Professional Documents
Culture Documents
Ecs Productdesc
Ecs Productdesc
Ecs Productdesc
Service Overview
Issue 19
Date 2019-04-25
and other Huawei trademarks are trademarks of Huawei Technologies Co., Ltd.
All other trademarks and trade names mentioned in this document are the property of their respective
holders.
Notice
The purchased products, services and features are stipulated by the contract made between Huawei and the
customer. All or part of the products, services and features described in this document may not be within the
purchase scope or the usage scope. Unless otherwise specified in the contract, all statements, information,
and recommendations in this document are provided "AS IS" without warranties, guarantees or
representations of any kind, either express or implied.
The information in this document is subject to change without notice. Every effort has been made in the
preparation of this document to ensure accuracy of the contents, but all statements, information, and
recommendations in this document do not constitute a warranty of any kind, express or implied.
Website: http://www.huawei.com
Email: support@huawei.com
Contents
1 ECS................................................................................................................................................... 1
2 Regions and AZs............................................................................................................................3
3 Storage............................................................................................................................................. 4
4 Network and Security................................................................................................................... 6
4.1 VPC................................................................................................................................................................................ 6
4.2 User Encryption.............................................................................................................................................................. 6
4.3 Cloud-Init........................................................................................................................................................................8
4.4 Project and Enterprise Project........................................................................................................................................ 9
5 IMS................................................................................................................................................. 11
6 ECS Instances............................................................................................................................... 13
6.1 Lifecycle....................................................................................................................................................................... 13
6.2 Spot ECSs..................................................................................................................................................................... 15
9 Change History............................................................................................................................ 85
1 ECS
An Elastic Cloud Server (ECS) is a computing server consisting of vCPUs, memory, image,
and Elastic Volume Service (EVS) disks that allow on-demand allocation and elastic scaling.
ECSs integrate Virtual Private Cloud (VPC), virtual firewalls, and multi-data-copy
capabilities to create an efficient, reliable, and secure computing environment. This ensures
the stable and uninterrupted operation of services. After creating an ECS, you can use it like
using your local computer or physical server.
ECSs support self-service creation, modification, and operation. You can create ECSs by
specifying the vCPU, memory, image specifications, and login authentication. After creating
an ECS, you can modify its specifications as required.
System Architecture
ECS works with other products and services to provide computing, storage, network, and
image installation functions.
l ECSs are deployed in multiple availability zones (AZs) connected with each other
through an internal network. If an AZ becomes faulty, other AZs in the same region will
not be affected.
l With the Virtual Private Cloud (VPC) service, you can build a dedicated network, set the
subnet and security group, and allow the VPC to communicate with the external network
through an EIP (bandwidth support required).
l With the Image Management Service (IMS), you can install images on ECSs, or create
ECSs using private images and deploy services quickly.
l The Elastic Volume Service (EVS) provides storage and Volume Backup Service (VBS)
provides data backup and recovery functions.
A region is a geographic area where resources used by your ECSs are located.
ECSs in the same region can communicate with each other over an intranet, but ECSs in
different regions cannot.
Public cloud data centers are deployed worldwide. Creating ECSs in different regions can
better suit certain user requirements. For example, applications can be designed to meet user
requirements in specific regions or comply with local laws or regulations. ECS pricing also
changes based on region.
Each region contains many AZs where power and networks are physically isolated. AZs in the
same region can communicate with each other over an intranet. Each AZ provides cost-
effective and low-latency network connections that are unaffected by faults that may occur in
other AZs.
3 Storage
l Common I/O: EVS disks of this type deliver a maximum of 2,200 IOPS. This disk type
is suitable for application scenarios that require large capacity, a medium read/write rate,
and fewer transactions, such as enterprise office applications and small-scale testing.
l High I/O: EVS disks of this type deliver a maximum of 5,000 IOPS and a minimum of 1
ms read/write latency. This disk type is designed to meet the needs of mainstream high-
performance, high-reliability application scenarios, such as enterprise applications, large-
scale development and testing, and web server logs.
l Ultra-high I/O: EVS disks of this type deliver a maximum of 33,000 IOPS and a
minimum of 1 ms read/write latency. This disk type is excellent for ultra-high I/O, ultra-
high bandwidth, and read/write-intensive application scenarios, such as distributed file
systems in HPC scenarios or NoSQL/RDS in I/O-intensive scenarios.
l Ultra-high I/O (latency-optimized): EVS disks of this type deliver a maximum of 1 GB/s
throughput and a minimum of 1 ms read/write latency. They can be used for key
enterprise services, such as SAP HANA.
NOTE
Ultra-high I/O (latency-optimized) EVS disks can be attached to SAP HANA ECSs only.
EVS disks with different I/O capacities provide different features at different prices. Choose
EVS disks based on your requirements. For more information about EVS disk specifications
and performance, see Elastic Volume Service User Guide.
directly access underlying storage media. SCSI EVS disks support both basic and
advanced SCSI commands.
NOTE
For more information about how to use SCSI EVS disks, for example, how to install the driver, see
section Do I Need to Install a Driver for SCSI EVS Disks?
4.1 VPC
VPC allows you to create customized virtual networks in your logically isolated AZ. Such
networks are dedicated zones that are logically isolated for your ECSs. You can define
security groups, virtual private networks (VPNs), IP address segments, and bandwidth for a
VPC. This facilitates internal network configuration and management as well as secure and
convenient network modification. You can also customize the ECS access rules within a
security group and between security groups to strengthen ECS security protection. For more
information, see Virtual Private Cloud User Guide.
Image Encryption
Key encryption supports encrypting private images. When creating an ECS, if you select an
encrypted image, the system disk of the created ECS automatically has encryption enabled,
implementing system disk encryption and improving data security.
For more information about image encryption, see Image Management Service User Guide.
l System disk encryption relies on the image. When creating an ECS, if you select an
encrypted image, the system disk of the created ECS automatically has encryption
enabled, and the encryption mode complies with the image encryption mode.
For more information about EVS disk encryption, see Elastic Volume Service User Guide.
Impact on AS
If you use an encrypted ECS to create an Auto Scaling (AS) configuration, the encryption
mode of the created AS configuration complies with the ECS encryption mode.
About Keys
The key required for encryption relies on Data Encryption Workshop (DEW). DEW uses a
data encryption key (DEK) to encrypt data and a customer master key (CMK) to encrypt the
DEK.
Table 4-1 describes the keys involved in the data encryption process.
DEK An encryption key that is used for encrypting Encrypts specific data.
data.
CMK An encryption key created using DEW for Supports CMK disabling
encrypting DEKs. and scheduled deletion.
A CMK can encrypt multiple DEKs.
NOTE
After disabling a CMK or scheduling the deletion of a CMK takes effect, the EVS disk encrypted using
this CMK can still be used until the disk is detached from and then attached to an ECS again. During
this process, the disk fails to be attached to the ECS because the CMK cannot be obtained. Therefore,
the EVS disk becomes unavailable.
For details about DEW, see Data Encryption Workshop User Guide.
4.3 Cloud-Init
Cloud-Init is an open-source cloud initialization program, which initializes specified
customized configurations, such as the hostname, key pair, and user data, of a newly created
ECS.
Using Cloud-Init to initialize your ECSs will affect your ECS, IMS, and AS services.
Impact on IMS
To ensure that ECSs created using private images support customized configurations, you
must install Cloud-Init or Cloudbase-Init before creating private images.
Impact on ECS
l When creating an ECS, if the selected image supports Cloud-Init, you can use user data
injection to inject customized configuration, such as ECS login password, for
initializing.
l After Cloud-Init is supported, ECS login modes will be affected.
l After Cloud-Init is supported, you can view and use metadata to configure and manage
running ECSs.
Impact on AS
l When creating an AS configuration, you can use user data injection to specify ECS
configurations for initialization. If the AS configuration has taken effect in an AS group,
the ECSs newly created in the AS group will automatically initialize their configurations.
l For an existing AS configuration, if its private image does not have Cloud-Init or
Cloudbase-Init installed, the login mode of the ECSs created in the AS group where the
AS configuration takes effect will be affected. To resolve this issue, see section "How
Does Cloud-Init Influence the AS Service?" in Auto Scaling User Guide.
For more information about AS, see Auto Scaling User Guide.
Notes
l When using Cloud-Init, enable DHCP in the VPC to which the ECS belongs.
l When using Cloud-Init, ensure that security group rules in the outbound direction meet
the following requirements:
– Protocol: TCP
– Port Range: 80
– Remote End: 169.254.0.0/16
NOTE
If you use the default security group rules in the outbound direction, the preceding requirements
are met, and the metadata can be accessed. Default security group rules in the outbound direction
are as follows:
l Protocol: ANY
l Port Range: ANY
l Remote End: 0.0.0.0/16
l Enterprise project
Enterprise projects are upgraded based on IAM projects and used to categorize and
manage resources of different projects of an enterprise. An enterprise project can contain
resources of multiple regions, and resources can be added to or removed from enterprise
projects. If you have enabled enterprise management, you cannot create an IAM project
and can only manage existing projects. In the future, IAM projects will be replaced by
enterprise projects, which are more flexible.
Both projects and enterprise projects can be managed by one or more user groups. Users who
manage enterprise projects belong to user groups. After a policy is granted to a user group,
users in the group can obtain the rights defined in the policy in the project or enterprise
project.
For details about how to create a project, create an enterprise project, and grant policies, see
Managing Projects and Enterprise Projects.
5 IMS
Introduction
IMS allows you to create ECSs using images. An image is an ECS template that contains an
OS and may also contain proprietary software and application software, such as database
software.
Images can be public, private, or shared. Public images are provided by the system by default,
private images are manually created, and shared images are private images that are shared by
another user. You can use any type of image to create an ECS. You can also create a private
image using an existing ECS. This provides you with a simple way to create ECSs that
comply with your service requirements. For example, if you use web services, your image can
contain a web server, static configurations, and dynamic page code. When you use this image
to create an ECS, your web server and applications will be available for use immediately.
Image Types
Image Description
Type
Public image A standard, widely used image. It contains an OS, comes with preinstalled
public applications, and is available to all users.
Private An image that is available only to the user who created it based on an ECS
image or EVS backup (system disk backup). Such a private image contains an OS,
preinstalled public applications, and the user's private applications.
A private image can be a system image or a data image.
l System image: contains an OS, comes with preinstalled application
software for running services. You can use a system image to create
ECSs and migrate your services to the cloud.
l Data image: contains only your service data. You can use a data image
to create EVS disks and migrate your service data to the cloud.
Image Description
Type
Marketplace A third-party image that has the OS, application environment, and software
image pre-installed. You can use the images to deploy websites and application
development environments with a few clicks. No additional configuration
operation is required.
6 ECS Instances
6.1 Lifecycle
A lifecycle indicates the ECS statuses recorded from the time when the ECS is created
through the time when the ECS is deleted or released.
Creating Intermediate The ECS has been created but is not BUILD or
running. BUILDING
In spot billing mode, you can purchase and use ECSs at a discount price. The performance of
the ECSs is the same as that of ECSs with the same specifications in other billing modes.
However, when inventory resources are insufficient, or the market price increases and exceeds
your expected price, the system will automatically release your ECS resources and reclaim the
ECSs. Compared with pay-per-use and yearly/monthly ECSs, spot ECSs offer the same level
of performance while at lower costs.
Working Rules
The market price for the ECSs of a certain flavor fluctuates due to supply-and-demand
changes. You can purchase and use spot ECSs at a low market price to reduce computing
costs.
When purchasing a spot ECS, you are required to set the highest price for a specified flavor
you are willing to pay for. This process is also known as price quoting. A higher quoted price
makes it more likely for you to purchase such an ECS.
l If the quotation is greater than or equal to the market price and the inventory resources
are sufficient, the spot ECS is purchased. The ECS is billed at the market price.
l If the quotation is less than the market price, the spot ECS cannot be purchased.
After purchasing a spot ECS, you can use it like using the ECSs in other billing modes.
However, the system will periodically compare your quotation with the market price and
check the inventory resources.
l If the quotation is greater than or equal to the market price and the inventory resources
are sufficient, you can continue using the ECS.
l If the quotation is less than the market price or the inventory resources are insufficient,
the system notifies you of releasing the ECS resources if notification is enabled and
automatically deletes the ECS in about 5 minutes.
Application Scenarios
l What are supported?
Spot ECSs are suitable for image rendering, stateless web service, gene sequencing,
offline analysis, function calculation, batch calculation, sample analysis, CI/CD, and test.
NOTE
When the market price is higher than your quotation or the inventory resources are insufficient, the
spot ECSs will be reclaimed. Therefore, back up data when using such ECSs.
l What are not supported?
To prevent ECS reclamation from interrupting services, do not use spot ECSs to run the
services requiring long-time operations or high stability.
Notes
l Only KVM ECSs support spot pricing payments. For details about the ECS flavors, see
the information displayed on the management console.
l The market prices of the ECSs of the same flavor may vary depending on AZs.
l Spot ECSs do not support OS changing.
l Spot ECSs do not support automatic recovery.
l Spot ECSs do not support specifications modification.
l Spot ECSs cannot be created using a Marketplace image.
l Spot ECSs cannot be switched to yearly/monthly ECSs.
l When a spot ECS is being reclaimed,
– It cannot be used to create system disk images and full-ECS images. However, data
disks of the ECS can be used to create data disk images.
– It cannot be deleted.
Billing Rules
For details, see section Billing Modes.
Reclaiming an ECS
HUAWEI CLOUD may reclaim and terminate your spot ECS at any time. A spot ECS that is
being reclaimed cannot be used to create images.
The reclamation may be due to:
l Higher market price than your quotation
l Insufficient inventory resources
NOTE
l If a spot ECS is reclaimed within the first hour after it is provisioned due to insufficient
inventory resources or lower quotation than the market price, the spot ECS is not billed.
l In the first settlement period (in hours) of a spot ECS, the spot ECS is billed, regardless of
whether it runs.
l The time required for reclaiming a spot ECS is 5 minutes. During the reclaiming, if the sharp
clock is exceeded, the spot ECS is billed at the market price for the time after the sharp clock.
l During the running of a spot ECS, its price is updated once an hour. After a spot ECS is
restarted, or it is stopped and then started, it is billed at the market price when the ECS starts.
Therefore, back up your data. Before reclaiming your spot ECS, the system will notify you of
resource releasing if notification is enabled. Use either of the following methods to enable
notification:
l Method 1: Use the spot metadata. For details, see section Managing ECS Metadata.
l Method 2: Use Cloud Trace Service (CTS) and Simple Message Notification (SMN)
provided on the public cloud. For details, see section Purchasing a Spot ECS.
FAQs
See section FAQs.
l A specifies the ECS type. For example, s indicates a general computing ECS, c a
computing ECS, and m a memory-optimized ECS.
l B specifies the type ID. For example, the 1 in s1 indicates a general computing first-
generation ECS, and the 2 in s2 indicates a general computing second-generation ECS.
l C specifies the flavor size, such as medium, large, or xlarge.
l D specifies the ratio of vCPUs to memory expressed in a digit. For example, value 4
indicates that the ratio of vCPUs to memory is 4.
Network Bandwidth
The intranet bandwidth and PPS of an ECS are determined based on ECS specifications.
l Assured intranet bandwidth: indicates the assured ECS bandwidth.
l Maximum intranet bandwidth: indicates the maximum ECS bandwidth.
l Maximum intranet PPS: indicates the maximum ECS capabilities in transmitting and
receiving packets.
NOTE
l Network bandwidth has been released in CN South, East, North, and Hong Kong, and will be
released in other regions.
l For instructions about how to test packet transmit and receive, see section How Can I Test Network
Performance?
l For instructions about how to enable NIC multi-queue, see section Enabling NIC Multi-Queue.
Compared with S1 ECSs, S2 ECSs are based on the latest-generation virtualization platform
and use non-uniform memory access (NUMA) binding to provide higher computing
performance.
Compared with S1 and S2 ECSs, S3 ECSs use latest-generation Intel Xeon Skylake CPUs,
which significantly improve the comprehensive performance. They provide a baseline level of
CPU performance with the ability to burst above the baseline and a balance of computing,
memory, and network resources. These ECSs are suitable for many applications.
Compared with S3 ECSs, SN3 ECSs feature improved network performance. Working in 25
Gbit/s networks, SN3 ECSs offer higher bandwidths, lower latency, more stable computing
performance for high PPS and cost-effectiveness.
S6 ECSs are suitable for applications that require moderate performance generally but
occasionally burstable high performance, such as light-load web servers, enterprise R&D and
testing environments, and low- and medium-performance databases. S6 ECS performance is
neither restricted by vCPU credits nor billed for additional credits. You can determine the
CPU usage and vCPU credits in monitoring details.
Specifications
Scenarios
l Web servers, small-load applications, and development and test environments
l Small-and medium-sized databases, cache servers, and search clusters
Specifications
Scenarios
l C3 ECSs
Small- and medium-scale databases, cache servers, and search clusters with high
requirements on stability; enterprise-class applications of diverse types and in various
scales
l C3ne ECSs
Network data forwarding, databases, cache servers, and search clusters with high
requirements on computing and network performance; enterprise-class applications of
diverse types and in various scales
l C6 ECSs
Websites and web applications, generalized databases and cache servers, and medium-
and heavy-workload enterprise applications with strict requirements on computing and
network performance
Working Rules
The credit mechanism is used on general-entry ECSs to prevent an overcommitted ECS from
occupying CPU resources for a long period of time. The initial credits are allocated to meet
the burstable performance requirement after the ECS is purchased. After the ECS runs, the
system allocates credits at a certain rate, and the credits can be accumulated. The upper limit
for the accumulated credits is the credit allocation rate multiplied by 24 hours. When the
upper limit is exceeded, the credits are not accumulated any more. The initial credits are not
counted in the upper limit for accumulated credits. If the actual ECS resource usage is higher
than the vCPU benchmark for a long period of time, the accumulated credits are continuously
consumed. When the accumulated credits reach 0, the vCPU usage cannot exceed the vCPU
benchmark.
The credit mechanism applies to applications with performance limited by vCPU benchmark
and credits.
Specifications
Specifications
Scenarios
l Applications
Memory-optimized ECSs are suitable for applications that process large volumes of data
and require a large amount of memory, rapid data switching and processing, and low-
latency storage resources.
l Application scenarios
Big data analysis, precision advertising, e-commerce big data analysis, IoV big data
analysis, relational databases, NoSQL databases, and memory data analysis
Specifications
Features
Large-memory ECSs use SR-IOV+OVS networks to provide a bandwidth as high as 10
Gbit/s.
Notes
l Large-memory ECSs do not support NIC hot swapping.
l E3 ECSs support the following OS that has been verified:
SUSE Enterprise Linux Server 12 SP2 64bit
l E1 and E2 ECSs can use the following types of EVS disks as system disk and data disk:
– High I/O (performance-optimized I)
– Ultra-high I/O (latency-optimized)
l The primary and extension NICs of a large-memory ECS have specified application
scenarios. For details, see Table 7-54.
l Up to 60 disks (including the system disk) can be attached to an ECS. For details about
restrictions, see section Can Multiple Disks Be Attached to an ECS?
An example is provided as follows:
If you create an e1.xlarge large-memory ECS, it can be attached with up to 60 disks,
where
– The number of system disks is 1.
– The number of EVS disks is at most 59.
Compared with D1 ECSs, D2 ECSs are developed based on KVM virtualization, use local
storage, and provide high storage performance and intranet bandwidth. They are designed for
distributed Hadoop computing, large data warehouse, distributed file system, data processing,
and log processing.
NOTE
D1 ECSs have been discontinued. D2 ECSs that feature higher performance are available for new orders
and capacity expansion for D1 ECSs.
Specifications
Scenarios
l Applications
Disk-intensive ECSs are suitable for applications that require large volumes of data to
process, high I/O performance, and rapid data switching and processing.
l Application scenarios
Big data computing, network file systems, data processing, MapReduce, Hadoop, and
data-intensive computing
Features of D2 ECSs
l D2 ECSs use local disks to provide high sequential read/write performance and low
latency, improving file read/write performance.
l D2 ECSs provide powerful and stable computing capabilities, ensuring efficient data
processing.
l D2 ECSs with a CPU/memory ratio of 1:8 process large volumes of data.
l D2 ECSs provide high intranet performance, including high intranet bandwidth and
packet per second (pps), meeting requirements for data exchange between ECSs during
peak hours.
l You are not allowed to buy additional local disks. The quantity and capacity of your
local disks are determined according to the specifications of your ECS. For D2 ECSs, if
additional local disks are required, buy them when creating the ECSs.
Ultra-high I/O ECSs can be used for high-performance relational databases, NoSQL databases
(such as Cassandra and MongoDB), and ElasticSearch.
Specifications
Features
Table 7-59 lists the IOPS performance of I3 ECSs.
i3.2xlarge.8 750000
i3.4xlarge.8 1500000
i3.8xlarge.8 3000000
i3.12xlarge.8 4500000
i3.15xlarge.8 5250000
Notes
l I3 ECSs do not support specifications modification.
l After an I3 ECS is deleted, the data on the local NVMe SSD is automatically deleted.
Back up the data before deleting such an ECS. Deleting local disk data is time-
consuming. Therefore, an I3 ECS requires a longer period of time than other ECSs for
releasing resources.
l The data reliability of local disks depends on the reliability of physical servers and hard
disks, which are SPOF-prone. Therefore, you are advised to perform data redundancy at
the application layer to ensure data availability. Use EVS disks to store long-term service
data.
l The device name of a local disk attached to an I3 ECS is /dev/nvme0n1 or /dev/
nvme0n2. For instructions about how to initialize a local disk, see section Initializing
EVS Data Disks.
l I3 ECSs support the following OSs:
o EulerOS 2.2
o CentOS 7.2
o CentOS 7.3
o Ubuntu Server 16.04
o SUSE Linux Enterprise Server 12 SP2
o Fedora 25 64bit
o OpenSUSE 42.2 64bit
NOTE
You are advised to use EulerOS 2.2 or Ubuntu Server 16.04 for optimal performance.
l If you use an I3 ECS created using a private image, make sure that the NVMe driver has
been installed during private image creation. If the driver has not been installed, install it
after creating the I3 ECS. For details, see section "(Optional) Installing Special
Drivers" in Image Management Service User Guide.
l An I3 ECS can be attached with a maximum of 60 disks (including VBD, SCSI, and
local disks). Among the 60 disks, the maximum number of SCSI disks is 30, and the
maximum number of VBD disks is 24 (including the system disk). For details about
restrictions, see section Can Multiple Disks Be Attached to an ECS?
l The basic resources, including vCPUs, memory, and image of an I3 ECS are still billed.
To stop billing such an ECS, delete it.
The processor and memory ratio of an HC2 ECS is 1:2 or 1:4. Each vCPU corresponds to the
hyperthreading of an Intel Xeon processor core. HC2 ECSs can be used for high-performance
computing services. They provide a large number of parallel computing resources and high-
performance infrastructure services to meet the requirements of high-performance computing
and massive storage and ensure rendering efficiency. HC2 ECSs are frequently used in the
following scenarios:
l Computing and storage systems for genetic engineering, games, animations,
biopharmaceuticals, and scientific computing
l Public rendering platforms for renderfarms and animation and film bases; other
rendering platforms for movies and videos
l High-performance frontend clusters, web servers, high-performance science and
engineering applications, advertisements, video coding, and distributed analysis
H3 ECSs use high-performance Intel Xeon SkyLake CPUs. Each vCPU corresponds to the
hyper-thread of an Intel Xeon processor core, providing stable computing capabilities. H3
ECSs are suitable for high-performance computing services. In addition, the ECSs use latest-
generation network acceleration engines and DPDK rapid packet processing mechanism to
provide high network performance.
Specifications
Scenarios
l Computing and storage systems for genetic engineering, games, animations, and
biopharmaceuticals
l Public rendering platforms for renderfarms and animation and film bases; other
rendering platforms for movies and videos
l High-performance frontend clusters, web servers, high-performance science and
engineering applications, advertisements, video coding, and distributed analysis
l Batch-processed workload, high-performance computing (HPC), and SAP applications
l Computing-intensive services, such as large-scale multiplayer online (MMO) gaming
Specifications
Features
l Large memory capacity and more processor cores than other types of ECSs
l InfiniBand NICs providing a bandwidth of 100 Gbit/s
l 100 Gbit/s EDR InfiniBand network
NOTE
l Remote login on the management console can be used for O&M, but it cannot be used in the
production environment. Physical GPUs cannot be used if an ECS is remotely logged in through the
management console.
l If remote login on the management console is unavailable, use Windows MSTSC or a third-party
desktop protocol, such as TeamViewer or VNC for login.
Graphics-accelerated Enhancement G3
Overview
G3 ECSs are based on PCI passthrough and exclusively use GPUs for professional graphics
acceleration. In addition, G3 ECSs use NVIDIA Tesla M60 GPUs and support DirectX and
OpenGL with up to 16 GB of GPU memory and 4,096 x 2,160 resolution. They are ideal for
professional graphics workstations.
Specifications
NOTE
Every NVIDIA Tesla M60 card is equipped with two M60 GPUs, each of which provides 2,048 CUDA
cores and 8 GB of GPU memory. M60 in G series of ECSs indicates M60 GPUs, but not M60 cards.
G3 ECS Features
l Professional graphics acceleration APIs
l NVIDIA M60 GPUs
l Graphics acceleration applications
l GPU passthrough
l Application flow identical to common ECSs
l Automatic scheduling of G3 ECSs to AZs where NVIDIA M60 GPUs are used
l A maximum specification of 16 GB of GPU memory and 4,096 x 2,160 resolution for
processing graphics and videos
Notes
Graphics-accelerated G1
Overview
G1 ECSs are based on NVIDIA GRID vGPUs and provide economical graphics acceleration.
They use NVIDIA Tesla M60 GPUs and support DirectX and OpenGL. The ECSs have up to
8 GB of GPU memory and 4,096 x 2,160 resolution, and are used for applications that require
high performance in graphics rendering.
Specifications
NOTE
M60-xQ support vGPUs. x can be 1, 2, 4, or 8, indicating that M60 GPUs are virtualized to vGPUs with
different specifications and models using GRID. x specifies the vGPU memory, and Q indicates that the
vGPU of this type is designed to work in workstations and desktop scenarios. For more details about
GRID vGPUs, see GRID VIRTUAL GPU User Guide.
G1 ECS Features
l NVIDIA M60 GPUs
l Graphics acceleration applications
l GPU hardware virtualization (vGPUs) and GPU passthrough
l Application flow identical to common ECSs
l Automatic scheduling of G1 ECSs to AZs where NVIDIA M60 GPUs are used
l A maximum specification of 8 GB of GPU memory and 4,096 x 2,160 resolution for
processing graphics and videos
Notes
l G1 ECSs do not support specifications modification.
l G1 ECSs support the following OSs:
– Windows Server 2008 R2 Enterprise SP1 64bit
– Windows Server 2012 R2 Standard 64bit
l g1.2xlarge.8 G1 ECSs do not support the remote login function provided by the public
cloud platform. To remotely log in to such an ECS, use MSTSC to log in to it and install
VNC on the ECS.
Non-g1.2xlarge.8 G1 ECSs support the remote login function provided by the public
cloud platform. For details, see section Login Using VNC.
l If a G1 ECS is created using a private image, install a GPU driver on the ECS after the
ECS creation. For details, see section Installing the GRID Driver.
Computing-accelerated P2v
Overview
P2v ECSs use NVIDIA Tesla V100 GPUs and provide flexibility, high-performance
computing, and cost-effectiveness. These ECSs use GPU NVLink for direct communication
between GPUs, improving data transmission efficiency. P2v ECSs provide outstanding
universal computing capabilities and have strengths in AI-based deep learning, scientific
computing, Computational Fluid Dynamics (CFD), computing finance, seismic analysis,
molecular modeling, and genomics.
Specifications
P2v ECSs are used in computing acceleration scenarios, such as deep learning training,
inference, scientific computing, molecular modeling, and seismic analysis. If the software
requires GPU CUDA parallel computing, use P2v ECSs. P2v ECSs support the following
commonly used software:
l Common deep learning frameworks, such as TensorFlow, Caffe, PyTorch, and MXNet
l CUDA GPU rendering supported by RedShift for Autodesk 3dsMax and V-Ray for 3ds
Max
l Agisoft PhotoScan
l MapD
Notes
l If a P2v ECS is created using a private image, make sure that the NVIDIA driver has
been installed during the private image creation. If not, install the driver after the P2v
ECS is created for computing acceleration. For details, see section Installing the
NVIDIA GPU Driver and CUDA Toolkit.
l P2v ECSs do not support specifications modification.
l Pay-per-use P2v ECSs support the following OSs:
– Ubuntu Server 16.04 64bit
– CentOS 7.4 64bit
– EulerOS 2.2 64bit
– Windows Server 2012 R2 Standard 64bit
– Windows Server 2016 Standard 64bit
Computing-accelerated P1
Overview
P1 ECSs use NVIDIA Tesla P100 GPUs and provide flexibility, high performance, and cost-
effectiveness. These ECSs support GPU Direct for direct communication between GPUs,
improving data transmission efficiency. P1 ECSs provide outstanding universal computing
capabilities and have strengths in deep learning, graphic databases, high-performance
databases, Computational Fluid Dynamics (CFD), computing finance, seismic analysis,
molecular modeling, and genomics. They are designed for scientific computing.
Specifications
P1 ECS Features
l Up to four NVIDIA Tesla P100 GPUs on a P1 ECS (If eight P100 GPUs are required on
an instance, use BMS.)
l GPU hardware passthrough
l 9.3 TFLOPS of single-precision computing and 4.7 TFLOPS of double-precision
computing
l Maximum network bandwidth of 10 Gbit/s
l 16 GB of HBM2 GPU memory with a bandwidth of 732 Gbit/s
l 800 GB NVMe SSDs for temporary local storage
l Comprehensive basic capabilities
Networks are user-defined, subnets can be divided, and network access policies can be
configured as needed. Mass storage is used, elastic capacity expansion as well as backup
and restoration are supported to make data more secure. Auto Scaling allows you to add
or reduce the number of ECSs quickly.
l Flexibility
Similar to other types of ECSs, P1 ECSs can be provisioned in a few minutes. You can
configure specifications as needed. P1 ECSs with two, four, and eight GPUs will be
supported later.
l Excellent supercomputing ecosystem
The supercomputing ecosystem allows you to build up a flexible, high-performance,
cost-effective computing platform. A large number of HPC applications and deep-
learning frameworks can run on P1 ECSs.
Supported Common Software
P1 ECSs are used in computing acceleration scenarios, such as deep learning training,
inference, scientific computing, molecular modeling, and seismic analysis. If the software
requires GPU CUDA parallel computing, use P1 ECSs. P1 ECSs support the following
commonly used software:
l Deep learning frameworks, such as TensorFlow, Caffe, PyTorch, and MXNet
l RedShift for Autodesk 3dsMax, V-Ray for 3ds Max
l Agisoft PhotoScan
l MapD
Notes
l It is recommended that the system disk of a P1 ECS be greater than 40 GB.
l P1 ECSs have local NVMe SSDs attached, which are still billed after the ECSs are
stopped. To stop billing such an ECS, delete it.
l The local NVMe SSDs attached to P1 ECSs are dedicated for services with strict
requirements on storage I/O performance, such as deep learning training and HPC. Local
disks are attached to the ECSs of specified flavors and cannot be separately bought. In
addition, you are not allowed to detach a local disk and then attach it to another ECS.
NOTE
Data may be lost on the local NVMe SSDs attached to P1 ECSs due to a fault, for example, due to
a disk or host fault. Therefore, you are suggested to store only temporary data in local NVMe
SSDs. If you store important data in such a disk, securely back up the data.
l If a P1 ECS is created using a private image, make sure that the NVIDIA driver has been
installed during the private image creation. If not, install the driver after the P1 ECS is
created for computing acceleration. For details, see section Installing the NVIDIA GPU
Driver and CUDA Toolkit.
NOTE
For instructions about how to create a private image, see Image Management Service User Guide.
l P1 ECSs do not support specifications modification.
l P1 ECSs do not support automatic recovery.
l P1 ECSs support the following OSs:
– Debian 8.0 64bit
– Ubuntu Server 16.04 64bit
– CentOS 7.3 64bit
– EulerOS 2.2 64bit
– Windows Server 2012 R2 Standard 64bit
l After you delete a P1 ECS, the data stored in local NVMe SSDs is automatically cleared.
Inference-accelerated PI1
Overview
PI1 ECSs use NVIDIA Tesla P4 GPUs dedicated for real-time AI inference. Working with P4
INT8 calculators, PI1 ECSs have shortened the inference latency by 15 times. Working with
hardware decoding engines, PI1 ECSs concurrently support real-time 35-channel HD video
transcoding and inference.
Specifications
l If a PI1 ECS is created using a private image, make sure that the GPU driver has been
installed during the private image creation. If not, install the driver after the PI1 ECS is
created for inference acceleration. For details, see section Installing the NVIDIA GPU
Driver and CUDA Toolkit.
l PI1 ECSs support the following OSs:
– Ubuntu Server 14.04 64bit
– Ubuntu Server 16.04 64bit
– CentOS 7.3 64bit
l PI1 ECSs do not support specifications modification.
l PI1 ECSs support automatic recovery when the hosts accommodating such ECSs
become faulty.
Specifications
NOTE
Network: 10 GbE
NOTE
Network: 10 GbE
Scenarios
l Applications
FPGA-accelerated ECSs are suitable for applications that require dense computing, high
concurrency, and high bandwidth, such as video processing, machine learning, genomics,
and financial risk analysis.
l Application scenarios
– Video processing: Video applications, such as automatic image recognition and
classification, image search, video transcoding, real-time rendering, Internet-based
live programs, and AR/VR, require high real-time computing performance, which
cannot be provided by common ECSs. FPGA ECSs offer cost-effective video
solutions, which are ideal for video scenarios.
– Machine learning: Multi-layer neural networks in machine learning require a large
number of computing resources. The training process involves handling massive
data, while the inference process requires ultra-low latency. In addition, machine
learning algorithms are being continuously optimized. FPGA ECSs meet the
preceding requirements due to their high parallel computing, programmable
hardware, low power consumption, and low latency. FPGA dynamically provides
the optimal hardware circuit design for different machine learning algorithms,
meeting strict requirements on massive computing and ultra-low latency. Therefore,
FPGA ECSs meet hardware requirements for machine learning.
– Genomics research: Precision medicine can be implemented through gene
sequencing and analysis as well as rapid analysis on massive biological and medical
data. Many fields, such as pharmaceutical development and molecular breeding,
also require the processing of massive data. These fields require hardware
acceleration to resolve performance bottleneck problems for biological
computation. FPGA ECSs meet such requirements due to their outstanding
programmable hardware computing performance.
– Financial risk analysis: The financial industry has strict requirements on computing
capabilities and real-time performance based on ultra-low latency and high
throughput for services, such as pricing tree model-based financial computing,
high-frequency trading, fund or securities trading algorithms, financial risk analysis
and decision-making, and transaction security assurance. Using programmable
hardware acceleration, FPGA ECSs offer an optimal hardware acceleration solution
for various scenarios. In certain scenarios, FPGA ECS performance has been
improved by thousands of times when compared with the performance of stand-
alone software.
Features
l FPGA development kit
FPGA cloud services provide the hardware development kit (HDK). The HDK contains
accelerator examples, the encoding environment, simulation platform, automatic
compilation tool, and code encrypting and commissioning tool package. You can quickly
develop and test your FPGA hardware accelerator by referring to the application
examples and development guide.
l Software development kit
FPGA cloud services provide the software development kit (SDK). The SDK contains
application examples, hardware abstract interfaces, accelerator abstract interfaces,
accelerator driver and runtime, and the version management tool. Accelerator abstract
interfaces allow your applications to invoke the hardware accelerator in the same way as
the software function library, helping you easily develop high-performance applications
based on the hardware accelerator.
l FPGA hardware configuration
Each FPGA ECS provides up to eight FPGAs and each of them contains 2.5 million
logic units. The PCIe 3.0 x 16 interface is supported, and the throughput reaches 100
Gbit/s. FPGAs communicate with each other through a mesh optical network with a
bandwidth of 300 Gbit/s. Each FPGA provides 64 GB DDR4 memory and an interface
rate of 2133 MHz. FPGA frees your applications from hardware restrictions.
l Hardware acceleration resource pool
FPGA cloud services have a pool of hardware acceleration resources. You can allocate
desirable FPGA resources in a cost-effective way like allocating CPU resources. With
the FPGA virtualization, isolation, and distributed technology, you can share FPGA
resources on a node, which is transparent to your resources and can meet hardware
acceleration requirements of your business to the largest extent.
Notes
l FPGA-accelerated ECSs support only the CentOS 7.3 64bit OS.
l FPGA-accelerated ECSs do not support modifying specifications.
l FPGA-accelerated ECSs cannot be migrated.
l FPGA-accelerated ECSs do not support automatic recovery.
l FP1 and FP1c ECSs are equipped with FPGAs, which are still billed after the ECSs are
stopped. To stop billing such an ECS, delete it.
Follow-up Procedure
After the ECS is created, you can develop and use accelerated engine images (AEIs) through
the hardware development kit (HDK) and application development kit (SDK) provided by the
ECS. For more details, see FACS User Guide.
Related Links
FACS User Guide
Specifications
Features
AI1 ECSs have the following features:
Notes
1. AI1 ECSs support the following OSs:
Ubuntu Server 14.04 64bit
2. AI1 ECSs do not support specifications modification.
3. AI1 ECSs support automatic recovery when the hosts accommodating such ECSs
become faulty.
mini_mind_studio_Ubuntu.rar https://support.huawei.com/carrier/navi?
coltype=software#col=software&detailId=PBI
1-23772363&path=PBI1-21430725/
PBI1-21430756/PBI1-22892969/
PBI1-22368564
ECS-related Services
Image Management Enables you to create ECSs l Creating a Data Disk Image
Service (IMS) using images. This improves Using an ECS Data Disk
the efficiency of ECS creation. l Creating a Full-ECS Image
You can also use an existing Using an ECS
ECS to create a private image
and export the data of the ECS
system disk or data disks.
9 Change History
Released On Description
Released On Description
Released On Description
Released On Description