All Products
Search

This Product

    Elastic Compute Service:Compute-optimized instance families (c series)

    Document Center

    Elastic Compute Service:Compute-optimized instance families (c series)

    Last Updated:Feb 28, 2026

    Compute-optimized instance families provide a 1:2 CPU-to-memory ratio for CPU-intensive workloads such as web servers, databases, batch computing, video encoding, game servers, data analytics, high-performance scientific computing, and machine learning.

    The ic5 family uses a 1:1 CPU-to-memory ratio instead of 1:2.

    Before you start:

    Choose an instance family

    Instance naming convention

    Instance type names follow a predictable pattern. For example, ecs.c9ae.4xlarge:

    SegmentMeaningExample
    cFamily: compute-optimizedc = compute
    9Generation9 = latest
    aProcessor vendora = AMD, i = Intel (recent generations), blank = Intel (older generations)
    eVariante = performance-enhanced
    4xlargeSize16 vCPUs

    Instance family summary

    The following tables list all compute-optimized instance families. Use them to narrow down your options before reading detailed specifications.

    Recommended families (x86)

    Instance familyProcessorArchitectureTurbo frequencyeRDMAJumbo framesvTPMEnclaveMax vCPUsMax network bandwidth
    c9iIntel Xeon 6 P-cores (Granite Rapids)CIPU3.9 GHzYesYesYesNo19264 Gbit/s
    c9aeAMD EPYC TurinCIPU3.7 GHzYesYesYesNo192100 Gbit/s
    c9aAMD EPYC TurinCIPU4.1 GHzYesYesYesNo6432 Gbit/s
    c8iIntel Xeon Emerald/Sapphire RapidsCIPU3.2 GHzYesYesYesYes192100 Gbit/s
    c8ineIntel Xeon Emerald/Sapphire RapidsCIPU3.2 GHzNoYesYesNo3244 Gbit/s
    c8aAMD EPYC GenoaCIPU3.7 GHzYesYesYesNo19264 Gbit/s
    c8aeAMD EPYC GenoaCIPU3.75 GHzYesYesYesNo12864 Gbit/s
    c7Intel Xeon Ice Lake3rd-gen SHENLONG3.5 GHzNoYesYesYes12864 Gbit/s
    c7aAMD EPYC MILAN3rd-gen SHENLONG3.5 GHzNoNoNoNo12832 Gbit/s
    c6eIntel Xeon Cascade Lake3rd-gen SHENLONG3.2 GHzNoNoNoNo10432 Gbit/s
    c6Intel Xeon Cascade LakeSHENLONG3.2 GHzNoNoNoNo10425 Gbit/s

    Recommended families (Arm)

    Instance familyProcessorArchitectureFrequencyeRDMAJumbo framesvTPMMax vCPUsMax network bandwidth
    c8yYitian 7104th-gen SHENLONG2.75 GHzYesYesYes12864 Gbit/s
    c6rAmpere Altra3rd-gen SHENLONG2.8 GHzNoNoNo6416 Gbit/s

    Not recommended families -- if sold out, use the families listed above instead.

    Instance familyProcessorUpgrade toWhy upgrade
    c5Intel Xeon Skylake/Cascade Lakec6, c6e, or c7Inconsistent server platform assignment
    ic5Intel Xeon Skylake/Cascade Lakec6 or c71:1 CPU-to-memory ratio, IPv4 only, no IPv6
    sn1neIntel Xeon Broadwell/Skylake/Cascade Lakec6, c6e, or c7Limited disk support (standard SSDs and ultra disks only)

    Match your workload

    WorkloadRecommended familiesWhy
    Web and application serversc9i, c9ae, c8i, c8aLatest generation, balanced performance
    Databases (large/medium)c9a, c8ae, c7High single-core turbo, consistent performance
    Game servers (MMO frontend)c9a, c8ae, c7, c7aHigh clock speed, ultra-high PPS
    Data analytics and batch computingc9ae, c9i, c8a, c8iHigh throughput, strong I/O
    Video encoding and transcodingc9ae, c9i, c8aHigh core count, compute-dense
    AI training and inferencec9ae, c8a, c8aeHigh-frequency AMD, GPU-free ML workloads
    HPC and scientific computingc8ae, c9iPerformance-enhanced, eRDMA for low latency
    Containers and microservicesc8y, c6rArm-based, cost-efficient for scale-out
    Network-intensive (gateways, forwarding)c8ineOptimized for high connection counts and PPS
    Confidential computingc8i, c7Enclave support, trusted boot, Intel TME

    c9ae, compute-optimized instance family

    Built on the Cloud Infrastructure Processing Unit (CIPU) architecture with AMD EPYC Turin processors. Uses physical cores (no Hyper-Threading) for consistent computing performance, with adjustable baseline storage and network bandwidth and chip-level security hardening.

    Use cases: Big data analytics (Spark, Flink, Elasticsearch), search and recommendation engines, core transactional processing (TP) systems, audio and video transcoding, AI training and inference, and enterprise applications (Java).

    Specifications

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline/burst bandwidth (Gbit/s)Packet forwarding rate (PPS)ConnectionsNIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENIDisk baseline IOPSDisk baseline/burst bandwidth (Gbit/s)
    ecs.c9ae.large242.5/burstable up to 25Up to 1,500,000Up to 500,0002366Up to 200,0002.5/burstable up to 20
    ecs.c9ae.xlarge484/burstable up to 25Up to 1,600,000Up to 500,000441515Up to 200,0003/burstable up to 20
    ecs.c9ae.2xlarge8166/burstable up to 25Up to 2,500,000Up to 500,000841515Up to 200,0004/burstable up to 20
    ecs.c9ae.4xlarge163210/burstable up to 25Up to 3,200,000Up to 500,0001683030Up to 200,0005.5/burstable up to 20
    ecs.c9ae.8xlarge326416/burstable up to 25Up to 5,000,000Up to 1,000,0003283030Up to 200,0008/burstable up to 20
    ecs.c9ae.12xlarge489625/none7,500,0001,500,0004883030150,00013/none
    ecs.c9ae.16xlarge6412832/none10,000,0002,000,0006483030200,00016/none
    ecs.c9ae.24xlarge9619250/none15,000,0003,000,00064153030300,00025/none
    ecs.c9ae.32xlarge12825664/none20,000,0004,000,00064153030400,00032/none
    ecs.c9ae.48xlarge192384100/none30,000,0006,000,00064155050600,00050/none

    c9a, compute-optimized instance family

    Built on the CIPU architecture with AMD EPYC Turin processors. Delivers consistent computing performance with chip-level security hardening.

    Use cases: Large and medium-sized databases, game servers, financial quantization, blockchain, web and application servers, and general-purpose enterprise applications.

    Specifications

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline/burst bandwidth (Gbit/s)Packet forwarding rate (PPS)ConnectionsNIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENIDisk baseline IOPSDisk baseline/burst bandwidth (Gbit/s)
    ecs.c9a.large242.5/burstable up to 15Up to 1,200,000Up to 500,0002366Up to 110,0002/burstable up to 15
    ecs.c9a.xlarge484/burstable up to 15Up to 1,400,000Up to 500,000441515Up to 110,0003/burstable up to 15
    ecs.c9a.2xlarge8166/burstable up to 15Up to 2,000,000Up to 500,000841515Up to 110,0004/burstable up to 15
    ecs.c9a.4xlarge163212/burstable up to 25Up to 3,000,000Up to 500,0001683030Up to 110,0005/burstable up to 15
    ecs.c9a.8xlarge326416/burstable up to 32Up to 4,000,000Up to 800,0003283030Up to 110,0008/burstable up to 15
    ecs.c9a.16xlarge6412832/none7,500,0001,500,0006483030120,00016/none

    c9i, compute-optimized instance family

    Built on the CIPU architecture with Intel Xeon 6 processors with P-cores (Granite Rapids). Delivers consistent computing performance with chip-level security hardening.

    Use cases: Machine learning inference, data analytics, batch computing, video encoding, frontend game servers, high-performance scientific and engineering applications, and web servers.

    Specifications

    The system may display different frequencies for this instance. The 3.6 GHz all-core turbo frequency is stable and guaranteed. The 3.9 GHz maximum single-core turbo frequency is a burst capability that depends on the overall CPU load of the physical server and is not covered by the Service-Level Agreement (SLA).

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline/burst bandwidth (Gbit/s)Packet forwarding rate (PPS)ConnectionsNIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENIDisk baseline/burst IOPSDisk baseline/burst bandwidth (Gbit/s)
    ecs.c9i.large242.5/burstable up to 151,000,000Up to 500,000236625,000/burstable up to 200,0002/burstable up to 10
    ecs.c9i.xlarge484/burstable up to 151,200,000Up to 500,00044151550,000/burstable up to 200,0002.5/burstable up to 10
    ecs.c9i.2xlarge8166/burstable up to 151,600,000Up to 500,00084151560,000/burstable up to 200,0004/burstable up to 10
    ecs.c9i.3xlarge122410/burstable up to 152,400,000Up to 500,000128151580,000/burstable up to 200,0005/burstable up to 10
    ecs.c9i.4xlarge163212/burstable up to 253,000,000500,0001683030100,000/burstable up to 200,0006/burstable up to 10
    ecs.c9i.6xlarge244815/burstable up to 254,500,000600,0002483030120,000/burstable up to 200,0007.5/burstable up to 10
    ecs.c9i.8xlarge326420/burstable up to 326,000,000800,0003283030200,000/burstable up to 300,00010/burstable up to 12
    ecs.c9i.12xlarge489625/burstable up to 329,000,0001,600,0004883030240,000/burstable up to 320,00012/burstable up to 15
    ecs.c9i.16xlarge6412828/burstable up to 3612,000,0002,000,0006483030300,000/burstable up to 400,00016/burstable up to 24
    ecs.c9i.24xlarge9619232/burstable up to 4818,000,0003,000,00064153030350,000/burstable up to 600,00020/burstable up to 28
    ecs.c9i.32xlarge12825636/burstable up to 5020,000,0004,000,00064153030400,000/burstable up to 650,00024/burstable up to 28
    ecs.c9i.48xlarge19238464/none24,000,0006,000,00064155050500,000/burstable up to 800,00032/none

    c8a, compute-optimized instance family

    Built on the CIPU architecture with chip-level security hardening for consistent computing performance.

    Use cases: Big data applications, web applications, AI training and inference, and audio and video transcoding.

    Specifications

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline/burst bandwidth (Gbit/s)Packet forwarding rate (PPS)ConnectionsNIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENIDisk baseline/burst IOPSDisk baseline/burst bandwidth (Gbit/s)
    ecs.c8a.large241.5/burstable up to 12.5900,000Up to 250,000236620,000/burstable up to 110,0001.5/burstable up to 10
    ecs.c8a.xlarge482.5/burstable up to 12.51,000,000Up to 250,00044151530,000/burstable up to 110,0002/burstable up to 10
    ecs.c8a.2xlarge8164/burstable up to 12.51,600,000Up to 250,00084151545,000/burstable up to 110,0002.5/burstable up to 10
    ecs.c8a.4xlarge16327/burstable up to 12.52,000,000300,000168303060,000/burstable up to 110,0003.5/burstable up to 10
    ecs.c8a.8xlarge326410/burstable up to 253,000,000600,000328303080,000/burstable up to 110,0005/burstable up to 10
    ecs.c8a.12xlarge489616/254,500,000750,0004883030120,000/none8/burstable up to 10
    ecs.c8a.16xlarge6412820/256,000,0001,000,0006483030160,000/none10/none
    ecs.c8a.24xlarge9619232/none9,000,0001,500,00064153030240,000/none16/none
    ecs.c8a.32xlarge12825640/none12,000,0002,000,00064153030320,000/none20/none
    ecs.c8a.48xlarge19238464/none18,000,0003,000,00064155050500,000/none32/none
    For ecs.c8a.large and ecs.c8a.xlarge, enable jumbo frames to reach the 12.5 Gbit/s burst network bandwidth.

    c8ae, performance-enhanced compute-optimized instance family

    Built on the CIPU architecture with chip-level security hardening for consistent computing performance. The "e" variant uses higher-frequency AMD processors for performance-sensitive workloads.

    Use cases:

    • AI workloads: deep learning, training, and inference

    • High-performance scientific computing (HPC)

    • Large and medium-sized databases, caches, and search clusters

    • MMO game servers

    • Enterprise applications with high performance requirements

    Specifications

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline/burst bandwidth (Gbit/s)Packet forwarding rate (PPS)vTPMConnectionsNIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENIDisk baseline/burst IOPSDisk baseline/burst bandwidth (Gbit/s)
    ecs.c8ae.large243/burstable up to 151,000,000YesUp to 300,000236630,000/burstable up to 200,0002/burstable up to 10
    ecs.c8ae.xlarge484/burstable up to 151,200,000YesUp to 300,00044151550,000/burstable up to 200,0002.5/burstable up to 10
    ecs.c8ae.2xlarge8166/burstable up to 151,600,000YesUp to 300,00084151560,000/burstable up to 200,0003/burstable up to 10
    ecs.c8ae.4xlarge163212/burstable up to 253,000,000Yes500,0001683030100,000/burstable up to 200,0006/burstable up to 10
    ecs.c8ae.8xlarge326420/burstable up to 256,000,000Yes1,000,0003283030200,000/none10/none
    ecs.c8ae.16xlarge6412832/none9,000,000Yes1,500,0006483030250,000/none16/none
    ecs.c8ae.32xlarge12825664/none18,000,000Yes3,000,00064153030500,000/none32/none
    For ecs.c8ae.large and ecs.c8ae.xlarge, enable jumbo frames to reach the 15 Gbit/s burst network bandwidth.

    c8i, compute-optimized instance family

    Built on the CIPU architecture with chip-level security hardening for consistent computing performance.

    Use cases: Machine learning inference, data analytics, batch computing, video encoding, frontend game servers, high-performance scientific and engineering applications, and web servers.

    Specifications

    The system randomly assigns one of the preceding processors when you purchase an instance. You cannot select a specific processor.

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline/burst bandwidth (Gbit/s)Packet forwarding rate (PPS)ConnectionsNIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENIDisk baseline/burst IOPSDisk baseline/burst bandwidth (Gbit/s)
    ecs.c8i.large242.5/burstable up to 151,000,000Up to 300,000236625,000/burstable up to 200,0002/burstable up to 10
    ecs.c8i.xlarge484/burstable up to 151,200,000Up to 300,00044151550,000/burstable up to 200,0002.5/burstable up to 10
    ecs.c8i.2xlarge8166/burstable up to 151,600,000Up to 300,00084151560,000/burstable up to 200,0004/burstable up to 10
    ecs.c8i.3xlarge122410/burstable up to 152,400,000Up to 300,000128151580,000/burstable up to 200,0005/burstable up to 10
    ecs.c8i.4xlarge163212/burstable up to 253,000,000350,0001683030100,000/burstable up to 200,0006/burstable up to 10
    ecs.c8i.6xlarge244815/burstable up to 254,500,000500,0002483030120,000/burstable up to 200,0007.5/burstable up to 10
    ecs.c8i.8xlarge326420/burstable up to 256,000,000800,0003283030200,000/none10/none
    ecs.c8i.12xlarge489625/none9,000,0001,000,0004883030300,000/none12/none
    ecs.c8i.16xlarge6412832/none12,000,0001,600,0006483030360,000/none20/none
    ecs.c8i.24xlarge9619250/none18,000,0002,000,00064153030500,000/none24/none
    ecs.c8i.48xlarge192512100/none30,000,0004,000,000641550501,000,000/none48/none

    c8ine, network-enhanced compute-optimized instance family

    Built on the CIPU architecture, optimized for network-intensive scenarios with stable computing performance.

    Use cases: Network access layer gateways, traffic and data forwarding, pre-processing middleware, large websites, e-commerce platforms, and AI applications as part of a cloud solution.

    Specifications

    • Compute:

      • CPU-to-memory ratio: 1:2

      • Intel Xeon Emerald Rapids or Intel Xeon Sapphire Rapids processors with a 2.7 GHz base clock and 3.2 GHz all-core turbo frequency

      • Hyper-Threading supported and enabled by default. To change this, see Change CPU options.

    • Storage:

    • Network:

    • Security: vTPM

    Instance types

    Instance typevCPUsMemory (GiB)Network bandwidth baseline/burst (Gbit/s)ConnectionsNIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENIEBS multi-queueDisk baseline/burst IOPSDisk bandwidth baseline/burst (Gbit/s)
    ecs.c8ine.large244/Up to 24600,000231010120,000/Up to 80,0002/Up to 8
    ecs.c8ine.xlarge487/Up to 281,200,000441515140,000/Up to 80,0002.5/Up to 8
    ecs.c8ine.2xlarge81612/Up to 352,000,000861515250,000/Up to 80,0004/Up to 8
    ecs.c8ine.4xlarge163223/Up to 443,500,0001683030280,000/Up to 100,0006/Up to 10
    ecs.c8ine.8xlarge326444/none7,000,00032830304100,000/none10/none

    c8y, compute-optimized instance family (Arm)

    Built on in-house Arm-based Yitian 710 processors and the fourth-generation SHENLONG architecture. Uses fast path acceleration on chips to improve storage, network, and computing performance.

    Use cases: Containers, microservices, websites, application servers, video encoding and decoding, HPC, and CPU-based machine learning.

    Specifications

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline/burst bandwidth (Gbit/s)Packet forwarding rate (PPS)ConnectionsNIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENIMax data disksDisk baseline/burst IOPSDisk baseline/burst bandwidth (Gbit/s)
    ecs.c8y.small121/10500,000Up to 250,0001233510,000/burstable up to 110,0001/burstable up to 10
    ecs.c8y.large242/10900,000Up to 250,0002366820,000/burstable up to 110,0001.5/burstable up to 10
    ecs.c8y.xlarge483/101,000,000Up to 250,000441515840,000/burstable up to 110,0002/burstable up to 10
    ecs.c8y.2xlarge8165/101,600,000Up to 250,0008415151650,000/burstable up to 110,0003/burstable up to 10
    ecs.c8y.4xlarge163210/253,000,000400,00016830301680,000/burstable up to 110,0005/burstable up to 10
    ecs.c8y.8xlarge326416/255,000,000750,000328303016125,00010
    ecs.c8y.16xlarge6412832/none10,000,0001,500,000648303032250,00016
    ecs.c8y.32xlarge12825664/none20,000,0003,000,0006415303032500,00032
    To use the ecs.c8y.32xlarge instance type, submit a ticket.

    c7a, compute-optimized instance family

    Built on the third-generation SHENLONG architecture with fast path acceleration on chips for improved storage, network, and computing performance.

    Use cases: Video encoding and decoding, high-PPS scenarios (live commenting, telecom data forwarding), web and game frontend servers, DevOps development and testing, data analytics and batch computing, high-performance scientific and engineering applications, and enterprise applications.

    Specifications

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline/burst bandwidth (Gbit/s)Packet forwarding rate (PPS)ConnectionsNIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENIDisk baseline/burst IOPSDisk baseline/burst bandwidth (Gbit/s)
    ecs.c7a.large241/burstable up to 10900,000Up to 250,000236612,500/burstable up to 110,0001/burstable up to 6
    ecs.c7a.xlarge481.5/burstable up to 101,000,000Up to 250,00044151520,000/burstable up to 110,0001.5/burstable up to 6
    ecs.c7a.2xlarge8162.5/burstable up to 101,600,000Up to 250,00084151530,000/burstable up to 110,0002/burstable up to 6
    ecs.c7a.4xlarge16325/burstable up to 102,000,000300,00088303060,000/burstable up to 110,0003/burstable up to 6
    ecs.c7a.8xlarge32648/burstable up to 103,000,000600,000167303075,000/burstable up to 110,0004/burstable up to 6
    ecs.c7a-nps1.8xlarge32648/burstable up to 103,000,000600,000167303075,000/burstable up to 110,0004/burstable up to 6
    ecs.c7a.16xlarge6412816/none6,000,0001,000,0003273030150,000/none8/none
    ecs.c7a-nps1.16xlarge6412816/none3,000,0001,000,0003273030150,000/none8/none
    ecs.c7a.32xlarge12825632/none12,000,0002,000,00032153030300,000/none16/none
    Ubuntu 16 and Debian 9 operating system kernels do not support AMD EPYC MILAN processors. Instances created from Ubuntu 16 or Debian 9 images cannot start.

    c7, compute-optimized instance family

    Built on the third-generation SHENLONG architecture with fast path acceleration on chips for improved storage, network, and computing performance.

    Use cases: High-PPS scenarios (live commenting, telecom data forwarding), MMO game frontend servers, web servers, data analytics, batch computing, video encoding, high-performance scientific and engineering applications, secure trusted computing, enterprise applications, and blockchain.

    Specifications

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline/burst bandwidth (Gbit/s)Packet forwarding rate (PPS)vTPMConnectionsNIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENIMax data disksDisk baseline/burst IOPSDisk baseline/burst bandwidth (Gbit/s)
    ecs.c7.large242/burstable up to 12.51,100,000YesUp to 500,0002366820,000/burstable up to 160,0001.5/burstable up to 10
    ecs.c7.xlarge483/burstable up to 12.51,100,000YesUp to 500,000441515840,000/burstable up to 160,0002/burstable up to 10
    ecs.c7.2xlarge8165/burstable up to 151,600,000YesUp to 500,0008415151650,000/burstable up to 160,0003/burstable up to 10
    ecs.c7.3xlarge12248/burstable up to 152,400,000YesUp to 500,0008815151670,000/burstable up to 160,0004/burstable up to 10
    ecs.c7.4xlarge163210/burstable up to 253,000,000Yes500,0008830301680,000/burstable up to 160,0005/burstable up to 10
    ecs.c7.6xlarge244812/burstable up to 254,500,000Yes550,000128303016110,000/burstable up to 160,0006/10
    ecs.c7.8xlarge326416/burstable up to 326,000,000Yes600,000168303024160,000/none10/none
    ecs.c7.16xlarge6412832/none12,000,000Yes1,200,000328303032360,000/none16/none
    ecs.c7.32xlarge12825664/none24,000,000Yes2,400,0003215303032600,000/none32/none

    c6r, compute-optimized instance family (Arm)

    Built on the third-generation SHENLONG architecture with fast path acceleration on chips for improved storage, network, and computing performance. Uses Arm-based Ampere Altra processors.

    Use cases: Containers and microservices, DevOps development and testing, websites and application servers, CPU-based machine learning and inference, and high-performance scientific and engineering applications.

    Specifications

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline/burst bandwidth (Gbit/s)Packet forwarding rate (PPS)ConnectionsNIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENIDisk baseline IOPSDisk baseline bandwidth (Gbit/s)
    ecs.c6r.large241/10900,000Up to 250,000236112,5001
    ecs.c6r.xlarge481/101,000,000Up to 250,0004415120,0001.5
    ecs.c6r.2xlarge8162/101,600,000Up to 250,0008415130,0002
    ecs.c6r.4xlarge16325/102,000,000300,0008830160,0003
    ecs.c6r.8xlarge32648/103,000,000600,00016730175,0004
    ecs.c6r.16xlarge6412816/none6,000,000900,000328301150,0008

    c6a, compute-optimized instance family

    Built on the SHENLONG architecture. Offloads virtualization features to dedicated hardware for predictable, consistent performance and reduced overhead.

    Use cases: Video encoding and decoding, high-PPS scenarios (live commenting, telecom data forwarding), web frontend servers, MMO game frontend servers, and DevOps development and testing.

    Specifications

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline/burst bandwidth (Gbit/s)Packet forwarding rate (PPS)ConnectionsNIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENIDisk baseline IOPSDisk baseline bandwidth (Gbit/s)
    ecs.c6a.large241/10900,000Up to 250,000226112,5001
    ecs.c6a.xlarge481.5/101,000,000Up to 250,0004315120,0001.5
    ecs.c6a.2xlarge8162.5/101,600,000Up to 250,0008415130,0002
    ecs.c6a.4xlarge16325/102,000,000300,0008830160,0003.1
    ecs.c6a.8xlarge32648/103,000,000600,00016730175,0004.1
    ecs.c6a.16xlarge6412816/none6,000,0001,000,000328301150,0008.2
    ecs.c6a.32xlarge12825632/none12,000,0002,000,0003215301300,00016.4

    c6e, performance-enhanced compute-optimized instance family

    Built on the third-generation SHENLONG architecture. Offloads virtualization features to dedicated hardware and uses fast path acceleration on chips for improved storage, network, and computing performance.

    Use cases: High-PPS scenarios (live commenting, telecom data forwarding), web frontend servers, MMO game frontend servers, data analytics, batch computing, video encoding, and high-performance scientific and engineering applications.

    Specifications

    For higher concurrent connections and network packet forwarding, consider the g7ne instance family.

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline/burst bandwidth (Gbit/s)Packet forwarding rate (PPS)ConnectionsNIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENIDisk baseline IOPSDisk baseline bandwidth (Gbit/s)
    ecs.c6e.large241.2/burstable up to 10900,000Up to 250,000236120,0001
    ecs.c6e.xlarge482/burstable up to 101,000,000Up to 250,0004415140,0001.5
    ecs.c6e.2xlarge8163/burstable up to 101,600,000Up to 250,0008415150,0002
    ecs.c6e.4xlarge16326/burstable up to 103,000,000300,0008830180,0003
    ecs.c6e.8xlarge326410/none6,000,000600,000168301150,0005
    ecs.c6e.13xlarge529616/none9,000,0001,000,000327301240,0008
    ecs.c6e.26xlarge10419232/none24,000,0001,800,0003215301480,00016

    c6, compute-optimized instance family

    Built on the SHENLONG architecture. Offloads virtualization features to dedicated hardware for predictable, consistent performance and reduced overhead.

    Use cases: High-PPS scenarios (live commenting, telecom data forwarding), web frontend servers, MMO game frontend servers, data analytics, batch computing, video encoding, and high-performance scientific and engineering applications.

    Specifications

    • Compute:

      • CPU-to-memory ratio: 1:2

      • 2.5 GHz Intel Xeon Platinum 8269CY (Cascade Lake) processors with turbo frequency of 3.2 GHz

      • Hyper-Threading supported and enabled by default. To change this, see Change CPU options.

    • Storage:

    A single instance in this family can deliver up to 200,000 IOPS.

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline/burst bandwidth (Gbit/s)Packet forwarding rate (PPS)ConnectionsNIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENIDisk baseline IOPSDisk baseline bandwidth (Gbit/s)
    ecs.c6.large241/burstable up to 3300,000Up to 250,000226110,0001
    ecs.c6.xlarge481.5/burstable up to 5500,000Up to 250,0004310120,0001.5
    ecs.c6.2xlarge8162.5/burstable up to 8800,000Up to 250,0008410125,0002
    ecs.c6.3xlarge12244/burstable up to 10900,000Up to 250,00086101030,0002.5
    ecs.c6.4xlarge16325/burstable up to 101,000,000300,0008820140,0003
    ecs.c6.6xlarge24487.5/burstable up to 101,500,000450,00012820150,0004
    ecs.c6.8xlarge326410/none2,000,000600,00016820160,0005
    ecs.c6.13xlarge529612.5/none3,000,000900,000327201100,0008
    ecs.c6.26xlarge10419225/none6,000,0001,800,0003215201200,00016

    Not recommended instance families

    The following families are previous-generation or limited-feature families. If they are sold out, use a recommended family listed above (c6, c6e, c7, or newer).

    c5, compute-optimized instance family

    Use cases: High-PPS scenarios, web frontend servers, MMO game frontend servers, data analytics, batch computing, video encoding, and high-performance scientific and engineering applications.

    Specifications:

    • CPU-to-memory ratio: 1:2

    • 2.5 GHz Intel Xeon Platinum 8163 (Skylake) or 8269CY (Cascade Lake) processors

    • I/O optimized

    • Supported disk types: ESSDs, ESSD AutoPL disks, standard SSDs, and ultra disks. For an overview, see Block Storage.

    • IPv4 and IPv6. For details, see IPv6 communication.

    Instances may be deployed on different server platforms. For a consistent platform, use c6, c6e, or c7 instead. A single instance can deliver up to 200,000 IOPS.

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline bandwidth (Gbit/s)Packet forwarding rate (PPS)NIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENI
    ecs.c5.large241300,0002261
    ecs.c5.xlarge481.5500,00023101
    ecs.c5.2xlarge8162.5800,00044101
    ecs.c5.3xlarge12244900,00046101
    ecs.c5.4xlarge163251,000,00048201
    ecs.c5.6xlarge24487.51,500,00068201
    ecs.c5.8xlarge3264102,000,00088201
    ecs.c5.16xlarge64128204,000,000168201

    ic5, compute-intensive instance family

    Use cases: Web frontend servers, data analytics, batch computing, video encoding, high-PPS scenarios, and MMO game frontend servers.

    Specifications:

    • CPU-to-memory ratio: 1:1

    • 2.5 GHz Intel Xeon Platinum 8163 (Skylake) or 8269CY (Cascade Lake) processors with an all-core turbo frequency of 2.7 GHz

    • I/O optimized

    • Supported disk types: ESSDs, ESSD AutoPL disks, standard SSDs, and ultra disks. For an overview, see Block Storage.

    • IPv4 only (no IPv6 support)

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline bandwidth (Gbit/s)Packet forwarding rate (PPS)NIC queuesENIsPrivate IPv4 addresses per ENI
    ecs.ic5.large221300,000226
    ecs.ic5.xlarge441.5500,0002310
    ecs.ic5.2xlarge882.5800,0002410
    ecs.ic5.3xlarge12124900,0004610
    ecs.ic5.4xlarge161651,000,0004820
    ecs.ic5.6xlarge24247.51,500,0006820
    ecs.ic5.8xlarge3232102,000,0008820
    ecs.ic5.16xlarge6464203,000,00016820

    sn1ne, network-enhanced compute-optimized instance family

    Use cases: High-PPS scenarios (live commenting, telecom data forwarding), web frontend servers, MMO game frontend servers, data analytics, batch computing, video encoding, and high-performance scientific and engineering applications.

    Specifications:

    • CPU-to-memory ratio: 1:2

    • 2.5 GHz Intel Xeon E5-2682 v4 (Broadwell), Platinum 8163 (Skylake), or 8269CY (Cascade Lake) processors

    • I/O optimized

    • Supported disk types: standard SSDs and ultra disks only. For an overview, see Block Storage.

    • IPv4 and IPv6. For details, see IPv6 communication.

    Instances may be deployed on different server platforms. For a consistent platform, use c6, c6e, or c7 instead.

    Instance types

    Instance typevCPUsMemory (GiB)Network baseline bandwidth (Gbit/s)Packet forwarding rate (PPS)NIC queuesENIsPrivate IPv4 addresses per ENIIPv6 addresses per ENI
    ecs.sn1ne.large241300,0002261
    ecs.sn1ne.xlarge481.5500,00023101
    ecs.sn1ne.2xlarge81621,000,00044101
    ecs.sn1ne.3xlarge12242.51,300,00046101
    ecs.sn1ne.4xlarge163231,600,00048201
    ecs.sn1ne.6xlarge24484.52,000,00068201
    ecs.sn1ne.8xlarge326462,500,00088201