I'm a Data Engineer by day at Riot Games (via X-Team ) and by night, I curate the last news/product announcements/resources about AWS Silicon (Graviton, AWS Nitro, Inferentia, and Trainium).
Share
AWS Graviton Weekly # 16: Week from December 16th, 2022 to December 23th, 2022
Published about 2 years ago • 9 min read
You're receiving this because you subscribed here OR here
This email may contain affiliate links. I receive a small commission for recommending products I use & love at no extra cost to you.
Issue # 16: December 16th, 2022 to December 23th, 2022
Hey Reader
Welcome to Issue # 16 of AWS Graviton Weekly, which will be focused on sharing everything that happened in the past week related to AWS Silicon: from December 16th, 2022, to December 23rd, 2022.
We are excited to announce support for the Amazon Graviton2 instance family in four additional regions. Supported instance types include general purpose (M6g), compute optimized (C6g), and memory optimized (R6g and R6GD) instances. Support for C6g, M6g, R6g and R6gd is available in the Europe (Paris) region. In addition, the Asia Pacific (Mumbai), South America (Sao Paulo), and Canada (Central) regions already supported C6g, M6g and R6g instance families, and we have now added support for R6gd in these regions.
Amazon OpenSearch Service Graviton2 instances support OpenSearch versions and Elasticsearch versions 7.9 and above. With Amazon OpenSearch Service, Graviton-based instances provide up to 30% better price-performance than comparable x86-based Amazon Elastic Compute Cloud instances. Further savings are available through reserved instance (RI) pricing for these instances.
The newest chip is the latest piece of Amazon’s effort to build more of the hardware that fills the massive data centers that power AWS. Amazon says making its own chips will give customers more cost-effective computing power than they could get by renting time on processors built by the likes of Intel, Nvidia or Advanced Micro Devices.
The move has put AWS in direct competition with those companies, which are also among its biggest suppliers. DeSantis said the chipmakers remain “great partners,” and that AWS plans to continue to offer high-performance computing services based on chips made by other companies.
Amazon Web Services has secured a five-year contract with the US Navy for cloud services, just weeks after scoring its share of a major US Department of Defense deal for cloud computing.
The cloud division of online marketplace Amazon has been awarded a contract worth $723.9 million by the Department of the Navy as a single-award fixed-price enterprise software license blanket purchase agreement. The details were disclosed in a contract notice posted on the Department of Defense website.
Take machine learning workloads for example. Software engineers have traditionally relied on expensive, power-hungry GPUs to do everything from model building to inference.
However, this one-size-fits all approach is not efficient—most GPUs aren’t optimized for these tasks. In the coming years, more engineers will see the benefits of moving workloads to processors specifically designed for things like model training (AWS Trainium) and inference (AWS Inferentia). As this happens, a new wave of innovation will begin. By realizing a 50% cost-to-train savings with a Trainium-based instance, or 50% better performance-per-watt on an Inferentia2-based instance, engineers and businesses alike will take notice, and we will begin to see a massive migration of workloads. The same will be true even for generalized applications, where there are still benefits to moving to custom silicon, such as Graviton3-based instances that use up to 60% less energy for the same performance than comparable EC2 instances.
In this post, we’ve illustrated how you can quickly and easily construct multi-architecture container images with GitLab, Amazon EKS, Karpenter, and Amazon EC2, using both x86 and Graviton instance families. We indexed on using as many managed services as possible, maximizing security, and minimizing complexity and TCO. We dove deep on multiple facets of the process, and discussed how to save up to 90% of the solution’s cost by using Spot instances for CI/CD executions.
These modifications can help you optimize your AWS costs without having to introduce any application re-architecture, system re-architecture, or engineering overhead.
To ensure cost proper prioritization, here are the proposed changes in order of required effort:
- Upgrading Amazon EBS Volumes from GP2 to GP3
- Swapping Amazon RDS/Amazon Aurora underlying compute to Graviton-based instances or latest generation instances depending on your DB engine
- Migrating existing Linux-based workloads to run on Graviton 64-Bit Arm-based Amazon EC2 instances.
AWS Graviton processors are custom built to deliver the best price performance for an organization’s cloud workloads. While providing significant price performance benefits, AWS Graviton processors help customers reinvent their businesses by innovating quickly and gaining better performance for a variety of workloads.
Discover how AWS Graviton’s optimized processors help provide a superior price-performance ratio. Available for AWS-managed services, you’ll gain insight on strategies, use cases, and insight on how to get the most out of AWS Graviton.
In a quest for lower energy consumption and higher performance in CPU-based machine learning workloads, Graviton3 is AWS’s next-generation ARM-compatible server processor range.
However, is the hype with Graviton3 processors and their core differential benefits really worth it? This article will help find the answers.
Transitioning an application to support Graviton depends on the bytecode-compiled language and can involve many steps. But the motivation to move towards Graviton support can be quite strong when optimising for cost.
Depending on the state of the containerised application, adopting Graviton may be as simple as replacing x86-based instances with Graviton instances.
What does that mean? It's a CPU that is designed to accelerate floating point and vector math. That may not sound exciting to you but for high performance computing environments that's exactly what the doctor ordered.
This custom unit is going to be aimed at giants that are looking to eake out the last drop of performance from those life sciences models. At the same time, Amazon introduced the Nitro version 5, the latest in their groundbreaking DPU family.
In that exact moment, Ian explained the incredible feautures behind the HPC7g, the new Amazon EC2 instances, focused on High Performance Computing, based in the new Graviton3E processor.
Are you looking for Amazon EC2 instances to deploy low-latency workloads that also require high-performance storage capacity (for example, MySQL, MongoDB, Hadoop, ElasticSearch, or Apache Kafka)?
This session dives deep on the different Amazon EC2 storage optimized instance offerings, including a discussion of SSD performance, AWS Nitro SSD advantages, price performance, and ways to optimize your clusters using Amazon EC2 storage optimized instances with Intel and AWS Graviton. Learn when to pick Amazon EC2 storage optimized instance offerings and/or Amazon EBS to run your high-performance storage workloads, based on your workload requirements.
Want to minimize costs on Amazon EKS without affecting the performance of your applications? Then this session is for you. Learn how to manage an efficient infrastructure on Kubernetes and optimize costs on AWS using Amazon EC2 Spot Instances, AWS Graviton, and Auto Scaling policies. EC2 Spot Instances allow you to run your applications in containers, obtaining an average savings of 65 percent on infrastructure without impact to your application.
Graviton-based processors are designed by AWS to deliver up to 40 percent better price performance over comparable x86-based instances. Finally, hear about other techniques and methods to further optimize Amazon EKS node groups.
In this exact moment of the keynote by Barry Cooks (VP of Kubernetes at AWS), you can see a demo provided by Sheetal Joshi (Senior Developer Advocate at AWS) where she explains how Karpenter first bin packs, then shifts to Graviton and then uses spot to bring cost down by 70% all while respecting pod disruption budgets!
Rising energy prices and worries about the infrastructure’s contributions to climate change mean the traditional approach to delivering more performance in the data center is proving too costly and energy-inefficient.
Arm and its partners are delivering an increasingly popular alternative to the old-school approach: World-leading silicon that achieves unparalleled performance efficiency along with an ecosystem of innovation that makes solutions more available to OEMs, software developers and end users than ever. In the first of a two-part series, Dermot O’Driscoll, VP of Arm’s Infrastructure Line of Business, sits down with host Geof Wheelwright to discuss how Arm technology is helping overcome these challenges.
In the second of a two-part series, Dermot O’Driscoll, VP of Arm’s Infrastructure Line of Business, chats with host Geof Wheelwright to changing design requirements in the data center and infrastructure.
There are 17 active candidates ready for interviews:
Events
No events in the time of the year
Quote of the week
With Graviton, we saw that the industry is moving in a direction of utilizing ARM chipsets and we understood that for a long-term investment in terms of savings, we need to be able to support that.
We didn't necessarily move our entire platform, in fact we use a mix of both ARM and AMD Spot Instances, we just understood that we didn't we wanted the option to use ARM when available.
We first moved our MySQL databases on RDS to Graviton without any kind of impact that we could perceive and so the next step was to move our applications to Graviton.
There was really no change needed in our application it was just a matter of getting the Docker containers to be able to run on either type of node .
Uh, it's slightly less expensive in the Spot Instances type, of course spot instances price fluctuate so you keep an eye on that but it gave us about a couple of percentage points like 5 to 10 percent of savings on the Spot instances.
I'm a Data Engineer by day at Riot Games (via X-Team ) and by night, I curate the last news/product announcements/resources about AWS Silicon (Graviton, AWS Nitro, Inferentia, and Trainium).
Issue # 115: November 29, 2024 to December 13, 2024 Hey Reader. Welcome to Issue # 116 of AWS Graviton Weekly, which will be focused on sharing everything that happened in the past week related to AWS Silicon: from November 29, 2024 to December 13, 2024. We are changing the cadence of this newsletter from weekly to bi-weekly. We think that bi-weekly will have a richer format and more meat for the brain. The past week was AWS:reInvent 2024, and the event was full of incredible announcements...
Issue # 115: November 22, 2024 to November 29, 2024 Hey Reader. Welcome to Issue # 115 of AWS Graviton Weekly, which will be focused on sharing everything that happened in the past week related to AWS Silicon: from November 22, 2024 to November 29, 2024. On Sunday, AWS:reinvent 2024 will start. I couldn't join this year (US visa is getting longer to get), so please if you make it: enjoy your time there, meet new people, network with intent to actually make true connections there. And more...
Issue # 114: November 15, 2024 to November 22, 2024 Hey Reader. Welcome to Issue # 114 of AWS Graviton Weekly, which will be focused on sharing everything that happened in the past week related to AWS Silicon: from November 15, 2024 to November 22, 2024. Enjoy. Recommendation of the week: CAST.AI Save big today NEWS Amazon and Anthropic deepen strategic collaboration Amazon EC2 added New CPU-Performance Attribute for Instance Type Selection The next generation of Amazon FSx for Lustre file...