This Cortex article recently ran in SiliconANGLE.
The rise of serverless technology as part of a broader cloud application modernization movement has been going on for a few years. But it seems like the serverless romance has just sparked up, as the execution rate of functions-as-a-service (FaaS) approaches like AWS Lambda spiked dramatically over the last year, according to observed industry data.
Average weekly invocations of Lambda functions increased by 209% between June and December of 2019, with an impressive ‘holiday spike’ of additional transactional use showing at the end of the year, according to a February 2020 study of AWS customers by cloud observability vendor New Relic titled “For the Love of Serverless.”
Whether this uptick in volume is an indicator of increased use among an existing set of early customers, or the beginning of a real hockey stick curve in the number of dev shops utilizing serverless, it’s clear new technical talent will need to be cultivated to explore this change.
Lambda evolves with deepening developer skillsets
“Understanding how to deploy serverless applications is a challenge,” said Dave Townsend, Principal Software Engineer in the Innovation & Architecture group at Matson, Inc in a related interview. “There is really no separation of code and infra now, so that takes some getting used to. It is still really early in this game, and I think as a community we’re evolving.”
Teams are inherently designing applications with the decoupling of logic and state in mind. The survey also shows that by quantity, the vast majority of functions customers monitored were various versions of Node.js (52.5%) and Python (35.8%) code, with lean code sizes and shorter (median 718ms) invocation times.
Several longer duration / larger memory sized functions noted in the serverless survey were attributed to oft-invoked Java coded functions, which were not always able to start up at the responsive <100ms time frames seen for other languages.
“We measured a number around function duration, and had a lively discussion about why Java runtimes could be taking longer to run,” said Andrew Tunall, GM, New Relic Serverless & Emerging Cloud Services.
“AWS could have been dogmatic about the technology, but instead they announced pre-provisioning functionality to customers with latency requirements — it’s meaningful that they meet the customer halfway to allow them to use functions-as-a-service.”
“In the past year, many serverless platform issues have been addressed, including provisioning of ‘warm’ functions and improved scalability when attached to virtual networks. Now that serverless can handle almost all use cases, the biggest technical challenges will continue to be developing the right workflows and tools for a broader developer base to adopt serverless,” said Farrah Campbell, Ecosystems Director, Stackery in the report.
What’s good for startups is good for enterprises
You can see how many companies get hooked — usually on a net-new application development exercise that is designed by nature to make rapid procedural calls to FaaS, perhaps with other adjacent serverless technologies for data management and observability, delivering a highly successful and scalable result at a very affordable cost.
“The combination of serverless architecture and stateless code such as AWS Lambda has given us the ability to create massive scale for our platform in a cost effective way, which incentivizes us to find creative ways to use these platform services,” said Michael Semick, CTO of startup Blueocean.ai.
“It’s attractive not having to maintain servers and operating systems, but the downside is tracking and organizing the growing number of Lambdas, which can create complexity for automated deployments as well as security and governance,” said Semick.
Many customers with existing application suites cite dependencies and technical debt as limiting factors to company-wide adoption, but the business case still needs to be justified first.
“It is often challenging (or even impossible) to show upper management a convincing like-for-like cost comparison of an on-prem or hosted environment versus a serverless estate. Without this direct comparison, it often delays the migration to serverless or management buy-in into serverless,” said Sheen Brisals, serverless solution architect at The LEGO Group.
Pervasive rise among peers in the category
All of this adoption data isn’t simply coming from one place. Rival APM vendor Datadog also recently produced its own State of Serverless report, demonstrating similar hockey-stick growth results for Lambdas — with the marquee statistic that 50% of their AWS customer shops are also exercising serverless functions.
But the most surprising result of that survey? The greatest rate of usage for Lambda functions is weighted toward large, enterprise-scale environments. So it’s really not all startups singing serverless praises…
According to Forrester, nearly 50% of companies are either using or planning to switch to serverless architecture within the next year, as cited in the recent New Relic report.
AWS can be expected to handily lead this charge, but similarly strong rises in Azure Functions and Google Cloud Functions should also be expected, especially for their ideal customers and workload types.
The Intellyx Take
The realization that ‘serverless isn’t really serverless’ is old hat — obviously, even an ephemeral function that lives in a container in a cloud for less than a second is still running on a server somewhere. Just not one you’d manage.
This high-growth takeaway is leading enterprise applications along the continuum toward an atomicity of function, as virtualized cloud instances of infrastructure gave way to hundreds of services and containers, even more orchestrated Kubernetes clusters, all of which may be calling on many thousands of Lambda compute functions, and other serverless instances like AWS DynamoDB for data persistence.
The initial successes of serverless may cause companies to embark on more daring net-new re-architectures and replacements of existing legacy dependencies.
Yes, there will always be dependencies. But at this price-per-function, what’s stopping your company or team from at least flirting with FaaS?
© 2020 Intellyx LLC. Intellyx advises enterprises on their digital transformation strategy, and publishes the weekly BrainCandy and Cortex newsletters. At the time of writing, New Relic is an Intellyx client. Microsoft is a former customer. None of the other companies mentioned here are Intellyx customers. Image source, Changwan Han, flickr, edited by bluefug.
