LAS VEGAS — Amazon on Wednesday unveiled a host of new AWS features that make its online-computing rental simpler for basic users and far more powerful for those who need the most massive computational engines.
Speaking at a keynote for the company’s fifth annual AWS re:Invent conference in Las Vegas, AWS CEO Andy Jassy revealed Amazon Lightsail, a new service intended to simplify launching virtual private servers which can be used to run a website.
“We do all the work behind the scenes — just hook up your IP address and it’s ready to go,” Jassy said. But those interested in tinkering can do so. The baseline package starts at $5 a month.
At the power-intense end of the spectrum, “we think a lot about hardware acceleration,” Jassy said. “We want everyone to have access to the same power that the biggest companies in the world have.” In preview is the new F1 instance, which uses field-programmable gate arrays (FPGAs). “They allow building your own accelerations. They include all you need to develop, test and debug your code,” Jassy said.
Here AWS appears to be playing catch-up with Microsoft Azure, which introduced FPGAs in October.
In between the two extremes, AWS introduced a number of new compute options, designed to let organizations use and pay for precisely the level and type of computing they require at any given time.
“I think it’s fair to say we love us some compute,” Jassy said.
There’s a new memory-optimized family, the R4, with 488 GB of memory, twice as much as the next-most-powerful offering. It’s twice as fast and has 1.8 times the cache of the next-most-powerful member of the family.
Among I/O-intensive instances, in the first quarter AWS will offer a new member of the I3 family, with nine times the IOPS, twice the memory, 2.3 times the storage and twice the number of CPUs. The C5 family, which is compute-intensive, in the first quarter will get a new member, with twice the CPU, twice the performance, three times the throughput and 2.4 times the memory of the next-most-powerful service.
GPU, which is more expensive than CPU, isn’t necessarily required in huge amounts. So AWS is putting in preview Elastic GPUs for EC2, which allows using GPU with any of the nine CPU instances. “It’s super-useful if you need some amount of GPU but not a full GPU instance,” Jassy said.
And, as expected, AWS added full PostgreSQL compatibility to Amazon Aurora.
re:Invent and AWS continue to grow
“The response has been overwhelming” to the announcement in October of the VMware cloud in AWS, which will let customers use the same software to manage virtual machines running in-house and for AWS instances. “In every aspect, we’re seeing the customers and the market and the roadmap of innovation,” said VMware CEO Patrick Gelsinger, on stage with Jassy for the second time in as many months.
At 32,000 attendees, this year’s crowd at re:Invent is far larger than the seating capacity of any NBA or NHL arena. In its first year, 2012, the conference had only 6,000 attendees, according to this nostalgic look back.
In the 10 years since its creation, AWS has become a cloud-services powerhouse. It has saved its parent company from posting an operating loss in several quarters and is on track toward a $13 billion per year run rate, making an understatement out of CEO Jeff Bezos’ April prediction that the service will deliver $10 billion in sales this year.
Microsoft Azure runs second by most measures. Although Microsoft’s commercial cloud business achieved a $12.1 billion run rate at the end of its fourth fiscal quarter ended June 30, Microsoft extrapolated that annualized rate from a month’s worth of revenue — not a full quarter, as AWS does — and on revenue from multiple products, including Office 365 Commercial and Dynamics Online, as well as Azure.
Amazon bases its run rate solely on Infrastructure as a Service (IaaS) and Platform as a Service (PaaS).
Alphabet, whose Google Cloud Platform ranks third, still doesn’t break out numbers for that offering.