Serverless computing is quickly gaining traction as a way to squeeze more cost and efficiency from cloud deployments, though mainstream use of the model could still be a year away.
According to Rich Sharples, senior director for product management at Red Hat’s JBoss Middleware division, more companies are beginning to use serverless computing, which is designed to reduce the amount of overhead associated with offering services in the cloud. This includes the ability for a cloud provider to dynamically manage server resources.
The market started to gain traction following the launch by Amazon Web Services (AWS) of its Lambda platform. Lambda allows an organization to pay only for the compute time consumed – there is no charge when the code is not running.
Sharples said he has seen initial serverless demand around peripheral applications like transforming website graphic images for use on mobile and similar single page applications specific to mobile platforms.
“There are a lot of cool use cases where cost and efficiency were very important and business cases were narrow enough,” Sharples said of these early deployments.
These initial use cases also highlight the overall economic focus on serverless computing, which are based on better use of resources resulting in less waste. In reducing potential waste, serverless computing pricing models typically break down access by tenths of a second.
Sharples noted that similar to containers, serverless computing is ideal for quick running, low overhead applications.
“Where serverless does not make sense is with any long-running applications or tasks that require significant memory,” Sharples explained.
Serverless Pricing Advantages
In a recent report, 451 Research noted serverless compute pricing is typically based on three parameters: script duration, or how long the code is used; the number of requests; and the memory required for the function.
451 Research favorably compared the total cost of ownership (TCO) of serverless computing to that of virtual machines (VMs). Owen Rogers, research director at the analyst firm, explained VMs needed to be up and running before a function request was placed, thus an enterprise would “need to pay for that, and there is an element of waste when capacity is not being used.”
“With serverless, we don’t have this problem,” Rogers said. “It scales instantly with a request. You just need to configure the request with the function and no other concerns.”
451 Research did note that it was difficult to nail down exact costs of serverless computing. But, in general it found among the larger cloud providers, IBM was less expensive than rival services from Microsoft, Google, and AWS.
IBM’s cost advantage was on usage based on tenths of a second, up to around two second of total usage per script. Helping to bolster IBM’s value is that the company does not round up usage to the nearest usage marker, which 451 Research found to be in 128 megabyte buckets.
“IBM has a simple model,” Rogers explained. “For very small duration scripts, it’s pretty much cheapest all the time, or at least in our parameters.”
For longer usage scenarios, 451 Research found Microsoft’s Azure platform could offer a less expensive product. Google and AWS pricing did not factor into the various usage models provided by 451 Research. Rogers did note that what is the least expensive to use may not provide the most value to developers. In other words: your mileage may vary.
Beyond cloud pricing, Sharples noted a significant serverless cost paradigm is that the developer doesn’t need to care about any of the infrastructure involved with actual deployments. This can increase developer efficiency because they can focus instead on just writing code.
“Developers are expensive and that’s where there will be a breakthrough in terms of the business case for serverless,” Sharples said. “Execution costs are still there, but developer costs are typically more.”
Open Source Focus
As for its own efforts, Sharples said Red Hat is currently using the Apache OpenWhisk serverless platform. Sharples cited the platform’s use of open source and support of Kubernetes as a reason for the move.
“As an open source company, we prefer to join rather than to build from scratch,” Sharples explained.
Red Hat expects to have its serverless platform in the hands of developers later this year, and have it running on its OpenShift platform “sometime next year.”
But, between now and then Sharples said work still needs to be done in terms of evolving various parts of the serveless computing ecosystem. He cited the need for maturation of the developer de-bug experience and diagnostic; more advances in serveless orchestration; and work on chain-of-command functions that could support the ability to launch serverless functions from within an Excel spreadsheet.