Oracle Cloud Infrastructure (OCI) today launched a multi-prong Arm initiative including instances (VM and bare metal) based on Ampere’s Altra microprocessor, and a three-tier Arm developer program seeking, among other things, to woo the huge base of mobile and IoT Arm-based developers to the cloud application space.
The announcements are being made at an Oracle Live (livestream) event which unambiguously promoted Arm: “The cloud was built on x86 processors, but the promise of Arm-based cloud computing—delivering linear scalability and price-performance advantages—is a future too bright to ignore. Oracle is committed to the growth of the Arm ecosystem for cloud computing and ensuring all customers are positioned to realize the benefits.”
Lately the proliferation of processors of all types (CPUs, GPUs, FPGAs, TPUs, specialized AI devices, etc.) has injected fresh energy into the processor landscape and cloud providers have played a key role in mainstreaming many options; AWS’s home-grown Graviton processor, based on Arm, is a good example. Widespread use of AMD’s Epyc x86 line, including by Oracle as far back as 2018, is another example.
The new Arm instances (OCI Ampere A1 Compute) are being offered at a penny per core hour, which Oracle says is the lowest in industry. OCI is offering flexible VM sizing from 1-to-80 OCPUs and 1-to-64 GB of memory per core or as a bare-metal service with 160 cores and 1 TB of memory. “Customers can deploy Arm-optimized applications on containers, bare metal servers, and virtual machines in the Oracle public cloud, or Dedicated Region Cloud@Customer, according to Oracle,” according to the official announcement.
In a pre-briefing with HPCwire, Bev Crair, SVP, OCI Compute, said “Ten years ago, I think the developer ecosystem wasn’t there. The tools weren’t there, [and] it was a little too scary. Now we’re seeing an inflection point in industry, where the number of different architectures that are available, particularly with the advent of the GPU, and the role that the GPU plays in AI and ML are offering. Having a compute infrastructure that can play in that same space from an AI ML perspective – not necessarily competing with the GPU, but doing something different, and additive – is an important part of the set of solutions that we need to make available to our customer base.”
“If you look at what’s happened with developers, the number of developers who are working in Arm now on mobile and developing on Arm in IoT is very large. In some sense, [our Arm initiative] is in part aimed at getting access to that community, right and enabling that community to consider not just a mobile device or a mobile solution or mobile application, but a server-side, cloud-side application solution. It’s also what Ampere is delivering. That system has really predictable performance, on the single-threaded core design that we’re seeing. As you increase the number of cores you’re using, you get linear scalability that we don’t necessarily see across all of our other shapes. And performance characteristics in specific workloads also matter; some of our benchmarking [shows] 30 percent better performance on our A1 shape versus the same workload on any of our other shapes,” said Crair.
OCI touts its new Arm developer program saying it provides “developers with more choice in compute instances and the best price-performance, compared to any other x86 instance on a per core basis.” The three-tier offering includes:
- Oracle Cloud Free Tier in which developers receive $300 in credits for 30 days.
- Always Free Arm which provides access to four A1 cores and 24 GB memory for an unlimited time.
- Arm Accelerator Program, which “provides open-source developers, ISV partners, customers and universities with Arm-based development projects that need more resources beyond what the Oracle Cloud Free Tier provides.” Developers must apply to this program to receive Oracle Cloud credits for a 12-month period.
Crair said, “We’ve created the new offerings not only to ensure that everyone has access to Arm, but also to make it easier for developers to get started on Arm on OCI. We’re building what we’re calling the Arm Accelerator Program, which provides users and developers who have an interest in moving the Arm ecosystem forward with the opportunity to come to us and say, ‘Hey, I’ve got this thing in mind, your free tier stuff doesn’t cut it completely because it’s more limited than I want and I’d like your help to do this for a longer period of time.”
In preparation, Oracle has assembled a stack and set of tools in effort to attract Arm developers. Here’s a brief description excerpted from the announcement:
“Oracle’s development stack is available on Arm-based A1 instances, including Oracle Linux, Java, MySQL, GraalVM, and the Oracle Container Engine for Kubernetes (OKE) service. To make it easy for developers to get started, Oracle created an Oracle Linux Cloud Developer image which enables customers to install, configure, and launch a development environment that includes OCI client tools, utilities, and common programming languages such as Java, GraalVM, Python, PHP, Node.js, Go and C/C++. The developer image is easily accessible and can be deployed from the OCI console.
“To help customers take advantage of the latest in Arm technology, Oracle is working closely with a wide variety of technology and open source partners, such as GitLab, Jenkins, Rancher, Datadog, OnSpecta, NGINX, and Genymobile. To help grow and enrich the Arm developer ecosystem, Oracle also announced that it is joining the Continuous Delivery Foundation (CDF), an open source, vendor-neutral community for sustaining the fastest growing CI/CD open source projects.”
Oracle is positioning the A1 instances for a variety of uses. Talking about HPC, Crair said, “We actually already have as part of our limited availability, a number of universities who are working HPC workloads on this environment, and doing the necessary work with us on tuning characteristics, to determine the places where this is going to work best.”
Simon McIntosh-Smith, head of the HPC Research Group, Department of Computer Science, University of Bristol is quoted in the announcement, “In our research, we need quick access to the latest computing technologies to help us solve complex scientific problems at breakneck speed. With Oracle Cloud Infrastructure, we have gained early access to the latest Arm-based Ampere A1 Compute to get the superior performance and power efficiencies needed to scale our high-performance computing workloads instantly. In terms of performance per dollar and in terms of cost, it gets even more compelling.”
(Overview: $0.01 per core per hour and $0.0015 per GB of RAM per hour. Ampere’s choice of using a single threaded core, plus sustained 3.0Ghz maximum frequency, results in linear scaling with respect to the cores. Ampere’s Altra processors can run all cores at the maximum frequency. In addition, the cores are completely isolated from the noisy neighbor impact of other workloads running on the same processor. Each core is single threaded by design with its own 64 KB L1 I-cache, 64 KB L1 D-cache and a huge one MB L2 D-cache.)
OCI describes A1 use cases as follows:
- General Purpose: The OCI Ampere A1 Compute provides “superior price-performance for general purpose workloads, such as web servers, application servers and containers. These shapes offer balanced performance and an optimal price point for cloud-based scale-out workloads, such as NGINX and web applications.”
- In-memory Caches and Databases: From databases to analytics, “Arm processors deliver predictable performance for databases, such as Redis and MySQL. Memory- heavy workloads and multithreaded applications, such as in-memory databases and key-value stores, experience superior performance.”
- Mobile Application Development: “Ampere Altra’s high core count (up to 160) is ideal for the density and scale needed for mobile application development and testing. In addition, developing iOS or Android-based applications on the OCI Ampere A1 Compute eliminates the need for an emulator or nested virtualization, leading to superior performance.”
- Computationally-intensive and Scientific Applications: Arm processors “provide the price-performance benefits that make it a commonly used platform for high- performance, compute-intensive and scientific applications such as AI/ML inferencing, media transcoding, and running HPC stacks like CFD, WRF, OPENFAM, GROMACS, BLAST, BeeGFS, and NAMD.”
Price-performance is clearly a dominant theme in Oracle’s early foray into Arm – not unlike the approach AMD took when reentering the datacenter and cloud server market. Nvidia, of course has also announced an Arm-based CPU (Grace). Crair declined to comment on whether or not Oracle, which was among the first hyperscalers to offer A100 (GPU) instances, planned new instances pairing Nvidia GPUs and Arm CPUs. We’ll see.
IDC was upbeat about the new Oracle offering. “As the Arm ecosystem continues to develop, the breadth of deployments and use cases will only continue to grow with it, enabling further adoption of Arm-based infrastructure. The A1 instance is a great first step for OCI towards bringing that future forward,” said Kuba Stolarski, research director, infrastructure systems, platforms and technologies group, IDC.
Oracle also provided a number of customer/partner testimonials in the official announcement (link below). The new offerings are availability starting today according to Crair.