Classification and naming of instance types - Elastic Compute Service

Alibaba Cloud provides various instance types to fit different use cases. This topic describes the relationships between instance family categories, instance families, and instance types. This topic also describes the naming conventions of instance types.

Differences between instance family categories

Alibaba Cloud groups instance families into the following categories based on the CPU architecture and supported scenarios: x86-based computing, Arm-based computing, ECS Bare Metal Instance, High-performance computing, Super Computing Cluster (SCC), and heterogeneous computing.

Instance family category	Description
x86-based computing and Arm-based computing instance family categories	x86-based computing instance family category: Instance families in this category are based on the x86 architecture, and each vCPU of the instance families corresponds to the hyperthread of a processor core. The instance families provide stable performance and are powered by different brands of processors, such as Intel, AMD, and Hygon. x86-based enterprise-level computing instance families: suitable for various scenarios, such as enterprise-level applications of various types and sizes, database systems, video encoding and decoding, and data analytics. x86-based entry-level computing instance families (x86-based shared instance families): intended for small and medium-sized websites or individual developers. Compared with enterprise-level instance types, shared instance types share resources to maximize resource utilization. As a result, shared instance types do not provide stable computing performance but cost less. Arm-based computing instance family category: Instance families in this category are based on the Arm architecture, and each vCPU of the instance families corresponds to a physical processor core. The instance families provide stable performance by using dedicated resources and are suitable for scenarios such as containers, microservices, website and application servers, high-performance computing, and CPU-based machine learning. The instance families are powered by different brands of processors, such as YiTian 710 and Ampere^® Altra^®.
ECS Bare Metal Instance family category	ECS Bare Metal Instance combines the strengths of physical machines and ECS instances to deliver powerful and robust computing capabilities. ECS Bare Metal Instance uses virtualization 2.0 to provide your business applications with direct access to the processor and memory resources of the underlying servers without virtualization overheads. ECS Bare Metal Instance retains the hardware feature sets, such as Intel VT-x, and resource isolation capabilities of physical machines, which is ideal for applications that need to run in non-virtualization environments in the cloud.
SCC instance family category	SCC is based on ECS Bare Metal Instance and uses high-speed Remote Direct Memory Access (RDMA)-based interconnections to significantly improve the network performance and acceleration ratio of large-scale clusters. SCC has all the benefits of ECS Bare Metal Instance and provides high-bandwidth, low-latency networks.
Heterogeneous computing instance family category	Elastic GPU Service: provides GPU-accelerated instance families that have GPU-accelerated computing capabilities and read-to-use, scalable GPU computing resources. As a member of Alibaba Cloud elastic computing service family, Elastic GPU Service combines the computing power of GPUs and CPUs to address the challenges in scenarios such as AI, high-performance computing, and professional graphics and image processing. For example, you can use Elastic GPU Service in a parallel computing scenario to significantly accelerate computing. Heterogeneous service instance family (video-trans instance family): suitable for scenarios such as video transcoding, image and video processing, and frame image extraction. Visual computing instance family: The ebmgi6s visual computing instance family, which is based on the Alibaba Cloud SHENLONG architecture and Intel^® Server GPUs, provides a security architecture that can be quickly scaled and the latest high-density instances that allow rendering of cloud mobile games.

Relationships between instance families and instance types

An instance family is a group of instance types that use the same processors and are suitable for similar business and use scenarios. Each instance family consists of multiple instance types that each have a combination of different capacities, including CPU and memory capacities. The ECS instance type defines the basic properties of an ECS instance, including CPU (CPU model and clock speed) and memory. When you create a instance, you must configure the instance type, network type, Elastic Block Storage (EBS) devices, and image.

The following figure shows the relationships between instance families and instance types.

Note

Only specific instance families and types are shown in the figure. For information about the other instance families and instance types, see Overview of instance families.

Naming conventions of instance types

Each instance family name is in the ecs.<Instance family> format. Each instance type name is in the ecs.<Instance family>.<Instance size> format.

ecs: the product code of ECS.
<Instance family>: consists of an instance family name body and an instance family name suffix.
<Instance size>: the number of vCPUs, which can be small, large, or <nx>large. small indicates 1 vCPU, large indicates 2 vCPUs, and xlarge indicates 4 vCPUs. A larger value of n in <nx> indicates a larger number of vCPUs. For example, 2xlarge indicates 8 vCPUs (= 2 × 4) and 3xlarge indicates 12 vCPUs (= 3 × 4).

x86-based computing instance families and Arm-based computing instance families

For example, ecs.g8ae.4xlarge is an instance type in a general-purpose instance family and is powered by enhanced AMD CPUs. The instance type has 16 vCPUs (= 4 × 4). The instance type has 64 GiB of memory based on the CPU-to-memory ratio of 1:4 of general-purpose instance families.

Instance type names vary in the <Instance family> part. The following table describes the <Instance family> part of each instance type name.

Instance family name body (lowercase letters + digits)

Instance family name suffix

Lowercase letter

Digit

Lowercase letter

An abbreviation that indicates the performance characteristics of the instance family.

c: compute-optimized
Compute-optimized instance families offer a CPU-to-memory ratio of 1:2 and are suitable for scenarios such as databases, web servers, high-performance scientific and engineering applications, game servers, data analytics, batch computing, video encoding, and machine learning.
g: general-purpose
General-purpose instance families offer a CPU-to-memory ratio of 1:4 and are suitable for scenarios such as general-purpose Internet applications, databases, websites, Java application services, game servers, and search promotions.
u: universal
Processors of universal instances may be deployed on different server platforms. Universal instance families offer CPU-to-memory ratios of 1:1, 1:2, 1:4, and 1:8 and are suitable for enterprise-level users who are price-sensitive in scenarios such as small and medium-sized enterprise-level applications, websites and application servers, small and medium-sized database systems, caches, and search clusters.
r: memory-optimized
Most memory-optimized instance families offer a CPU-to-memory ratio of 1:8. Memory-optimized instance families are suitable for in-memory databases, data analytics and mining, distributed in-memory caches such as Redis, big data applications such as Kafka and Elasticsearch, and memory-intensive, general-purpose, and enterprise-level applications such as Java.
re: memory-enhanced or high memory
hf, hc, hg, and hr: high clock speeds or high frequency
Instance families with high clock speeds (also called high-frequency instance families) offer CPU-to-memory ratios of 1:2, 1:4, and 1:8 and are suitable for scenarios such as massively multiplayer online (MMO) games, high-performance scientific computing, and large and medium-sized database systems.
i: local SSDs
Instance families equipped with local SSDs offer CPU-to-memory ratios of 1:4 and 1:8 and are suitable for scenarios such as online transaction processing (OLTP) and high-performance relational databases, NoSQL databases such as Cassandra and MongoDB, search scenarios that use solutions such as Elasticsearch, and E-MapReduce big data scenarios such as storage and computing separation.
d: big data
Most big data instance families offer a CPU-to-memory ratio of 1:4 and are suitable for big data computing and storage scenarios in which services such as Hadoop MapReduce, Hadoop Distributed File System (HDFS), Hive, and HBase are used, and search and log data processing scenarios in which solutions such as Elasticsearch and Kafka are used.
s: shared
- t: burstable
- e: economy

The generation to which the instance family belongs. A larger number indicates a newer generation of instance families that is more cost-effective.

Examples: 8, 7, 6, and 5.

An abbreviation that indicates the other characteristics of the instance family.

y: equipped with in-house Arm-based YiTian 710 processors
a: equipped with AMD processors
ae: AMD-enhanced
i: equipped with Intel processors
h: equipped with Hygon processors
re: RDMA-enhanced
se: storage-enhanced
ne or nex: network-enhanced
t: security-enhanced
p: persistent memory-optimized
g: general-purpose
r: memory-optimized
c: compute-intensive

Heterogeneous computing instance families, ECS Bare Metal Instance families, and SCC instance families

For example, ecs.ebmgn7ix.32xlarge is an instance type in a GPU-accelerated, compute-optimized ECS Bare Metal Instance family. The instance type is powered by NVIDIA A10 GPUs with 24 GB of memory each and enhanced AMD CPUs. The instance type has 128 vCPUs (= 32 × 4). The digit 7 indicates the Ampere architecture.

Instance type names vary in the <Instance family> part. The following table describes the <Instance family> part of each instance type name.

Instance family name body (lowercase letters)

Instance family name suffix (lowercase letters + digits)

gn: NVIDIA GPU-accelerated compute-optimized
vgn: NVIDIA GRID vGPU-accelerated dedicated
sgn: NVIDIA GRID vGPU-accelerated shared
gi: Intel GPU-accelerated compute-optimized
f: FPGA-accelerated compute-optimized
ebmc, ebmg, ebmr, ebmgn, or ebmhf: ECS Bare Metal Instance
sccc, sccg, scch, sccgn, or scchf: SCC

6v: 6 indicates the Volta or Turing architecture, and v indicates V100 GPUs with 16 GB of memory each.
For example, gn6v is a compute-optimized instance family that is powered by NVIDIA V100 GPUs with 16 GB of memory each based on the Volta or Turing architecture.
6e: 6 indicates the Volta or Turing architecture, and e, which is short for extend, indicates second-generation V100 GPUs with 32 GB of memory each.
For example, gn6e is a compute-optimized instance family that is powered by NVIDIA V100 GPUs with 32 GB of memory each based on the Volta or Turing architecture.
6i: 6 indicates the Volta or Turing architecture, and i indicates T4 GPUs.
For example, gn6i is a compute-optimized instance family that is powered by NVIDIA T4 GPUs based on the Volta or Turing architecture.
6s: 6 indicates the Volta or Turing architecture, and s indicates sixth-generation SG-1.
For example, ebmgi6s is a visual computing instance family that is powered by Intel^® Server GPUs and sixth-generation SG-1.
7: indicates the Ampere architecture.
7i: 7 indicates the Ampere architecture, and i, which is short for inference, indicates A10 GPUs with 24 GB of memory each.
7e: 7 indicates the Ampere architecture.
7s: 7 indicates the Ampere architecture, and s indicates seventh-generation A30 GPUs.

Instance type metrics

Metric	Description
Processor	The physical CPU model of an ECS instance. The performance of the processor varies based on the processor type. CPU base frequency: also known as core frequency, refers to the standard rate at which a processor is running without overclocking or special optimization. The CPU base frequency is the operating speed specified by the designer of a processor and is the typical clock speed at which the processor reliably runs under typical workloads. CPU turbo frequency: the maximum clock frequency that can be achieved when the core frequency of a processor is temporarily increased.
vCPU	Each vCPU of x86-based instance types corresponds to a hyperthread of a processor core. Each vCPU of Arm-based instance types corresponds to a physical processor core. The instance types deliver stable performance by using dedicated resources.
Burst performance	Average baseline CPU performance: The amount of CPU capacity that can be continuously provisioned to a burstable instance. CPU credits per hour: When a burstable instance starts, the instance begins to consume CPU credits to maintain its computing performance. The instance earns CPU credits at a specific rate that varies based on the instance type. The number of CPU credits that a vCPU can earn per hour varies based on the instance type. The CPU credits per hour column in the instance type table indicates the CPU credits that all vCPUs of a burstable instance can earn per hour. Maximum CPU credit balance: The maximum number of CPU credits that a burstable instance can earn within a 24-hour period. The CPU credit balance is valid for up to 24 hours. Each burstable instance earns CPU credits at a specific rate based on the instance type and can accrue only a limited number of credits in the CPU credit balance.
Memory	Memory: stores and retrieves data that can be quickly accessed on an ECS instance. Memory temporarily stores data generated when the instance is running programs and processing data. If the instance is shut down or restarted, the data stored in memory is lost. Persistent memory: Persistent memory can be used as memory or local storage. Persistent memory is listed separately from memory in the instance type tables of instance families that support persistent memory. The usage of persistent memory varies based on the instance type. For more information, see Configure the usage mode of persistent memory. Encrypted memory: Encrypted memory is provided based on the Intel^® Software Guard Extensions (SGX) technology. In the instance type tables of instance families that support encrypted memory, the Memory column indicates the total amount of memory, including the amount of encrypted memory. For more information, see Build an SGX confidential computing environment.
Network bandwidth	Network baseline bandwidth: the maximum amount of data that can be transmitted over a network connection in a specific period of time, which is the initial bandwidth setting for a network connection. You can select network baseline bandwidths based on your business requirements. For more information, see Network bandwidth. Network burst bandwidth: the maximum data transmission rate to which an ECS instance can burst beyond its network baseline bandwidth by consuming network burst credits within a short period of time Specific instance types in sixth-generation or later instance families support network burst bandwidth. Network burst bandwidth is a way to utilize idle resources. No service level agreement (SLA) commitments are made for network burst bandwidth. For more information, see the Burst bandwidth section of the "Network bandwidth" topic. Full-duplex network bandwidth: supported by instance types in seventh-generation or later instance families. Based on the full-duplex network bandwidth, the inbound and outbound network bandwidths can reach the preset values, without affecting each other. For more information, see the corresponding performance metrics of each instance family in the "Overview of instance families" topic. An ECS instance can send and receive packets at full rates. For more information, see the Internal bandwidth section of the "Network bandwidth" topic. Note Instance type specifications are all verified and obtained in a pure forwarding test environment. In actual business scenarios, the performance of an ECS instance may vary based on other factors such as instance workload type, packet size, connection type (persistent or short-lived), image version, and networking model. We recommend that you perform business stress tests on instances to select appropriate instance types.
Packet forwarding rate (pps)	The sum of maximum inbound and outbound packet forwarding rates. For information about how to test the packet forwarding rate of an ECS instance, see Test the network performance of an instance. Note Instance type specifications are all verified and obtained in a pure forwarding test environment. In actual business scenarios, the performance of an ECS instance may vary based on other factors such as instance workload type, packet size, connection type (persistent or short-lived), image version, and networking model. We recommend that you perform business stress tests on instances to select appropriate instance types.
Connections	A connection, also called a session, is the process of connecting a client to a server and transferring data between the client and the server. A connection is uniquely defined by the network communication quintuple that consists of a source IP address, a destination IP address, a source port, a destination port, and a protocol. The connections of an ECS instance include TCP, UDP, and Internet Control Message Protocol (ICMP) connections. If your business is sensitive to concurrent network connections, select an instance of an instance type of which the maximum number of connections meets your business requirements.
NIC queues	The maximum number of network interface controller (NIC) queues that is supported by the instance type per elastic network interface (ENI). It is also the maximum number of traffic queues that an instance can process on each ENI. A larger number of NIC queues indicates more efficient distribution and processing of network data, less time that packets wait to be processed, improved network performance, and lower packet loss rate and network latency. Specify a proper number of NIC queues based on the network workload, hardware performance, and system configurations. For more information, see Manage NIC multi-queue.
ENIs	The maximum number of ENIs that can be bound to an instance. You can bind one or more ENIs to each ECS instance. You can bind secondary ENIs to or unbind secondary ENIs from different ECS instances to allow for more flexibility and scalability in network configurations and meet network requirements in different scenarios. For example, you can use ENIs to create multi-IP address, multi-NIC, or high-availability networks. For more information, see Overview.
ERIs	The maximum number of elastic RDMA interfaces (ERIs) that can be bound to an instance. ERIs are Remote Direct Memory Access (RDMA)-capable ENIs. ERIs reuse the networks to which ENIs belong. This allows you to use the RDMA feature in the original networks and enjoy the low latency provided by RDMA without the need to modify service networking. For more information, see Overview.
Support for Jumbo Frames	Specifies whether the instance type supports the Jumbo Frames feature. Alibaba Cloud supports jumbo frames that are 8,500 bytes in size and allows you to send Ethernet frames that have 8,500 bytes of payload. Jumbo frames increase the payload size per packet and the percentage of the packet that is not packet overhead to provide high throughput and high network performance. For information about how to enable the Jumbo Frames feature, see Jumbo frames.
Private IPv4 addresses per ENI	The maximum number of private IPv4 addresses that is supported by the instance type per ENI.
IPv6 addresses per ENI	The maximum number of IPv6 addresses that is supported by the instance type per ENI.
I/O optimized	I/O optimization provides better network capabilities and storage performance for instances and cloud disks. For example, you can attach a standard SSD to an I/O optimized instance to maximize the storage performance of the standard SSD.
Local storage	Local storage, also called local disks, refers to disks that are attached to the physical machines on which ECS instances are hosted. Local storage provides temporary block storage for ECS instances. Local storage cannot be separately created and is measured in a binary unit of GiB. Warning The durability of data stored in a local disk is determined by the reliability of the associated physical machine. Local disks are subject to single point of failure (SPOF) risks. Data stored on local disks may be lost. Do not store long-lived business data on local disks. For more information, see Local disks.
Disk bandwidth	Disk baseline bandwidth: the maximum disk bandwidth that an instance type can consistently support for cloud disks in compliance with an SLA commitment. Disk burst bandwidth: the cloud disk bandwidth to which an instance type can burst. Disk burst bandwidths are time-limited and dependent on the number of disk bandwidth resources provided by an instance. No SLA commitments are made for disk burst bandwidths.
Disk IOPS	Baseline IOPS: the maximum IOPS that an instance type can consistently support for cloud disks in compliance with an SLA commitment. Burst IOPS: the IOPS to which an instance type can burst. Burst IOPS is time-limited and dependent on the number of I/O resources provided by an instance. No SLA commitments are made for burst IOPS.
vTPM	Trusted computing capabilities: Trusted Platform Modules (TPMs) or Trusted Cryptography Modules (TCMs) serve as trusted computing bases (TCBs) on the underlying physical servers that host trusted instances to ensure the tamper-protected, trusted boot of the instances. Virtual TPM (vTPM) can be used to measure the critical components of the boot chain of instances.