re: Invent Amazon has designed its own 64-bit Arm server processors called Graviton and is currently renting them in AWS.
Just for the annual record: the meeting meeting in Las Vegas, today the internet titan, A1.medium from A1.medium has uncovered A1 family A1 family up to 1 vCPU, 2GB RAM and 3.5Gbps EBS and 10Gbps network bandwidth. $ 0.0255 per hour, ie 16 vCPU and 32GB of RAM with a1.4xlarge up to $ 0.408 per hour.
This can work cheaper than comparable T samples in terms of CPU; properties vary between sample types. Here you can compare and compare for yourself. Amazon claims that the A1 sample is 45 percent cheaper than x86 virtual machines, depending on the configuration.
Some basic benchmarks of a1.4xlarge and c5.4xlarge were made using the Phoronix C-Ray benchmark. Render, 108 & c5, received 112s on the a1. c5 $ 0.68 / hr, a1 $ 0.408 / hr. It really looks like a bargain. #reinvent
– rbranson: EVP, Thoughts (@rbranson) November 27, 2018
In any case, you should not look at raw prices: Due to the differences between Arm and x86, you need to test the software stack in A1 VMs for performance and compatibility before it is delivered to the architecture.
The Graviton CPUs are running AArch64, Amazon Linux 2, Red Hat Enterprise Linux 7.6 or Ubuntu, and in the US East (Northern Virginia and Ohio), the US West (Oregon), and Europe (Ireland) regions of the Amazon. On request for dedicated hosting. Consider scaling platforms, web host, and so on.
If you're running things written in portable scripts or interpreted languages, you can probably pass that code to A1 virtual machines and give it a swirl. If not, then remove the Armv8-A tool chain, get pre-built packages from your preferred Linux distribution, or try pre-built pairs depending on availability.
. If you haven't looked closely, you may not even notice that you are using an Arm server. This was a design requirement, iy conquered Jon Masters, who worked for the acquisition of RHEL supported by A1's Red Hat examples. "I've been playing Amazon's A1 examples for some time, and the team's engineering and the quality of their talents is a chasm."
If you're wondering where all this comes from, we've warned you in 2014 that you're considering using the Arm chips for Amazon's cloud computing resources. An important step in this search was to sell Annapurna Labs for several hundred million dollars in 2015. This focused on Arm on a chip designer and licensee and initially on the projects of Internet of Things. AWS then announced the use of special chips designed by Annapurna in the cloud called Nitro. These network and storage tasks loaded by EC2 hypervisors allow x86-based sample hosting machines to focus on running client workloads, while Nitro ASICs have shifted packages and data.
The chip design team has now used this experience to promote and distribute the Graviton Arm processors in the AWS cloud, so the customer code has been run on the general purpose virtual machines. As you would expect from the extremely mysterious Amazon, there is not much more to the technical details that are currently available. For more information, we trained the PR people of AWS.
By the way
/ Proc / cpuinfo In the case of A1, the CPU ID number is 0xD08, indicating that Graviton is based on the 2015 Cortex-A72 plans of the Arm. We also note that a vCPU is mapped to a physical kernel. Example of 16 vCPU organized In the four quad-core clusters containing 2MB of shared L2 cache per core and four KB cores, the 32 KB L1 cache and 48 KB L1 script cache per core.
Last month, Arm claimed that a million CU-backed data center servers would be sent in 2018 and most of them combined their systems. We had to guess that Amazon would make up a large part of this figure. Amazon also offers AMD Epyc processors in the EC2 cloud. Meanwhile, Microsoft Azure wants to be powered by at least 50 percent of the Arm.
They're putting more pressure on Intel. According to analysts, Chipzilla's x86 processors account for more than 90 per cent of the world data center computing market. Epyc, Graviton, Cavium's ThunderX2 and other competitors are losing their market share here and there. It's a pun. ®