As Alibaba Cloud's flagship product, Elastic Compute Service (ECS) can help you build your services on the cloud. Alibaba Cloud ECS provides a high-performance and highly scalable computing platform for your services to grow, with much more powerful computing capabilities than what you'll find in a typical on-premises setup.
In this article, we're going to show you some of the featured technologies and advantages of the newest generation of this cornerstone product, which is already in its sixth generation. As a quick summary of what we'll cover in this article, this newest generation of Alibaba Cloud ECS instances offer significant performance improvements, better stability, as well as higher availability. This is in part due to new processors that have significantly optimized and customized for better performance and the newest generation of Alibaba's proprietary X-Dragon architecture. Read on to learn more details about this newest generation.
The sixth generation of Alibaba Cloud ECS instances offers great improvements in terms of CPU computing power, memory read and write performance, as well as scenario adaptability, network throughput, and storage, having rather noticeable improvements in all of these areas over the previous generation.
A Comparison of Sixth of Fifth Generation ECS Instances
These improvements in performance relate to several changes in the overall architecture and feature set of Alibaba Cloud's sixth-generation ECS instances. In particular, this newest generation of cloud servers uses customized Intel processors, the newest generation of Alibaba Cloud's proprietary X-Dragon architecture, as well as the network management system build in MoC and optional ESSD storage.
All sixth-generation ECS instance come equipped with Intel® Xeon® Platinum 8269CY processors. These processors were customized based on the Cascade Lake microarchitecture, which is designed for the second-generation Intel® Xeon® Scalable processors. These processors have a turbo boost with an increased burst frequency of 3.2 GHz, and can provide up to a 30% increase in floating performance over the fifth generation ECS instances.
Compared with the previous Intel® Xeon® Platinum 8163 processors that used the Skylake micro-architecture, 8269CY processors using the Cascade Lake microarchitecture offer the several notable benefits. First, the processors feature lower power consumption, higher memory bandwidth and lower memory latency. Next, they also offer high performance thanks to turbo speeds featuring in all instance cores, and increased deep learning inference and reasoning capabilities by a factor of 11.
Sixth-generation ECS instances are built around Alibaba's very own X-Dragon architecture. Among other things, the newest generation of this architecture helps improve performance by offloading a large number of virtualization operations to dedicated hosts, thereby significantly reducing overhead. And, through this and other optimizations, this newest generation of cloud servers opens up a new frontier in CPU and memory resource utilization, improving the overall capabilities of hosts and high-specification instances.
The Third-Generation of Alibaba's X-Dragon Architecture Visualized
This next-generation of ECS and the latest generation of ECS bare-metal instances are all based on Alibaba Cloud's proprietary X-Dragon computing platform. The X-Dragon computing platform consists of the MoC NIC, the X-Dragon software system, and X-Dragon Hypervisor. Through this architecture, the management, network virtualization and storage virtualization software that run on physical servers in the traditional KVM virtualization solution can be offloaded onto the MoC NIC, making the servers fully accessible to users all the time.
Supreme performance from optimized architecture
The core of the MoC NIC is the X-Dragon chip. The X-Dragon software system runs on the X-Dragon chip to provide virtual private cloud (VPC) and EBS disk capabilities. It offers these capabilities to ECS instances and ECS Bare Metal Instance through VirtIO-net and VirtIO-blk standard interfaces.
Compared with the previous generation, the newest generation of ECS instances offload Alibaba Cloud's proprietary VSwitches from physical servers to the MoC NIC for the first time. Following this, virtual private cloud (VPC) and Server Load Balancer (SLB) capabilities are provided. When a physical server is running, the DPDK loop process of the corresponding VSwitch needs to run on a single CPU node.
Network Management System
As a result, ECS instances that run on a physical server may exhibit differences in terms of such capabilities as packet forwarding rates and network latency. Meanwhile, the CPU turbo frequency and memory access bandwidth of your ECS instances on the same CPU node as the DPDK loop process are also affected. After the switch to the MoC NIC, this process no longer interferes with your ECS instances. In addition, Cloud Network Management and the MoC NIC have been comprehensively optimized to provide better network performance for your ECS instances.
Now, with the newest generation of ECS instances, in addition to the typical SSDs, Alibaba Cloud customers can opt for ultra-stable and ultra-high-performance ESSDs. ESSDs are a good option, especially for enterprise-level customers, as they can reduce the read and write latency of a disk by up to 70% and support a maximum single-disk IOPS of one million. Moreover, compared with SSD storage, the ESSD storage option has a 60% capacity increase.
In an Alibaba Cloud ECS instance, the ESSD storage consists of a VirtIO-blk frontend driver, a Kunpeng backend driver, and an Alibaba exclusive Apsara Distributed File System cluster. The VirtIO-blk frontend driver runs on your ECS instance or ECS Bare Metal Instance, the Kunpeng backend driver runs in the MoC NIC, and the Apsara File System cluster is deployed separately.
ESSDs offer the multi-queue capability of the block layer of Linux kernel in the VirtIO-blk frontend driver. An ESSD can provide multiple hardware queues, which significantly improves the performance of concurrent access from multiple I/O processes to this ESSD. The ESSD Kunpeng backend driver uses multiple technologies to improve storage performance. As one example, it uses the SPDK framework to interact with NVMe SSDs and adopts a user-state driver and multi-queue adaptive polling mode. Similarly, the Apsara Distributed File System clusters were also upgraded to improve storage performance. For example, the related storage servers and NVMe SSDs and now use two 25 Gbps NICs, and the high-performance Remote Direct Memory Access (RDMA) protocol is used to reduce end-to-end communication latency across VSwitches to a mere two microseconds.
Over the past several years, Alibaba Cloud Elastic Compute Service has maintained a leading position in the market. As the backbone of this service, Alibaba Cloud's exclusive X-Dragon computing platform, which is a one-of-a-kind system, consists of the proprietary X-Dragon chip, MoC NIC, lightweight X-Dragon Hypervisor, and X-Dragon server. With all of these pieces in place, you can expect world-class elasticity and high-performance. The X-Dragon computing platform helps make zero resource consumption and performance loss for Alibaba Clod ECS virtual machines and ECS Bare Metal instances possible.
Alibaba Cloud's virtualization team has improved the computing stability of the sixth-generation instances to 99.9999% based on the X-Dragon computing platform. Even when a ping flood test is performed on a sixth-generation instance in a load-free environment, only one packet among one million messages experiences latency more than one millisecond. This is world leading stability and service availability.
Much of this world-leading technology comes from the e-commerce cloud infrastructure expertise of Alibaba Group. At Alibaba, we need to be able to predict performance in terms of several different dimensions while dealing with peak traffic spikes seen during the Double 11 shopping event, which is China's and arguably the world's largest online shopping promotion. We strive to enable performance prediction for instances, networks, and storage.
Given this experience, sixth-generation ECS instances provide instance-level QoS for network links to prevent session conflicts. Instance-level QoS is also provided for disk I/O to ensure that each instance has predictable disk throughput, bandwidth, and IOPS. Alibaba Cloud was the first among large-scale cloud vendors in China to provide instance-level QoS assurance for network links and disk I/O. In other words, this is in many ways an industry first.
At Alibaba, we have also developed powerful delivery capabilities and elasticity for instances. Our instances can handle up to 540,000 orders per second during the big Double 11 promotion, starting up from common images within only 22 seconds, and deliver a maximum of 160,000 vCPUs within five minutes per zone per user. As such, six-generation ECS instances can meet users' requirements for the ultimate level of elasticity and deliverability on the cloud.
Alibaba Cloud's six generation of ECS provides industry-leading stability backed by powerful engineering capabilities. Compared with typical on-premises data centers and your run-of-the-mill server vendors, Alibaba Cloud is exceptional in that it adopts much more stringent standards for its data centers, as well as server access, and O&M practices so to ensure superior data reliability and high availability. This is especially true for Alibaba Cloud's cornerstone product of ECS.
At Alibaba Cloud, we have accumulated much experience in operating data centers and managing large-scale IT infrastructure. With over 10 years of production experience, we have analyzed massive amounts of data to identify the causes of infrastructure faults and explore ways of predicting and preventing such faults. In collaboration with the scientists of the Alibaba DAMO Academy , we process and model the data to predict faults more accurately. The accuracy of our predictions exceed 99%, effectively ensuring our ultra-high availability. Next, the service-level agreement (SLA) that we offer for a single ECS instance is now up to 99.975%. This a world first, you could say.
As part of Alibaba Cloud's Elastic Compute Service, disks offer block-level data storage, which features low latency as well as high performance, persistence, and reliability. The disks of Alibaba Cloud ECS use triplicate technology to provide data reliability up to 99.9999999%, and they also automatically copy your data within a zone to ensure data availability and prevent service interruptions in the case of hardware faults. Also, with these ECS disks, like how you would with a hard disk, you can partition and format the disk attached to your ECS instance, create a file system, and persistently store data in it.
Alibaba Cloud has passed a host of international information security certifications, such as ISO 27001 and Multi-Tier Cloud Security (MTCS), which demand strict confidentiality of user data and user information, as well as high user privacy protection.
But besides this, Alibaba Cloud also has a rich service and product portfolio that offers additional security features to protect your data running on your six-generation ECS instance. To ensure the utmost in security, for example, you can use your ECS instances in a VPC.
Alibaba Cloud's virtual private cloud (VPC) provides a stable, secure, and controllable network environment that enables quick delivery. VPC features traffic isolation and attack isolation to protect your business from cyber-attacks. After you build your business in a VPC, the VPC establishes the first line of defense for your business.
With the latest generation of Alibaba Cloud's Elastic Compute Service, Alibaba Cloud is capable of creating the IT resources required by almost any medium-sized Internet enterprise within just a few minutes, and therefore can easily provide the scalability and elasticity required for any business or large-scale enterprise around the clock.
And, besides elastic computing, Alibaba Cloud offers a full portfolio of cloud services including storage and networking services, and even the planning and drafting of business architecture on the cloud, allowing you to fully equipped on the cloud.
In terms of computing, each Alibaba Cloud region has multiple zones, and you can create and deploy active/standby or active/active ECS instances in multiple zones to achieve higher availability of services. You can build fault-tolerant systems across multiple regions and zones to implement a finance-oriented solution that spans three data centers across two regions. Alibaba Cloud provides mature solutions for fault tolerant services such as disaster recovery.
With Alibaba Cloud ECS, it is easy to achieve both vertical and horizontal scaling. With vertical scaling, you can modify the configurations of an ECS instance according to your business volume in relative real time. And, with horizontal scaling, you can quickly achieve resource re-allocation among different applications. That is, with the elasticity offered by Alibaba Cloud's ECS instances, you can easily provide the resources required for peak hours, coping with any peak loads, and you can also scale back as these peaks subside, saving your overall operation costs. In other words, above anything else, Alibaba Cloud ECS can provide the scalability, elasticity and cost effectiveness your business requires.
Alibaba Cloud ECS - January 8, 2020
Alibaba Clouder - May 22, 2018
Alibaba Clouder - October 28, 2019
Alibaba Clouder - February 12, 2020
ApsaraDB - November 28, 2019
Alibaba Developer - January 9, 2020
Quickly Improves High Availability For ApplicationsLearn More
A HPCaaS cloud platform providing an all-in-one high-performance public computing serviceLearn More
An online computing service that offers elastic and secure virtual cloud servers to cater all your cloud hosting needs.Learn More
More Posts by Alibaba Cloud ECS