All Products
Search
Document Center

MaxCompute:2024

Last Updated:Mar 20, 2026

This topic describes MaxCompute announcements for 2024. Learn about new features, syntax enhancements, and permission changes to improve your project development efficiency.

December 23, 2024 — MaxFrame distributed computing framework adds support for international regions

Dear MaxCompute user,

The MaxCompute distributed computing framework MaxFrame will support Singapore, Japan (Tokyo), and Indonesia (Jakarta) starting Monday, December 23, 2024 (UTC+8).

MaxFrame is a distributed computing framework developed by Alibaba Cloud. It provides a Python programming interface and uses MaxCompute compute resources and data interfaces directly. MaxFrame supports Pandas and XGBoost APIs and automatically handles distributed processing. This enables Python developers to process large-scale data, perform ML/AI data preprocessing, and develop models more efficiently and conveniently on MaxCompute. For details, see Distributed AI Computing Engine MaxFrame.

If you have questions or need help, contact us through the DingTalk group application form to join the MaxCompute Developer Community (DingTalk group ID: 11782920), your dedicated DingTalk group, or the official MaxFrame user support group (DingTalk group ID: 37130012987).

Thank you for your cooperation!

November 28, 2024 — Update to MaxCompute Service Level Agreement (SLA)

Dear MaxCompute user,

Alibaba Cloud will update the Service Level Agreement (SLA) for MaxCompute, a cloud-native big data computing service. The changes are as follows:

  • Add definitions of service unavailability (Sections 1.1–1.10).

  • Add exceptions (Subsections (k), (l), and (m) of Section 1.11).

For SLA details, see MaxCompute Service Level Agreement.

The updated SLA takes effect on Wednesday, January 1, 2025 (UTC+8). If you have questions, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your support and attention to Alibaba Cloud!

November 14, 2024 — MaxCompute LogView security upgrade

Dear MaxCompute user,

MaxCompute will begin a full security upgrade for LogView in the following regions starting at 2:00 PM on Monday, December 2, 2024 (UTC+8): China (Hong Kong), China East 2 Finance Cloud, China North 2 Finance Cloud (invitational preview), China North 2 Alibaba Gov Cloud 1, China South 1 Finance Cloud, Japan (Tokyo), Singapore, Malaysia (Kuala Lumpur), Indonesia (Jakarta), Germany (Frankfurt), UK (London), US (Silicon Valley), US (Virginia), UAE (Dubai), and SAU (Riyadh - Partner Region).

  • Have an Alibaba Cloud account: Users must hold a valid Alibaba Cloud account and complete identity verification.

  • Be a member of the job’s MaxCompute project: A project administrator can add users to the project.

Note that if the DataWorks workspace’s MaxCompute data source access identity differs from the user’s identity, the user may be unable to access the LogView for their own submitted jobs. This occurs when the user is not added to the MaxCompute project where the job runs. Example: In standard mode, a user submits a job from the development environment to the production environment. If the user is not added to the production environment’s MaxCompute project, they cannot access the LogView for their job, even though it belongs to the production project.

To avoid this, ensure the following:

  • All team members who need LogView access register an Alibaba Cloud account and complete identity verification.

  • Add team members to the relevant MaxCompute projects and grant them required permissions on specified objects.

    Note

    For more information about user management and authorization, see User Planning and Management and MaxCompute Permissions.

After the upgrade, LogView will no longer show job execution results by default.

If you encounter issues accessing LogView or configuring permissions during or after the upgrade, or need technical support, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

November 13, 2024 — MaxCompute intelligent job diagnostics feature release

Dear MaxCompute user,

The MaxCompute intelligent job diagnostics feature will roll out across all regions starting Wednesday, November 13, 2024 (UTC+8). This feature analyzes SQL jobs based on detailed execution information:

  • For failed jobs: Matches error messages to error categories and provides error descriptions and solutions.

  • For slow-running jobs: Analyzes resource consumption and execution details to identify anomalies such as insufficient resources, data skew, data bloat, or plan regression. It also explains causes and provides solutions.

This feature helps you quickly detect job failures or unexpected run times and resolve them independently. For more information and usage instructions, see Intelligent Job Diagnostics.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

October 30, 2024 — MaxCompute data transmission service upgrade and cross-region traffic governance

MaxCompute will upgrade its data transmission service and implement cross-region traffic governance starting Monday, November 18, 2024 (UTC+8). Details follow:

  • Data tunnel read/write performance and traffic control

    • Streaming Tunnel: The per-concurrent slot (Slot) write rate limit increases from 1 MB/s to 10 MB/s. The per-Slot write request limit increases from 10 requests per second to 100 requests per second. Batch Tunnel: The per-Slot write rate limit is set to 10 MB/s.

    • Streaming Tunnel Slot usage calculation improves. Instead of specifying the number of Slots when creating a StreamUploadSession, each RecordPack now occupies one Slot and releases it upon completion. This matches Batch Tunnel behavior. For details, see Data Transmission Service Overview.

  • Cross-region traffic governance

    Connections to a MaxCompute project using an endpoint outside the project’s region are prohibited. Use the endpoint or tunnel endpoint for the project’s region. For endpoint details, see Endpoint.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

October 28, 2024 — Impact of dynamic-to-static partition optimization on UPDATE/DELETE/MERGE INTO operations

Dear MaxCompute user,

Starting Thursday, October 24, 2024 (UTC+8), MaxCompute enables dynamic-to-static partition optimization for all DML operations by default to improve query performance. This optimization changes behavior for UPDATE/DELETE/MERGE INTO operations. When you run these operations on non-existent partitions, you may see errors like:

  • ODPS-0110061: Failed to run ddltask - Operation DELETE expect target partition exist, but object not found.

  • ODPS-0422141: Partition not found - Partition 'projectName.tableName(pt=xxx)' does not exist.

If this occurs, disable the optimization at the session level by setting SET odps.optimizer.cbo.rule.filter.black=re;. Then run UPDATE/DELETE/MERGE INTO operations normally.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

October 22, 2024 — MaxCompute Go SDK launch

Dear MaxCompute user,

MaxCompute officially launched its Go SDK on Tuesday, October 22, 2024 (UTC+8). The Go SDK provides a Go programming interface for MaxCompute. You can use Go code to operate and manage MaxCompute services, such as accessing and managing projects, working with tables, and handling data transmission. For more information, see Go SDK Overview.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

October 15, 2024 — Discontinuation of Alibaba Cloud classic network connection and endpoint updates

Dear MaxCompute user,

The Alibaba Cloud classic network connection method for MaxCompute will be discontinued on Saturday, November 30, 2024 (UTC+8). Replace it with the Alibaba Cloud VPC network connection or Alibaba Cloud product interconnection network. Confirm and switch promptly to avoid business impact from network interruptions. Also, endpoints for Alibaba Cloud VPC or product interconnection networks in Finance Cloud, Alibaba Gov Cloud, and other international regions have been updated. Review and update your configurations. For endpoint details, see Endpoint.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

October 14, 2024 — Support for multiple subscription compute resources in the same region

Dear MaxCompute user,

Starting Monday, October 14, 2024 (UTC+8), MaxCompute will support purchasing multiple subscription compute resources in the same region under one account. This means one Alibaba Cloud account can purchase multiple top-level quotas in the same region, with resources isolated between quotas.

This feature optimizes the following scenarios:

  • Strong isolation of subscription compute resources for different businesses under one account.

  • Splitting billing for different businesses under one account. Previously, MaxCompute subscription compute resources could not be split-billed using sub-quotas. With this feature, purchase multiple top-level quotas to split billing by order.

For more information about MaxCompute quotas, see Compute Resources — Quota Management.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

September 27, 2024 — MaxCompute storage cost optimization feature release

Dear MaxCompute user,

The MaxCompute storage cost optimization feature will roll out across public cloud regions in China starting Friday, September 27, 2024 (UTC+8). This feature recommends changing table or partition storage classes from Standard to Infrequent Access (IA) or Archive based on fixed rules: — Standard storage with no access in the last 30 days → IA storage — Standard storage with no access in the last 180 days → Archive storage It uses the most recent access time for tables or partitions and estimates potential storage cost savings (list price) to help you configure tiered storage policies.

For more information and usage instructions, see Tiered Storage Configuration Optimization.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

September 26, 2024 — MaxCompute MaxFrame distributed computing framework goes commercial

Dear MaxCompute user,

MaxFrame is a distributed computing framework developed by Alibaba Cloud. It provides a Python programming interface and uses MaxCompute compute resources and data interfaces directly. MaxFrame supports Pandas and XGBoost APIs and automatically handles distributed processing. This enables Python developers to process large-scale data and perform ML/AI data preprocessing more efficiently and conveniently on MaxCompute. For details, see:

Note

MaxCompute MaxFrame will gradually replace PyODPS DataFrame and Mars interfaces. It offers significant improvements in operator compatibility and distributed performance. New users and new jobs should use MaxFrame for Python development. We recommend migrating existing PyODPS DataFrame and Mars jobs to MaxFrame for better development experience and service support.

The MaxCompute MaxFrame distributed computing framework goes commercial on Friday, September 27, 2024 (UTC+8) in China (Beijing), China (Hangzhou), China (Shanghai), China (Shenzhen), China (Chengdu), China (Ulanqab), and China (Hong Kong). After going commercial, MaxFrame charges based on compute resource usage for jobs. It supports pay-as-you-go and subscription billing. For billing details, see Compute Fees (Subscription).

If you have questions or need help, contact us through the DingTalk group application form to join the MaxCompute Developer Community (DingTalk group ID: 11782920), your dedicated DingTalk group, or the official MaxFrame user support group (DingTalk group ID: 37130012987).

Thank you for your cooperation!

September 20, 2024 — RAM permission controls added to MaxCompute console features

Dear MaxCompute user,

To improve security, MaxCompute will add Resource Access Management (RAM) permission controls to some console features on Tuesday, October 8, 2024. By default, the Alibaba Cloud account (root account) has access or operation permissions for all console features. RAM users need specific permissions to access features. The features and corresponding RAM permissions added are listed below:

Module

Feature

Permission Action

Workspace > Notebook

Notebook

  • odps:CreateNotebookTemplate

  • odps:ListNotebookTemplates

  • odps:GetNotebookTemplate

  • odps:UpdateNotebookTemplate

  • odps:DeleteNotebookTemplate

  • odps:CreateNotebookStorage

  • odps:ListNotebookStorage

  • odps:CreateNotebookInstance

  • odps:ListNotebookInstances

  • odps:GetNotebookInstance

  • odps:StartNotebookInstance

  • odps:StopNotebookInstance

  • odps:UpdateNotebookInstance

  • odps:DeleteNotebookInstance

Tenant Management > External Data Sources

External Data Sources

  • odps:ListTenantObjectBindings

  • odps:UpdateTenantObjectBindings

Data Transmission > Migration Service (MMA)

Migration Service (MMA)

  • odps:ListMmsDataSources

  • odps:GetMmsDataSource

  • odps:CreateMmsDataSource

  • odps:UpdateMmsDataSource

  • odps:DeleteMmsDataSource

  • odps:CreateMmsFetchMetadataJob

  • odps:ListMmsJobs

  • odps:GetMmsJob

  • odps:CreateMmsJob

  • odps:DeleteMmsJob

  • odps:StartMmsJob

  • odps:StopMmsJob

  • odps:RetryMmsJob

  • odps:ListMmsTasks

  • odps:GetMmsTask

  • odps:ListMmsTaskLogs

  • odps:GetMmsAsyncTask

  • odps:UpdateMmsAsyncTask

  • odps:DeleteMmsAsyncTask

  • odps:ListMmsDbs

  • odps:GetMmsDb

  • odps:ListMmsTables

  • odps:GetMmsTable

  • odps:ListMmsPartitions

  • odps:GetMmsPartition

  • odps:ListMmsAgents

  • odps:CreateMmsAuthFile

For details about permission actions, see RAM Permissions.

To avoid losing access to console features after the update, add the required RAM permissions to RAM users before Tuesday, October 8, 2024.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

September 19, 2024 — MaxCompute new features from Apsara Conference go into invitational preview

Dear MaxCompute user,

At the 2024 Apsara Conference, MaxCompute launched several new features in invitational preview. These include Lakehouse 2.0 with OpenLake support, Object Tables for processing unstructured data using SQL or MaxFrame, Delta Table incremental formats, incremental computing based on incremental materialized views, and MCQA 2.0 SQL engine query acceleration. These features will be available for trial in the China (Hangzhou), China (Shanghai), China (Beijing), and China (Shenzhen) public cloud regions. If you want to learn more, register using the New Features Invitational Preview Application Form.

After registration, try the new features in two ways:

  • In the MaxCompute console, select Project Type as Apsara Conference New Features Invitational Preview Project when creating a new project.

  • In the public preview version of DataStudio (Data Development) in DataWorks, select Project Type as Apsara Conference New Features Invitational Preview Project when creating an internal project while binding MaxCompute compute resources.

Important
  • Invitational preview features are only available in new projects marked as Apsara Conference invitational preview. Because features are still under iteration and improvement, invitational preview projects and features do not guarantee production readiness. Avoid running critical data or business workloads on them.

  • After the invitational preview ends, Apsara Conference invitational preview projects will automatically convert to standard production projects. Data and configuration in these projects will remain intact. MaxCompute will refine and iterate these features based on feedback, but does not guarantee all features will be included in standard projects.

September 5, 2024 — MaxCompute console quota (Quota) management UI optimization

Dear MaxCompute user,

MaxCompute will roll out an optimized UI for quota (Quota) management in the console across all regions starting Wednesday, September 11, 2024. This version optimizes for two scenarios: “No time-based scaling needed” and “Time-based scaling needed.” If you do not need time-based scaling, use the Basic Configuration tab to manage subscription compute quotas. If you need time-based scaling, first complete Basic Configuration, then go to the Scaling Configuration tab for related settings.

When you configure subscription compute quotas on the Quota Management page in the MaxCompute console, you will see these changes:

  • The former Quota Template is renamed Basic Configuration, with these capabilities:

    • Use Edit Basic Configuration to add sub-quotas and edit basic parameters: Exclusive, Per-job CU Limit, Enable Priority, Non-reserved CUs, and Scheduling Policy. You can also edit Reserved CU Range [minCU, maxCU] and Elastic Reserved CUs. However, these edits apply only to the currently active quota configuration (formerly Quota Plan). If you use time-based scaling, go to the Scaling Configuration tab to configure Reserved CU Range [minCU, maxCU] and Elastic Reserved CUs.

    • In the Quota list on the Basic Configuration tab, delete sub-quotas. This works the same as the original Quota template deletion logic.

  • The former Quota Plan and Time Plan are merged into Scaling Configuration, which now manages only Reserved CU Range [minCU, maxCU] and Elastic Reserved CUs. Capabilities include the following:

    • New configuration scheme: Formerly Add Plan. Supports only configuring Reserved CU Range [minCU, maxCU] and Elastic Reserved CUs.

    • Configuration scheme list: Formerly Quota plan list. Now supports expanding each scheme to view and compare configurations.

    • Time-based management: Formerly Time Plan. Functionality remains unchanged.

    • Scaling preview: New feature. Select a sub-quota to view CU allocation trend charts for each time period.

    • Global scaling view: New feature. Shows all configuration schemes in a table format, making it easy to review Min/Max CU amounts and Elastic Reserved CUs for all quotas across time periods.

For detailed usage instructions, see Compute Resources — Quota Management. If you have questions or need help, contact us through your dedicated DingTalk group, the official MaxFrame user support group (DingTalk group ID: 37130012987), or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

August 29, 2024 — RAM permission controls added to MaxCompute console features

Dear MaxCompute user,

To improve security, MaxCompute will add Resource Access Management (RAM) permission controls to some console features on Tuesday, September 10, 2024. By default, the Alibaba Cloud account (root account) has access or operation permissions for all console features. RAM users need specific permissions to access features. The features and corresponding RAM permissions added are listed below:

Module

Feature

Permission Action

Workspace > Project Management

Outbound Network Configuration

odps:ListOutboundInternetAddress

Workspace > Project Management

Role Permissions

odps:UpdateUsersToRole

Workspace > Quota Management

Quota Rules

  • odps:ListQuotaRoutingRules

  • odps:CreateQuotaRoutingRule

Workspace > Resource Monitoring

Storage Resources

  • odps:GetStorageSizeSummary

  • odps:GetStorageAmountSummary

  • odps:GetStorageSummaryCompared

  • odps:ListStorageProjectsInfo

  • odps:ListStorageTablesInfo

  • odps:ListStoragePartitionsInfo

Job Performance Monitoring

odps:ListTopJobInfo

Data Transmission Service

  • odps:GetTableAccessInfoTopK

  • odps:GetTableIpAccessInfoTopK

  • odps:GetTableAccessInfo

  • odps:ListTableSlotDetail

  • odps:GetTunnelThroughputSummary

Workspace > Job O&M

Job O&M

odps:ListJobSnapshotInfos

Workspace > Materialized Views

Materialized Views

  • odps:ListGlobalConfig

  • odps:GetGlobalConfig

  • odps:CloseGlobalConfig

  • odps:UpdateGlobalConfig

  • odps:AddMvRecommendationSupportProject

  • odps:ListMvRecommendationSupportProjects

  • odps:CheckMvRecommendationSupportProjects

  • odps:RemoveMvRecommendationSupportProject

  • odps:CreateMaterializedView

  • odps:GetMaterializedViewStatus

  • odps:ListMaterializedViews

  • odps:GetMaterializedView

  • odps:UpdateMaterializedView

  • odps:DeleteMaterializedView

  • odps:ListProjectMvRecommendations

  • odps:GetProjectMvRecommendation

  • odps:ListMvRecommendations

  • odps:ListMvRecommendationsByProject

  • odps:GetMvRecommendation

  • odps:GetMvRecommendationByProject

  • odps:ListMvRecommendationJobInfo

  • odps:ListMaterializedViewJobInfo

Tenant Management > Network Connectivity

Network Connectivity

  • odps:ListNetworkLinks

  • odps:CreateNetworkLink

  • odps:GetNetworkLink

  • odps:RemoveNetworkLink

Tenant Management > Image Management

Custom Images

  • odps:ListImages

  • odps:AddImage

  • odps:GetImage

  • odps:RemoveImage

For details about permission actions, see RAM Permissions.

To avoid losing access to console features after the update, add the required RAM permissions to RAM users before Tuesday, September 10, 2024. For example, to give a RAM user quota management permissions, configure the following RAM policy:

{
  "Version": "1",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": [
        "odps:UpdateQuota",
        "odps:UpdateQuotaPlan",
        "odps:UpdateSubQuotas",
        "odps:UpdateQuotaSchedule",
        "odps:CreateQuotaPlan",
        "odps:DeleteQuotaPlan",
        "odps:CreateQuotaSchedule",
        "odps:ListQuotaRoutingRules",
        "odps:CreateQuotaRoutingRule",
        "odps:GetQuotaRoutingRule",
        "odps:RemoveQuotaRoutingRule",
        "odps:UpdateQuotaRoutingRule"         
      ],
      "Resource": "*"
    }
  ]
}

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

August 19, 2024 — MaxCompute resource monitoring feature upgrade

Dear MaxCompute user,

MaxCompute resource monitoring adds new metrics for storage resources and job performance monitoring.

  • Storage resources

    View storage usage trends and distribution details. See total storage usage and the percentage of each storage class in your selected region. Monitor storage changes over time by project and time range. View detailed storage information for tables or partitions. Identify abnormal storage spikes and optimize top-consuming projects or tables/partitions.

  • New job performance monitoring metrics

    • Show the top 10 jobs consuming the most CPU time or memory time within your filter range. Show the top 10 Signatures and ExtNodeIds with the highest total or average CPU time or memory time.

    • Show the top 10 jobs with the longest total runtime within your filter range. Show the top 10 Signatures and ExtNodeIds with the highest total or average runtime.

    • View job scan volume trends over time.

    • View trends for job scan volume processed per CU-hour over time.

    Use these metrics to quickly locate high-resource or long-running jobs and take action. Compare job scan volume per CU-hour to identify performance anomalies. Use job O&M features to find root causes and resolve issues.

For more information and usage instructions, see Resource Monitoring.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

July 31, 2024 — PyTorch algorithm service retirement

Dear MaxCompute user,

Due to system upgrades, Platform for AI (PAI) will retire legacy PyTorch algorithm components in all clusters on August 30, 2024. If you submit PyTorch jobs to MaxCompute using PAI commands like pai -name pytorch100/pytorch131/pytorch151, migrate them promptly. We recommend using PAI-DLC to submit PyTorch jobs. For details, see Create Training Jobs. Starting August 31, 2024, legacy PyTorch algorithm component jobs will no longer be covered by the Service Level Agreement (SLA).

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

July 18, 2024 — Retirement of some OpenAPIs

Dear MaxCompute user,

MaxCompute will retire some OpenAPIs on Wednesday, July 31, 2024 (UTC+8). Details follow:

  • Retired OpenAPIs and replacement OpenAPIs

    Retired OpenAPI version and name

    Replacement OpenAPI version and name

    Version

    OpenAPI Name

    2019-06-12

    CreateProject

    2022-01-04: CreateProject

    DeleteProject

    Not available

    GetProject

    2022-01-04: GetProject

    OpenMaxComputeService

    Not available

    2022-01-04

    GetQuota

    2022-01-04: QueryQuota

  • Reason for retirement

    • The OpenAPIs from the 2019-06-12 version are retired because their functionality has been replaced and no external customers have called them for a long time.

    • The 2022-01-04 version’s GetQuota API is retired because its response format is being updated to provide more complete quota information, and no external customers have called it for a long time.

  • Impact of Going Offline

    After retirement, calls to these OpenAPIs will return InvalidApi.NotFound or InvalidAction.NotFound error codes.

  • Replacement or action items

    If your code depends on these OpenAPI SDKs (specific requests, responses, or related code), dependency issues may occur during future SDK upgrades. Replace or remove the affected methods. If your code directly depends on these OpenAPIs, replace or remove them.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

July 4, 2024 — Job O&M adds job start time, wait time, and runtime parameters

Dear MaxCompute user,

MaxCompute adds job start time, wait time, and runtime parameters to the job O&M list and job analysis overview page. Parameter descriptions:

  • Start time: When the job receives its first compute resource. For very short jobs or jobs that do not consume compute resources (such as DDL statements), this is the job submission time.

  • Wait time: Time between job submission and start time.

  • Execution Duration: the duration from the start time to the end time of the job.

Use these parameters to quickly determine if long wait times cause overall job slowness. Then use job analysis to check if the issue is due to insufficient resources or multiple jobs competing for resources. Optimize job execution, adjust job priority, or manage compute resources based on your business needs.

For more information and usage instructions, see Job O&M.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

July 3, 2024 — MaxFrame feature release

Dear MaxCompute user,

The distributed computing framework MaxFrame is released in China (Shanghai), China (Beijing), China (Ulanqab), China (Shenzhen), and China (Chengdu) on Wednesday, July 3, 2024 (UTC+8). Other regions will follow.

MaxFrame is a distributed computing framework developed by Alibaba Cloud. It provides a Python programming interface and uses MaxCompute compute resources and data interfaces directly. This enables Python developers to process large-scale data and develop AI models more efficiently and conveniently. MaxFrame fully supports Pandas APIs and handles distributed processing automatically. You can use MaxCompute’s massive compute resources and data for data processing, visual exploration, scientific computing, and ML/AI development in a familiar and efficient way.

For MaxFrame feature details and usage instructions, see Distributed AI Computing Engine MaxFrame.

If you have questions or need help, contact us through your dedicated DingTalk group, the official MaxFrame user support group (DingTalk group ID: 37130012987), or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

Compute Cost Optimization Feature Released in China (Hong Kong), Singapore, and Two Other Regions on July 1, 2024

Dear MaxCompute user:

The MaxCompute compute cost optimization feature is now available in the China (Hong Kong), Singapore, Indonesia (Jakarta), and Germany (Frankfurt) regions. This feature generates optimized resource configuration plans for subscription primary Quota compute resources based on your actual job requests and desired resource configurations. It also provides recommended resource configuration plans and projected outcomes for switching projects from pay-as-you-go to subscription compute resources. This helps you further optimize compute costs and improve resource utilization efficiency. After this release, the feature is available in the following ten regions: China (Hangzhou), China (Shanghai), China (Beijing), China (Zhangjiakou), China (Shenzhen), China (Chengdu), China (Hong Kong), Singapore, Indonesia (Jakarta), and Germany (Frankfurt).

For more information and usage instructions, see Compute Resource Configuration Optimization.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

May 17, 2024 — MaxCompute LogView security upgrade

Dear MaxCompute user,

MaxCompute will begin a phased security upgrade for LogView starting at 14:00 on Monday, June 3, 2024 (UTC+8) in the following regions: China (Chengdu), China (Zhangjiakou), China (Shenzhen), China (Beijing), China (Shanghai), China (Hangzhou), and China (Ulanqab). After the security upgrade, LogView links generated by jobs submitted from Data Studio and DataWorks data analytics jobs using shared resource groups for scheduling will redirect you to the MaxCompute management console. To access the LogView for the job's project, you must satisfy all the following conditions:

  • Have an Alibaba Cloud account: Users must hold a valid Alibaba Cloud account and complete identity verification.

  • Be a member of the job’s MaxCompute project: A project administrator can add users to the project.

Note that if the DataWorks workspace’s MaxCompute engine access identity differs from the user’s identity, the user may be unable to access the LogView for their own submitted jobs. This occurs when the user is not added to the MaxCompute project where the job runs. Example: In standard mode, a user submits a job from the development environment to the production environment. If the user is not added to the production environment’s MaxCompute project, they cannot access the LogView for their job, even though it belongs to the production project.

To avoid this, ensure the following:

  • All team members who need LogView access register an Alibaba Cloud account and complete identity verification.

  • Add team members to the relevant MaxCompute projects and grant them required permissions on specified objects.

    Note

    For more information about user management and authorization, see User Planning and Management and MaxCompute Permissions.

If you encounter any issues with Logview access, permission configuration, or require technical support during or after the upgrade, you can contact us through your dedicated DingTalk group or the MaxCompute Developer Community group (DingTalk group ID: 11782920).
Thank you for your cooperation!

May 13, 2024 — MaxCompute ETCD certificate upgrade

Dear MaxCompute user,

MaxCompute will upgrade ETCD certificates for the Singapore region cluster between 10:00 AM and 4:00 PM on Monday, May 20, 2024 (UTC+8). The upgrade is expected to not affect job execution. In rare cases, if a job fails, retry it.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

May 8, 2024 — MaxCompute job performance monitoring feature release

Dear MaxCompute user,

The MaxCompute job performance monitoring feature launches at 8:00 PM on Wednesday, May 8, 2024 (UTC+8). Features include viewing trends for job count, CU usage, and job runtime over time.

Note

All metrics support filtering by time range, project, and quota. They also support grouping by job type, status, project, and quota.

Use these metrics to identify abnormal spikes in job count or resource usage and take action. Compare job runtimes across time periods to spot unusually slow jobs that degrade overall performance.

For more information and usage instructions, see Job Performance Monitoring.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

April 30, 2024 — MaxCompute storage service upgrade

Dear MaxCompute user,

MaxCompute will upgrade storage in the China East 2 Finance Cloud (Shanghai) region between 10:00 AM and 4:00 PM on Wednesday, May 15, 2024 (UTC+8) (low-risk change). The upgrade is expected to not affect job execution. In rare cases, if a job fails, retry it.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

April 23, 2024 — Lifecycle changes for MaxCompute projects in international regions

Dear MaxCompute user,

Starting Thursday, April 25, 2024 (UTC+8), all new pay-as-you-go standard service projects in China (Hong Kong), Japan (Tokyo), Singapore, Malaysia (Kuala Lumpur), Indonesia (Jakarta), Germany (Frankfurt), UK (London), US (Silicon Valley), US (Virginia), UAE (Dubai), and SAU (Riyadh - Partner Region) will align their lifecycles with the pay-as-you-go standard service lifecycle. All project storage, downloads, and other pay-as-you-go billing details will appear under MaxCompute (Pay-as-you-go). Usage details will also use the MaxCompute (Pay-as-you-go) metering specification. If you purchase subscription resources in these regions, only prepayment resources align with the subscription instance lifecycle.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

April 7, 2024 — New discounts for MaxCompute subscription products

Dear MaxCompute user,

New discounts for MaxCompute subscription products take effect at 10:00 AM on April 7, 2024 (UTC+8). Discounts apply to standard compute resources, Data Transmission Service, and other subscription resource specifications. Discount rates:

One-year purchases get 15% off. Two-year purchases get 30% off. Three-year purchases get 50% off.

March 14, 2024 — MaxCompute supports migrating ActionTrail events to MaxCompute

Dear MaxCompute user,

MaxCompute supports creating ActionTrail trails to migrate operation events to MaxCompute.

When you need to analyze operation events older than 90 days, create an ActionTrail trail to deliver events to MaxCompute for querying and analysis.

For more information and usage instructions, see Migrate ActionTrail Events to MaxCompute.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

February 29, 2024 — DataWorks exclusive resource group PyODPS node upgrade

Dear MaxCompute user,

DataWorks will upgrade exclusive resource groups across regions between Monday, April 8, 2024 (UTC+8) and Thursday, May 30, 2024 (UTC+8) to enhance PyODPS support. Upgrade details and regional schedules are in the table below. After the upgrade, the PyODPS package version for nodes remains unchanged.

Upgrade Content

Region

Upgrade Date

  • Add load_resource_package support to DataWorks PyODPS nodes.

  • Add multi-process support.

  • Add automatic error prompts.

China (Chengdu), China (Zhangjiakou)

April 8–11, 2024

China (Shenzhen)

April 15–18, 2024

China (Beijing)

April 22–25, 2024

China (Shanghai)

May 6–9, 2024

China (Hangzhou)

May 13–16, 2024

China (Hong Kong), international regions

May 20–23, 2024

China East 1 Finance Cloud, China East 2 Finance Cloud

May 27–30, 2024

Upgrades may affect running PyODPS jobs. If you encounter PyODPS job failures on exclusive resource groups or other issues, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

February 28, 2024 — MaxCompute single-node downtime drill

Dear MaxCompute user,

MaxCompute will conduct reliability drills for access services in China (Zhangjiakou) and China (Chengdu) between 3:00 PM and 4:00 PM on Wednesday, February 28, 2024 (UTC+8). This is expected to not affect job execution. In rare cases, if a job fails, retry it.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!

January 25, 2024 — MaxCompute compute cost optimization feature release

Dear MaxCompute user,

The MaxCompute compute resource optimization feature was upgraded to the compute cost optimization feature on Thursday, January 25, 2024 (UTC+8).

The compute cost optimization feature builds on the original compute resource optimization feature (generating optimal configurations for subscription top-level quotas). It adds recommendations for switching from pay-as-you-go compute resources to subscription compute resources, including projected outcomes. This helps you further optimize compute costs and improve resource utilization. For more information and usage instructions, see Compute Resource Configuration Optimization.

If you have questions or need help, contact us through your dedicated DingTalk group or the MaxCompute Developer Community (DingTalk group ID: 11782920).

Thank you for your cooperation!