The need for higher computing speed is never-ending
It’s been the rule since the beginning of digital computer history. There’s always more data to be crunched, so the need for faster hardware and software is always there. It’s a race without a finish line, and even in today’s leading-edge cloud computing applications, the stakes are the same.
Therefore, it’s no surprise that Amazon Web Services (AWS) chose its 2019 AWS re:Invent conference to announce two significant efforts designed to accelerate cloud applications, as well as a host of flexible options that enhance the performance their cloud instances provide.
AWS introduces Graviton2-based instances
The biggest news on the speed front is the release of the new Graviton2 cloud instances, which deliver a whopping 40% increase in the price/performance ratio – more power for less cost – than existing instances based on older processors (x86 “Skylake” for the chip geeks). For the benefit of those same geeks, Graviton2 instances are based on the Arm Neoverse N1 chip. The big difference here is that x86 chips were designed for the legacy servers that cloud instances have mostly replaced; the N1 is made for the specific demands of cloud-server workloads.
AWS is making these Graviton2-based instances for mainstream use (M6g), compute-intensive applications (C6g), and memory-intensive applications (R6g). This is a speed/cost advantage that will have a ripple effect across the entire XaaS (Anything-as-a-Service) industry, eventually lowering the cost of entry for end-users and allowing developers to build more robust and capable applications to serve a growing and diversifying market.
M6g instances are available now in preview, while C6g and R6g instances are coming later in 2020.
Amazon EC2 Inf1 Instances are here, featuring Inferentia
In 2018, Amazon first broke word on Inferentia, an AWS-built chip designed for machine learning. It was huge news because AWS designed Inferentia expressly to make machine-learning inferencing more cost-effective. Inferencing allows the software to learn by using models they’ve already used to perform tasks, and it lets programming make accurate predictions of future outcomes based on previous results.
Now, Inferentia is here. As AWS says,
“With Amazon EC2 (Elastic Compute Cloud) Inf1 instances, customers receive the highest performance and lowest cost for machine learning inference in the cloud. Amazon EC2 Inf1 instances deliver two times higher inference throughput, and up to 66 percent lower cost-per-inference than the Amazon EC2 G4 instance family, which was already the fastest and lowest cost instance for machine learning available in the cloud.”
Cloud machine-learning workloads are divided into training and inference. Both types of machine learning have long been the domain of repurposed graphics-processing units (GPUs). They were adapted to this use because they were designed to handle many parallel calculations using thousands of cores, with the memory bandwidth to deal with these cores’ demands.
AWS’s deployment of Amazon EC2 Inf1 instances with the Inferentia processor marks the launch of the first dedicated machine-learning platform unique to the cloud. It’s designed to take full advantage of its unique operating environment and leaves the vestiges of legacy processing solutions behind.
Other big news from AWS re:Invent 2019
- Local Zones: AWS extends the cloud to more locations and closer to end-users. This is excellent news for applications that perform better with low-latency (less lag time) connections.
- AWS Wavelength: AWS now embeds computational and storage infrastructure inside wireless providers’ 5G networks to help mobile-application developers deliver data with “single-digit millisecond” latency times.
- Amazon SageMaker Studio: Boosts developer productivity with beginning-to-end machine-learning workflows. Get products to market quicker.
The cloud keeps getting faster – stay up to speed with CloudHesive
If you’ve been considering taking your IT to the cloud, you can’t do better than going with AWS, the leading cloud-services provider, and CloudHesive, the partner you need to help make it all happen – from the actual migration to keeping your cloud up and running. Learn more by getting in touch with CloudHesive at 800-860-2040 or through our online contact form.
