A recent conversation about application performance set Catchpoint CEO Mehdi Daoudi on edge—metaphorically speaking.
“[I’m] a little bit tired of hearing some CDN vendors and their customers say CDN is a commodity, and no one cares about performance,” he wrote on LinkedIn. “Really? The whole point is performance, speed, reach!”
Content delivery networks are the original edge computing application and work by serving content from the regional edge (or, in some cases, an access edge), as defined by the LF Edge open glossary. One can point to Akamai as the originator of the CDN market, which used distributed compute, storage, and network services closer to the end-user to enable better application performance. Static web pages were the original “application” to run on CDNs, but dynamic content, streaming video, and security services are now among the many applications and functions to be served from the edge.
LF Edge’s edge continuum
Daoudi spoke to Edge Industry Review about why performance at the edge still matters. Catchpoint has skin in the game—the company offers observability tools for application and network performance. Catchpoint’s customers obviously believe performance matters.
Building (and monitoring) a better internet with the edge
“Covid has shown we need a better internet. Telehealth experiences are choppy. Hospital networks are not well equipped to offer telehealth—[doctors] maybe on WiFi connections. Low latency is critical,” Daoudi said. “Then I hear performance doesn’t matter. But it does when we’re talking about self-driving cars, about homeschooling.”
Solving performance issues when users complain means triaging hardware, network, and software issues. Was it network latency that caused an issue with video playback? Was the latency due to a temporary routing issue, and if so, which users were affected? Was there an issue with API calls? Solving these issues requires the ability to triangulate among the different variables, and observability tools are needed to do that. It stands to reason that enterprises are indeed interested in performance.
“At the end of the day, we don’t do it (observability) just for the sake of observability. We do it because IT needs to deliver on their business,” Daoudi said. Observability can mean many different things, but Daoudi believes that monitoring from the end-user perspective is ultimately what matters most.
Edge performance has become even more important during the pandemic as companies sometimes have to let thousands (or tens of thousands) of workers work remotely. One CIO told Daoudi he used to have 2000 employees in three offices; “Now I have 2100 offices to manage!” Catchpoint responded by introducing its Employee Experience Monitoring product in 2020. So what’s next?
People are directly impacted by poor edge performance. Are we talking about the same thing with applications running on the access edge managed by a company Vapor IO or a telco like Verizon or T-Mobile? What about applications running on an edge server in a manufacturing plant, where machine-to-machine communication is paramount, not machine-to-human communication?
Observability at the Programmable Edge
In either case, Daoudi would argue that delivering to the last mile will be the key challenge for AWS and other cloud providers. “Take vaccine distribution as an example – sending [bulk shipments] to LA is not an issue. It’s the last mile delivery” to pharmacies, hospitals, and doctor’s offices before the vaccines expire, he said.
Daoudi has already seen interesting uses of edge worker-type applications that are performing operations closer to the end-user.
“Applications are so rich and demanding, you need to keep gaining the milliseconds. On a typical webpage with 300 to 500 objects, you are adding up the wait. It’s heavy,” he noted. That leads to an interesting challenge for Catchpoint in scaling its technology as well. One customer is launching measurements on AWS’ Wavelength service. That’s only the start, as far as Daoudi is concerned.
“What we’re seeing now with customers, PoS (point of sales systems), gaming consoles, thermostats are becoming the “user,” according to Daoudi. In the end, the edge computing trend will result in the reinvention of the CDN and other edge computing services as companies seek not only to enhance application performance, but ensure that applications are consistently performing well, and continuously available. (For more on Daoudi’s take on edge computing and application performance, you can read his Guest Opinion piece on Edge Industry Review here.
So does it make sense that someone at a CDN vendor might say that CDN services are a commodity? Yes. And no. For those focused on selling a service like software delivery or streaming video delivery, there is a lot of competition and price pressure. But even there, customers who are driving revenue from those services (think Hulu or Disney+), performance still matters. If the vendor’s performance isn’t good in Brazil, for instance, the customer will use a different CDN. If the vendor’s performance suffers during peak loads, the customer shifts workloads to another vendor in near real-time. So, in the end, the vendor that says performance doesn’t matter is either selling to the wrong type of customer or they have trouble meeting performance requirements after a certain point.
Now that the definition of edge computing is evolving to include more advanced applications, CDN vendors like Akamai, Cloudflare, Fastly, and Lumen (formerly Level 3) are continuing to advance new forms of programmable edge services. How much will the performance of these services matter? Will it be measured in seconds? Sub-seconds? Milliseconds?
Whatever the developer’s requirement is, performance will matter, meaning that monitoring network and application performance is paramount. Developers need to be able to continuously monitor the entire application lifecycle to effectively harness the promise of edge computing.
Akamai | Catchpoint | CDN | Cloudflare | DevOps | Fastly | Lumen | monitoring | observability