As an application performance monitoring (APM) vendor that mainly serves the digital transformation market, New Relic has a grandstand view of what's happening in this next-generation evolution of enterprise IT. Its 15,000 customers are the early adopters of technologies destined to go mainstream over the next few years, providing an invaluable guide to future trends.
The most surprising trend that's coming up fast, says Lee Atchison, Senior Director, Strategic Architecture, is adoption of serverless computing, in which applications are assembled out of on-demand cloud services, such as AWS Lambda. A recent survey of customers found that this new technology has already been adopted by the vast majority:
We were surprised at the number of customers that said Lambda was a significant part of their infrastructure. 90% said it was significant. I'm personally surprised at how fast that's been adopted. People are moving away from large, monolithic applications.
Meanwhile, the use of container technology, such as Docker, is already endemic, and the pattern of usage in New Relic's customer base points to a wider trend towards dynamic cloud infrastructure, Atchison believes. Since an application can be brought up on a container almost instantly, whereas a virtual server takes longer to set up, containers are often used for very short periods of time.
Analysing aggregate Docker usage across millions of instances at New Relic customers, it turns out that most run for less than an hour in total and as many as one in ten are running for less than one minute. Atchison concludes that this illustrates a new approach to operations that's best suited to environments that must scale fast, cope with sudden spikes in usage, and accommodate rapid change:
Dynamic cloud allows you to build applications faster, but the way you run operations is different. Servers you might allocate for hours or minutes, Docker containers for minutes or seconds, serverless for milliseconds.
This is really the future of the dynamic cloud. This dynamic allocation of resources is the way the world is moving in the dynamic cloud.
This impacts performance monitoring too, because managing the virtual infrastructure becomes an integral part of how the application operates when servers are no longer tied to specific physical assets:
Now resource provisioning is part of your application. But how do you monitor something that only runs for milliseconds?
There are no serial numbers and spreadsheets of resources. In the dynamic world, things can't be managed in the same way they were in the static world.
So as enterprises move to cloud, he concludes, it becomes essential to monitor the entire stack, not just the application on top of it, because the interaction between the two becomes a material factor.
These findings provided a convenient backdrop last week to New Relic's launch of Health Map, which provides a unified view of applications and the underlying infrastructure, as shown in the screenshot at the top of the story. This builds on the vendor's core APM product along with last year's introduction of New Relic Infrastructure, which tracks real-time health metrics and configuration changes across cloud and hybrid infrastructure, from Amazon Web Services and Docker containers to bare-metal servers.
Health Map brings together data from both applications and infrastructure into a high-density overview that gives users a wide range of search, filtering and drill-down options, on web and mobile. Designed to help customers manage the kind of dynamic cloud environment that Atchison decribes, it was an obvious follow-on from the introduction of Infrastructure, says Greg Unrein, Vice President, Product Management:
As we were developing the Infrastructure product for launch, we recognized [that] if we could merge those two views it could become really useful.
This is especially true in the DevOps world, where the underlying infrastructure behaves more like programmable code and therefore has to be managed in much the same way as the applications running on top of it. As well as supporting the change in mindsets, giving real-time visibility into what's going on is crucial to building confidence as enterprises move to DevOps, says Unrein.
It supports the change to DevOps because it encourages people to think more in terms of code.
On the enterprise side we see New Relic helps cultural change because it gives more visibility.
Extending the reach
The challenge the vendor now faces is to extend the reach of its products further into the hybrid infrastructure that its growing enterprise customer base has to manage. The release of an SDK for its infrastructure tool means that partners and customers can now bring additional resources under management, he says.
As we've grown the enterprise side of our business, that's translated into pushing visibility deeper into the legacy assets.
An API-first approach is very deliberate because we see the first priority is enabling anyone else to go out and take advantage of the product.
As analyst James Governor pointed out in a blog post on retailer Ocado's use of New Relic, its customers are creative enough even to adapt its products to monitor IoT landscapes. But as Unrein admits, Health Map doesn't yet encompass serverless infrastructure. That's on the roadmap, he implies, as well as adding the ability to show business-level metrics such as customer experience:
Right now it's very orientated to applications running on containers or even on bare metal rather than serverless cloud usage.
If you had an app running fully serverless, that isn't what Health Map is optimized for yet. The other thing [to come] is pulling in more of the customer experience side of the data.
Even for vendors, it's a race to keep up with the fast-moving pace of adoption. Atchison also worries that as adoption grows, customers may find they lack the right skills to manage serverless environments:
The adoption rate of small services is significantly faster than I would have expected ...
One of the fears I have is, we're moving to too-small services. Complexity isn't going away — it's being replaced by service interdependencies. The danger is you need a higher-end engineer to understand the interactions after a move to microservices.
You're moving who is responsible. The bar is being raised, as far as their experience level goes.
It's up to early adopters to navigate a landscape in which both tools and skillsets are still evolving, beating a path that others can follow. These are valuable insights from New Relic's customer base about the latest trends and the pitfalls that surround them. There's more to come — I'll be writing up customer stories from last week's event, including Ocado and Ryanair, in the coming days.