is a tremendous amount of focus on building infrastructure around compute, storage and service delivery platforms, not enough attention is being paid to cloud networking- the paths that connect those platforms to users. Some public cloud vendors are investing in dedicated links (like AWS Direct Connect or Google Fiber), but it’s only within the bounds of that particular single cloud provider. Both of these are static, dedicated paths to their own consumers. Enterprises, of course, can and do deploy MPLS and/or SD-WAN, which both deliver performance benefits within the organization itself. But private networks, by their very nature, are constrained by hardware limitations, purchased bandwidth limitations, or both, and are of no help to an employee traveling in China who needs to quickly access the corporate Box account. That’s where cloud networking has to come into play.
In the age of multi-cloud enterprise deployments and global users, both monolithic offerings from individual public cloud providers and private, dedicated networks like MPLS just don’t meet the needs of a global, mobile community of user and workers.
Cloud Networking And Internet Overlay Networks
Think of an overlay network simply as a network running on top of another network. It requires the underlying network to be able to function, and is typically intended to provide some security or performance advantages. MPLS, SD-WANs using IPSec VPNs, and Virtual Leased Lines are all examples of Internet overlay networks. All of these examples are widely used by SaaS providers and enterprises looking to deliver better service to their users, and they all come with inherent issues of scalability, lack of control over the Internet backbone, or both.
This is where the massive, global infrastructures of the major public cloud providers can be put to use in service of overcoming these obstacles through cloud networking. First, use the globally distributed compute power of the cloud to implant measurement nodes throughout the major cloud providers that continually monitor the performance of the Internet backbone. Then use that data to feed an orchestrator that makes real-time routing decisions so that traffic between the user and the server always travels the best-performing path, even across multiple public cloud providers. This orchestrator could also take advantage of the cloud’s ability to instantly harness resources when needed by automatically spinning up or moving routing nodes wherever needed to extract the best performance.
This is exactly what we do at Teridion, unlocking the power of public cloud networking. We use the entire surface area of the public cloud to create a secure Internet network overlay for each of our customers, using machine learning to set up and modify routes predictively, avoiding congestion on the Internet. Teridion delivers three key benefits for our customers:
- Low latency Internet application performance – Teridion’s Internet Overlay routes around Internet congestion
- Increased Internet reliability – Teridion’s Internet Overlay has stable, efficient routing to ensure packets get to and from where they need to go with little or no packet loss.
- Application acceleration – Teridion improves the performance of TCP and provides better throughput for cloud applications.
We deliver great SaaS application performance, particularly when there is a demand for bidirectional data-in-motion or delivery of dynamic content. In some cases, our throughput is 15x or more than what our customers get across the public Internet. That might sound hard to believe, but seeing is believing. We encourage you to try it yourself, for free.