Amazon Web Services (AWS) has set out how its investments in artificial intelligence (AI) chips and software are saving customers money and helping them migrate their legacy Windows and VMware workloads off-premise much quicker.
AWS CEO Matt Garman used the opening keynote at the public cloud giant’s Re:Invent customer and partner conference in Las Vegas, which is the first he has delivered since taking over the company reins in June 2024, to talk up the potential for generative AI (GenAI) to digitally transform the way that businesses operate. He also talked at length about the work that goes into ensuring the AWS cloud infrastructure is equipped to cope with the growing demand from its customers for the compute power they need to run AI and GenAI workloads.
As previously reported by Computer Weekly, the demand for GenAI workloads from its customers was recently cited as the reason for a “significant re-acceleration” in AWS’s annual growth rate, with the company reporting a 19.1% year-on-year uptick in revenue during its third-quarter results.
Garman touched on Amazon’s 14-year-long collaboration with Nvidia, which he said has enabled it to roll out a succession of increasingly more powerful graphics processing unit (GPU) instances based on the latter’s technology so it can keep pace with its customers’ AI demands.
The company has also doubled down on the creation of its own AI silicon – namely its family of Tranium chips – to support a wider range of instances that are designed to improve the cost performance of running compute-intensive workloads. To this point, Garman used the keynote to announce that the second generation of Tranium instances had now become generally available, claiming the latest iteration can deliver “30-40%” better price performance than “current GPU-powered instances”.
This is based on feedback from early adopters of the technology, with Garman naming Adobe as among the customers who have seen some “promising” early wins with the technology.
Another is AI-focused software engineering startup Poolside, who has reportedly committed to training all future versions of their large frontier model on Tranium 2. The company is also anticipating the move will generate savings in the region of 40%. “Databricks is one of the largest data and AI companies in the world,” he said. “[It] plans to use Trainium 2 to deliver better results and [to] lower the total cost of ownership for our joint customers by up to 30%.”
Opening up about Amazon’s use of GenAI
The conversation later moved on to how GenAI is also changing the way that AWS operates, with particular focus on how its own offerings are helping to speed up the time it takes to refactor legacy, on-premise workloads and ready them for migration to the public cloud.
Central to this bit of the discussion was Amazon Q, which is the company’s generative AI chatbot assistant that is designed for in-house use by software developers, business analysts and contact centre employees to make the work they do more efficient.
The migration of customer workloads out of private datacentres and into the public cloud is a process that fuelled the company’s growth for a decade or more after its inception in 2006.
However, despite the company previously acknowledging that a large proportion of enterprise workloads remain on-premise, it was an area that was markedly less talked about during the keynote, until Garman flagged how Amazon Q can assist with this task.
“Our goal at AWS is to help every builder be able to innovate, [and] we want to free you from the undifferentiated heavy lifting to really focus on those creative things that make your building unique … [and] generative AI is a huge accelerator of this capability,” he said.
As an example, he talked about how Amazon Q Developer, an iteration of the chatbot specifically designed to help developers speed up their CodeDeploy processes, is helping customers deploy faster, more secure and better-quality software updates.
Garman then went onto announce several new features that were being added to Amazon Q Developer that will generate unit tests, documentation and code reviews on behalf of developers, so they can spend more time each day writing code than dealing with the admin associated with it.
Addressing the legacy
The software is also reducing the amount of time they have to spend managing legacy applications, it is claimed.
“One of [the software’s] most powerful capabilities we already have is [its ability to] automate Java version upgrades,” said Garman. “What it can do is transform a Java application from an old version of Java to a new version in a fraction of the time it would take to do manually. This is work that no developer loves to do, but is critically important.”
According to Garman, integrating this capability into Amazon’s own internal systems saw it “migrate literally tens of thousands of production applications” to Java 17 in a “small fraction of the time” it would typically take. “The estimate from our teams is this saved us 4,500 developer years … [and] this is a mind-blowing amount of time saved, and because we’re now running on modern Java, we can use less hardware, too. So, we saved $260m a year through this process.”
Java upgrades are one thing, but – in Garman’s opinion – a migration that a lot of enterprises want assistance with is moving from Windows to Linux. And this is something AWS can assist with now through the preview release of a new version of Amazon Q Developer.
“Customers love an easy button to get off of Windows,” he said. “They’re tired of constant security issues, the constant packing or patching, all the scalability challenges that they have to deal with, and they definitely hate the onerous licensing costs.
“But we do recognise today that this is hard. Actually, modernising away from Windows is not easy, [but] with Q Developer, modernising windows just got a lot easier … [as it allows you] to transform .Net applications that are running on Windows to Linux in a fraction of the time.”
Signature IT
As an example, Garman flagged digital transactions, signing software company Signature IT, and the work it has done to modernise its legacy .Net applications and migrate them from Windows to Linux. “It was a project they estimated was going to take six to eight months, [and] they actually completed it in just a few days,” he said. “That is a game-changing amount of time.”
But it’s not just Windows workloads that enterprises are having a hard time modernising. “Windows is not the only legacy platform in the datacentre that is slowing down all your modernisation efforts … oftentimes it is VMware workloads that customers would really love to modernise to cloud-native services,” said Garman.
“VMware is deeply entrenched in many datacentres, and has been for a really long time. And what happens is … because it’s been there for a long time, there ends up [being] this kind of spaghetti mess of interconnected applications.”
“[So] really the hardest part about modernising is finding out what are the dependencies of those applications,” he said. “And the migrations are error-prone, because it’s hard to understand if you move something, if it is going to break something else. And again, of course, licensing is expensive.”
To assist with this, Q Developer also has capabilities that will allow VMware-based datacentre workloads to be reconfigured to become cloud-native, with the system able to identify the dependencies and create a migration plan for the user.
“[This] really reduces a ton of the migration time, and significantly it reduces [the organisation’s] risk,” said Garman. “It also launches agents that can convert on-premise VMware network configurations into modern AWS equivalents. This takes what used to be months and months of work into hours to weeks.”
The next complex datacentre migration project the company is looking to simplify for enterprises, with the help of Amazon Q, concerns mainframes, which Garman described as “by far the most difficult to migrate to the cloud”.
“When you talk to customers, just the effort of trying to analyse, document and plan mainframe modernisation is often too much, [and] people give up [because] it’s too hard. Turns out, Q can help with this, too,” he said.
The software has a number of agents in it that are able to do mainframe code analysis, refactor applications and create documentation in real time for legacy COBOL code so enterprises can fill in any knowledge gaps about what it might do.
“Most customers will tell you their mainframe migration will probably take three to five years … but planning a project for three to five years is nearly impossible,” said Garman. “A lot of the time, they just don’t get done.”
And while it’s beyond the capabilities of Amazon Q to make mainframe migrations a “one-click” job right now, he said early testing suggests the software could significantly accelerate the pace of these projects.
“We think Q can actually turn what was going to be a multi-year effort into a multi quarter effort, cutting by more than 50% the time to migrate mainframes,” said Garman. “If you can take a multi-year effort and bring it down to a couple of quarters, that’s something that people can really get their heads around. And customers are incredibly excited about this.”
