If you’re an enterprise digital professional who’s ever marveled at the speed and simplicity of Facebook, Twitter, or Instagram, and wondered why your own web and mobile efforts can’t offer the same performance, this week’s O’Reilly Velocity conference in New York City had some important lessons for you.
Known mostly as a confab where web-scale performance techs compare notes, this year’s Velocity welcomed increasing numbers of traditional enterprises – all of whom realize that their customers expect and demand performance to equal the fastest web scale company.
Web and Mobile Performance: The Buck Stops with the User
For this exceptionally technical crowd, the Velocity audience was surprisingly focused more so on people than on technology: how best to work within teams, how to support the business and its goals, and most importantly, how to improve and optimize the user experience. All the underlying technology, from the back-end systems of record to the web infrastructure to the cloud, all line up to deliver value to the customer on their phone or sitting in front of their computer.
Furthermore, the stakes have never been higher. Today’s web and mobile users are increasingly fickle; a delay of a fraction of a second can lead to abandoned shopping carts, frustrated users, and quick switches to a competitor’s site or mobile app. Improving the speed of the experience, however, is simply the price of admission.
Users also expect increasingly rich, interactive experiences. Today’s Web sites deliver dozens of images, video files, embedded applications, sophisticated ads, and other dynamic content. Even mobile apps participate in an increasingly large and complex ecosystem of other apps and Internet services, and they must all deliver in milliseconds.
Achieving maximum performance thus means pushing increasingly complex technology to its limits. Performance engineers work to squeeze every millisecond out of the technology, which means pushing end-to-end infrastructure to its breaking point. Failure, therefore, is always a single click away.
As a result, performance engineering is more about failure than about speed. For this reason, the primary mission for these teams is resilience.
Resilience is the ability to respond to and recover from failure. Knowing when something is about to fail, quickly identifying the problems, understanding the root cause of each issue, gracefully degrading performance while people are working on solutions, and delivering those solutions quickly and permanently are all important pages in the web and mobile performance playbook.
Resilience is a key aspect of business agility, along with responsiveness and innovativeness – and yet, business executives traditionally underinvest in resilience. Performance engineers call this lack of support the “faster better cheaper” (FBC) problem. Stakeholders continually push technology teams to quickly achieve their goals with constrained budgets, but such FBC-style decision making leads to increased brittleness, as better typically means more features, instead of improved resilience.
The FBC problem is not a technology problem – it’s an organizational problem. The solution is for the performance engineering personnel and the digital business stakeholders to communicate better and thus work more effectively as a team.
It’s essential, therefore, for the digital professional to understand the tradeoffs between features and performance: add more bells and whistles and the site or the app will slow down. Push it to run faster and you increase the chance it will fail. Such failure translates directly into lost revenue and other adverse business impacts. Each digital team must therefore balance the goals of better functionality and resilience in order to optimize increasingly dynamic business outcomes.
Vendors Driving Optimized Business Outcomes
I interviewed eleven vendors at Velocity. Several focused on web site monitoring, where the primary capability was combining synthetic and real user monitoring. Synthetic monitoring uses automated scripts to simulate users, and works well for testing apps before they go live and for establishing performance baselines. Real user monitoring gives businesses visibility into how users are experiencing the site from moment to moment.
In this market, AppNeta offers end-to-end, deep diagnostics of each issue, providing detailed information on transactional performance. SOASTA ties user monitoring to business metrics, with predictive “what if” analytics that help web teams improve their sites. Catchpoint offers extensive third-party monitoring including last mile probes that test real-world behavior, even going so far as to put their probes in people’s homes.
SmartBear’s AlertSite can identify the particular line of code responsible for an issue, and provides answer-centric alerts that aid with root cause analysis. Finally, Keynote offers a cloud-based tool that includes testing real mobile devices in the cloud. Keynote also provides an industry scorecard so customers can compare their performance to competitors.
Web performance testing is only part of the performance battle. Radware provides active Web and mobile optimization on an appliance that goes in your data center in front of the Web server, now with virtual or cloud-based options. Radware is able to optimize individual parts of each page, ensuring that relevant content arrives in the appropriate sequence. Chef is familiar in the cloud world for automating cloud deployment, helping enterprises follow web scale continuous deployment practices.
Crittercism focuses entirely on mobile application performance management, providing mobile app developers with deep insight into how their apps are behaving in the mobile ecosystem. In contrast, ThousandEyes provides deep insight into network performance, including the internal network, mobile infrastructure, as well as in the cloud. ThousandEyes reduces fingerpointing among software-as-a-service vendors, cloud providers, connectivity providers, and on-premise infrastructure teams.
Several content delivery networks (CDNs) also exhibited and presented at Velocity. CDNs replicate content across the Internet to bring it closer to users, thus improving performance and lightening the load on the servers that originally contain the content. CDN vendor Fastly focuses on delivering dynamic objects including mobile video and user-generated content. Highwinds is a CDN who has built out their own global network, offering last-mile peering arrangements with Internet service providers around the world. They focus primarily on the media/entertainment, advertising, and gaming industries.
Empathy: The Unexpected Catchword of O’Reilly Velocity
While all the exhibiting vendors had solid technology stories to tell, and most of the sessions were deeply technical, perhaps the most important meme of the conference was empathy. Technologists focusing on web operations and performance run the risk of leading insular, technology saturated lives – but this community is increasingly coming to the realization that such an attitude is bad for the individual, bad for their team, and bad for business.
This change in the organizational context for web operations and performance was clearest during discussions of DevOps, a cultural and organizational shift spreading wildfire-fast through enterprise technology organizations around the world. The basic idea behind DevOps is to bring developers and operations personnel together to increase the speed an organization can deliver quality software.
Many software developers believe the driving force behind DevOps is the automation of the operational environment. The better automated the production environment is, so the argument goes, the more developers can handle deployment and administration tasks. You might think, therefore, that the Velocity crowd would be resistant to DevOps, as automation might put their jobs at risk.
On the contrary. This audience fully supports DevOps, as for operations people, DevOps means collaboration, communication, and empathy – basically better teamwork, including developers as well as business stakeholders on extended teams. While automation is critically important to ops people because it simplifies their lives and raises the level of value they can provide to their organizations, it’s not the primary driving force behind DevOps.
This focus on teamwork carried through the conference keynotes. As keynote speaker Mikey Dickerson, Administrator of the new US Digital Service explained, the secret to his team’s dramatic rescue of the Healthcare.gov web site wasn’t technological. The key was resolving communication issues among the numerous participants in this complex and vitally important web initiative – proving that even the best technology in the world won’t give you top performance without effective teamwork.