Amazon Web Services announced five new Amazon Elastic Compute Cloud (Amazon EC2) instances, two new AWS Outposts form factors, and three new AWS Local Zones locations that extend its lead in offering the broadest and deepest portfolio of compute offerings in the cloud.
AWS already has more compute instance types than any other cloud provider, with instances based on the fastest processors from Intel, cost-optimized instances with AMD processors, the most powerful GPU instances from NVIDIA, instances that feature up to 400 Gbps networking performance, and the only Arm-based instances in the cloud offering customers 40% better price performance with AWS-designed AWS Graviton2 processors.
AWS Outposts is unique in the industry as it enables customers who want to run AWS on-premises to do so with the same Application Programming Interfaces (APIs), tools, and hardware that they use to run AWS in AWS’s Regions.
AWS Local Zones gives customers who need low latency infrastructure in major metropolitan areas but don’t want to provision or maintain datacenter space in those locales the ability to use AWS in these metropolitan areas. Today’s announcement provides even more options for customers looking to choose the best compute for their unique needs.
AWS Graviton2-powered C6gn instances
New AWS Graviton2-powered, Arm-based C6gn instances deliver 100 Gbps networking performance (4x higher than C6g) and 38 Gbps EBS bandwidth (2x higher than C6g) for compute-intensive workloads. They also provide 40% better price performance over comparable current generation network-optimized x86-based instances.
Today, customers use AWS network optimized instances (e.g. C5n, M5n, R5n) for a variety of network-intensive workloads like firewall, router, and load-balancing appliances, video transcoding, and analytics that can take advantage of high networking throughput and packet rate performance.
However, because these instances have a limit in maximum packet rate, some customers need to overprovision compute resources to drive higher packet processing power to achieve the desired networking throughput.
C6gn instances provide the best packet processing performance on EC2, enabling customers to migrate to C6gn and consolidate their workloads onto fewer instances or smaller instance sizes, and reduce infrastructure costs.
With the new Amazon EC2 C6gn instances, AWS is expanding its Graviton2 portfolio to help customers meet the demands of network-intensive workloads while lowering costs.
C6gn supports Elastic Fabric Adapter (EFA), a network interface for EC2 that makes it possible for applications using popular HPC technologies like Message Passing Interface (MPI) to scale to thousands of CPU cores. C6gn instances will be available later this month in eight sizes providing up to 64 vCPUs, 100 Gbps of network bandwidth, and 38 Gbps of EBS bandwidth.
Graphics-optimized G4ad instances powered by AMD GPUs
New AMD-powered GPU instances offer the best price performance for graphics-intensive applications in the cloud. Today, customers use G4dn instances powered by NVIDIA GPUs and custom Intel CPUs for graphics-intensive applications like remote graphics workstations, virtual desktop infrastructure (VDI), video transcoding, photo-realistic design, and game streaming.
New G4ad instances feature AMD GPUs and CPUs to offer the lowest cost in the cloud for graphics intensive-applications, providing up to 45% better price performance when compared to NVIDIA GPU-based G4dn instances.
G4ad instances feature the latest AMD Radeon Pro V520 GPUs and 2nd generation AMD EPYC processors and offer up to 2.4 TB of local NVMe storage for fast data access. This enables customers to efficiently create photo-realistic and high-resolution 3D content for movies, games, and AR/VR experiences.
With access to AMD Radeon Pro Software for Enterprise at no additional cost, G4ad instances offer professional-grade graphics rendering for virtual workstations.
Customers looking to use remote workstations in the cloud for running graphics applications (e.g. ESRI ArcGIS Pro, Autodesk Revit, Autodesk Maya, or 3D Studio Max) can use G4ad instances to give them the flexibility to provision resources on a per-project basis and not be limited by availability of on-premises hardware.
G4ad instances will be available in the coming days in three sizes, with 1, 2, or 4 GPUs each.
General purpose M5zn instances powered by the fastest Intel Xeon Scalable CPUs
New M5zn instances feature the highest single-threaded performance (for applications that execute a series of individual tasks sequentially) from Intel Xeon Scalable (Cascade Lake) processors available in the cloud.
Today, customers use Amazon EC2 z1d instances when they need a high-frequency CPU (to perform single-threaded tasks more quickly and run more jobs per core) and local, low latency storage for workloads like electronic design automation (EDA) and relational database applications that require both high per-core performance and a large memory footprint.
While many customers benefit from the high compute performance of z1d instances, they do not always fully utilize the available memory and local storage, which results in overprovisioning compute and leads to higher infrastructure costs.
Customers performing complex calculations and real-time analysis for their financial, analytics, and gaming workloads have asked for an EC2 instance that balances faster compute performance, lower memory, and higher networking throughput.
M5zn instances deliver an all-core turbo frequency of up to 4.5 GHz, 4 GiB memory per vCPU, 100 Gbps of network bandwidth, and are available in seven sizes (up to 48 vCPUs and 192 GiB memory).
M5zn instances offer up to 45% better per-core performance compared to M5 instances, allowing them to right-size their instances to save on infrastructure costs and reduce per-core software licensing.
M5zn instances support Elastic Fabric Adapter (EFA), which provides low latency, high throughput networking, and the ability to scale up to tens of thousands of processor cores.
The combination of high per-core performance and network throughput make M5zn instances ideal for gaming, analytical, and simulation applications like those used by the financial, automotive, aerospace, energy, and telecommunication industries, where performance per-core is a bottleneck.
M5zn instances are available in US East (N. Virginia), US East (Ohio), US West (N. California), US West (Oregon), Europe (Ireland), Europe (Frankfurt), and Asia Pacific (Tokyo), with availability planned for additional Regions soon.
Highest capacity D3/D3en storage-optimized instances
Next-generation Intel-powered storage-optimized instances deliver the highest capacity HDD instance storage in the cloud.
Today, customers rely on Amazon EC2 to run dense storage workloads that require high throughput access to large quantities of data like massively parallel processing data warehousing (e.g. Amazon Redshift and HP Vertica), big data and analytics distributed file systems (e.g. Hadoop and MapReduce), network file systems (e.g. Lustre and Windows File System), and log or data processing applications (e.g. Kafka and Elasticsearch).
While customers running these workloads on D2 instances today enjoy the available storage per vCPU and low cost, they also want higher CPU performance and network speed to meet the increasing performance requirements of these dense storage workloads.
At the same time, customers running compute and storage clusters on-premises have expressed a need for dense storage with even more storage per vCPU at a lower cost for migrating and scaling their network and distributed file systems on AWS.
Next-generation D3 instances deliver up to 30% better processing performance, and up to 2.5x higher network performance over previous generation D2 instances.
D3 instances feature 2nd generation Intel Xeon (Cascade Lake) processors with a sustained all-core frequency of 3.1 GHz, and they offer up to 48 TB of storage, 32 vCPUs, 256 GiB of memory, and 25 Gbps of network bandwidth.
A new extended storage and high-speed networking variant, D3en instances, provides additional HDD storage capacity compared to D2 instances, offering 336 total TB of storage (7x higher than D2 instances), 75 Gbps of network bandwidth (7.5x higher than D2 instances), and up to 6.2 GiBps of disk throughput (2x higher than D2 instances).
D3en instances deliver highest capacity local HDD storage in the cloud, an up to 80% reduction in cost-per-TB compared to D2 instances, and let customers architect petabyte-scale file storage clusters so they can consolidate their high capacity big data analytical workloads.
D3/D3en instances are available in US East (N. Virginia), US East (Ohio), US West (N. California), US West (Oregon), Canada (Central), Europe (Ireland), Europe (Frankfurt), Europe (Paris), Europe (Sweden), Europe (London), Asia Pacific (Singapore), Asia Pacific (Tokyo), Asia Pacific (Sydney), Asia Pacific (Seoul), Asia Pacific (Mumbai), China (Beijing), China (Ningxia), GovCloud (US East), and GovCloud (US West).
R5b memory-optimized instances
New memory-optimized instances enhanced for Amazon Elastic Block Store (EBS) deliver up to 60 Gbps of bandwidth and 260,000 IOPS of instance-to-EBS performance for the most demanding database workloads.
Today, many customers use R5 instances for large relational database workloads like commerce platforms, ERP systems, and health record systems, and rely on EBS to deliver scalable, durable, and highly available block storage.
While R5 instances offer sufficient storage performance for many use cases, some customers have on-premises workloads that would benefit from even more EBS performance, including large commercial database workloads from SAP, Oracle, and Microsoft SQL Server that have high performance storage requirements.
These customers have to scale their R5 instances to achieve their desired EBS performance, increasing their infrastructure and database licensing costs and reducing the compute and memory utilization.
With 60 Gbps of bandwidth and 260,000 IOPS, R5b instances increase instance-to-EBS performance by 3x compared to same-sized R5 instances. This drives significantly increased performance for large database workloads that process large data sets in memory.
The increased instance-to-EBS performance of R5b instances allows customers to move their storage performance-intensive workloads like relational databases and data analytics from on-premises data centers to AWS to realize reduced costs and greater security, scalability, and reliability.
R5b instances also support EBS io2 Block Express volumes (in preview) enabling customers to have a single 99.999% (five 9s) durable EBS volume with up to 256,000 IOPS and 4,000 MB/second of throughput to further consolidate their storage intensive workloads.
Existing EC2 customers with workloads sensitive to storage performance can consolidate their workloads on fewer R5b instances, or on smaller instance sizes by migrating from R5 to R5b.
This enables them to reduce both infrastructure and licensing costs. R5b instances are supported by Amazon RDS for Oracle and Amazon RDS for SQL Server. This simplifies the migration path for running commercial database applications on AWS and improves storage performance for current RDS customers by up to 3x.
R5b instances are available in US West (Oregon), Asia Pacific (Tokyo), US East (N. Virginia), US East (Ohio), Asia Pacific (Singapore), and Europe (Frankfurt), with availability in additional regions to follow.
Smaller AWS Outposts form factors
Before today’s announcement, AWS Outposts provided customers with the ability to run their workloads on-premises and seamlessly connect with the broad array of AWS services in the cloud in a form factor that scales from one rack (roughly the size of a refrigerator) to dozens of racks, with capacity equivalent to hundreds or thousands of servers.
Some customers also want to extend the benefit of AWS Outposts to locations that may have constraints on space, power, or networking, or do not need a full rack of capacity.
For example, retail stores may need to run their sales or security systems on-premises for low latency and local network access, but still want to aggregate inventory data and analyze customer behavior using data lakes and machine learning models running in an AWS Region.
Other customers may have workloads that do not require a full rack of capacity, such as a hospital’s patient portal, a product line-monitoring workload in a factory, or a telecommunications provider’s 5G network management application.
Also, for many of these use cases, customers may need to set up servers across hundreds or thousands of remote locations, which can take weeks or months.
In all of these cases, customers then need to manage applications across servers, monitor health and performance of the applications and infrastructure, and ensure that the applications have the latest security patches, which is time-consuming and can impact application availability.
Two new smaller AWS Outposts form factors will allow customers to bring the same AWS services, infrastructure, and operating models on-premises to space-constrained locations like branch offices, factories, hospitals, cell towers, or retail stores.
AWS Outposts will now be available in two new smaller form factors – a 1U (a 1 ¾-inch tall one rack unit server) and a 2U (a 3 ½-inch tall two rack unit server) version – that require significantly less power and network connectivity than the full 42-rack unit AWS Outposts.
The 1U form factor AWS Outposts is suitable for 19-inch wide, 24-inch deep cabinets in space-constrained locations (e.g. cell sites or some retail shops) and provides 64 vCPUs, 128 GiB memory, and 4 TB of local NVMe storage.
The 2U form factor is suitable for standard 19-inch wide, 36-inch deep cabinets, and provides up to 128 vCPUs, 512 GiB memory, and 8 TB of local NVMe storage, with configurations that support accelerators like AWS Inferentia or GPUs.
Each of these smaller form factors allow customers to run AWS services like Amazon Elastic Compute Cloud (EC2), Amazon Elastic Container Service (ECS), Amazon Elastic Kubernetes Services (EKS), and Amazon Virtual Private Cloud (VPC) on-premises.
Once AWS Outposts is connected to a customer’s network, AWS will remotely manage the infrastructure just like the hardware running in an AWS Region, including pushing automated security patches and monitoring hardware health, which saves customers time and money.
As before, AWS Outposts gives customers access to the same familiar AWS APIs, control plane, tools, and hardware on-premises as in an AWS Region at virtually any location, regardless of space constraints or capacity requirements.
It also seamlessly connects back to AWS for the full array of AWS services – providing for a truly consistent hybrid experience. The smaller AWS Outposts form factors will be available in 2021.
AWS Local Zones expand to Boston, Houston, and Miami
AWS spans 77 Availability Zones within 24 geographic regions around the world, with announced plans for 15 more Availability Zones and five more AWS Regions in India, Indonesia, Japan, Spain, and Switzerland. The vast majority of customers get the necessary performance for their applications in public AWS Regions.
However, for some of the more latency-sensitive and throughput-sensitive workloads (e.g. remote real-time gaming, machine learning inference, and live video streaming), customers want AWS infrastructure closer to their end-users.
These latency-sensitive workloads have traditionally required customers to procure, operate, and maintain IT infrastructure in their own data center or co-location facility, adding cost and operational complexity.
Customers also had to build and run these low latency application components with a different set of APIs and tools than the other parts of their applications running in AWS.
Prior to today, customers could use AWS Local Zones in Los Angeles to deliver single-digit latency access to applications for end-users located in the Southern California area. However, customers outside of Southern California also want this capability for end-users in other cities across the US.
With today’s announcement, customers can now use new AWS Local Zones in Houston, Boston, and Miami to run AWS compute, storage, database, analytics, and machine learning services, and deliver applications with single-digit millisecond latencies to local end-users nearby.
With an additional 12 Local Zones launching in 2021 in Atlanta, Chicago, Dallas, Denver, Kansas City, Las Vegas, Minneapolis, New York, Philadelphia, Phoenix, Portland, and Seattle, customers will be able to deliver ultra-low latency applications to end-users in cities across the US.
AWS Local Zones are managed and supported by AWS, meaning customers no longer need to incur the expense or effort of procuring, operating, and maintaining data centers or co-location facilities in various cities to support ultra-low latency applications.
AWS Local Zones provide customers a high-bandwidth, secure connection between their local workloads and those running in the closest AWS Region. This gives customers the ability to use the same AWS APIs and tools to run latency-sensitive workloads nearby to end-users, while seamlessly connecting to the full range of services in the AWS Region.
“Whether your organization has been all-in on the cloud since day one or is just beginning to move workloads to the cloud, all customers want to optimize for price performance,” said Dave Brown, Vice President, EC2, AWS.
“As customers bring more and more workloads to the cloud, AWS continues to expand the industry’s leading compute portfolio to meet their increasingly diverse needs. With the new EC2 instance types, AWS Outposts and AWS Local Zones options we’re introducing today, we’re providing customers with an unmatched breadth and depth of capabilities to help them innovate more cost-effectively, with the right compute for the right job.”
Datadog, Inc. is the monitoring and security platform for cloud applications. “We’re excited about the M5zn instances and the impact they can have on our high throughput compute and network intensive applications,” said Rob Boll, Director, Runtime Platforms, at Datadog.
“Many of our compute intensive workloads will benefit from the improved performance per core, and with 100 Gbps networking throughput, we’ll be able to pack workloads more efficiently onto fewer nodes. We expect to see significant infrastructure cost savings as we move our larger workloads over to M5zn instances.”
Ubitus is a cloud gaming technology leader. Through their platforms, users can enjoy a AAA gaming experience on any device including smartphones, tablets, gaming consoles, smart TVs, and computers as long as they’re connected to a broadband network.
“We used AWS to partner with 505 Games to bring Control Ultimate Edition to a highly popular portable gaming device,” said Wesley Kuo, CEO of Ubitus. “With the new EC2 G4ad instances we have a lower cost structure, which will enable us to bring more games like Control Ultimate Edition to the mobile users globally.”
Netflix is the world’s leading streaming entertainment service with over 190 million paid memberships in over 190 countries enjoying TV series, documentaries, and feature films across a wide variety of genres and languages.
“We have many artists and content creators who can benefit from bringing the on-premises infrastructure they use day-to-day to the cloud. However, every second counts, especially when it comes to the time it takes to compile and render 3D models and animations.
“For some time, we have wanted to move these artists’ workstations to AWS in order to take advantage of the performance, elasticity, and cost benefits of the cloud. However, due to the interactive nature of these applications, our artists need very low latency access to their workstations to have a good working experience,” said Nils Pommerien, Director, Cloud Infrastructure Engineering, Netflix.
“AWS Local Zones, which bring cloud resources closer to our artists, have been a game-changer for these applications. By taking advantage of access to AWS’s highly performant and cost-effective compute resources, we have been able to migrate portions of our content creation process to AWS Local Zones, while ensuring an even better experience for artists.”
Mindbody is the leading technology platform for the fitness, wellness, and beauty industries. “We have a portfolio of interdependent applications running in our existing on-premises data centers. We have been looking to move these workloads to the public cloud for some time. However, it is daunting to migrate such complex, interdependent applications to the cloud at the same time while ensuring a seamless experience for our end users,” said John Strong, Senior Director of Production Engineering, Mindbody.
“AWS Local Zones have solved a significant problem for us here. We are using Local Zones to migrate our complex, legacy on-premises applications to AWS without an expensive revamp of our architecture.
“With a Direct Connect to Local Zones, we are able to establish a hybrid environment that provides ultra-low latency communication between applications running in the Local Zone and our on-premises installations. In turn, this has enabled us to migrate applications incrementally, simplifying our migrations drastically while reducing any business risk with on-going hybrid deployments.”