All Products
Search

This Product

    MaxCompute:Specifications overview

    Document Center

    MaxCompute:Specifications overview

    Last Updated:Mar 26, 2026

    MaxCompute offers five resource package types that differ in how compute, storage, and data transfer resources are provisioned and billed. Review the differences below to identify which package fits your workload before purchasing.

    Compare resource packages

    The following table summarizes resource availability and billing across all package types.

    Specification type Computing resource Storage resource Upload/download resource Billing
    Subscription

    • Reserved (dedicated pool)

    		 Shared pool, on-demand Shared pool, on-demand Compute pricing (Subscription) + Storage pricing + Data transfer pricing (Internet download)
    Elastically reserved CUs Reserved in dedicated pool; requires Subscription first — — Compute pricing (pay-by-the-hour)
    Auto-scaling CUs Elastic; requires Subscription first — — Compute pricing (pay-by-the-hour)
    Pay-as-you-go Shared pool, preemptible Shared pool, on-demand Shared pool, preemptible
    Pay-as-you-go Off-Peak Edition Shares pool with Pay-as-you-go; isolated from Subscription — —
    The subscription-based storage-intensive 160, storage-intensive 320, and storage-intensive 600 packages, along with subscription-based non-reserved computing resources and the pay-as-you-go developer edition, are no longer available for new purchases. These services were discontinued at 00:00:00 on October 31 (Tuesday), 2023 (UTC+8).

    Choose a resource package

    Subscription

    Choose Subscription for production workloads with stable, predictable compute demand. Reserved CUs (compute units) guarantee dedicated resources with no contention from other tenants.

    Storage and data transfer draw from shared pools and are billed separately.

    Elastically reserved CUs

    Choose Elastically reserved CUs when you already have a Subscription package and want to manually scale compute out during peak hours and back in during off-peak hours. This package adds reserved capacity on top of your existing Subscription resources.

    Elastically reserved CUs require an active Subscription resource package.

    Auto-scaling CUs

    Choose Auto-scaling CUs when your workload has frequent, unpredictable spikes and you want the system to handle scaling automatically. MaxCompute detects workload fluctuations and scales compute up or down without manual intervention — you pay only for the CUs consumed.

    Use this package to balance job performance against cost efficiency. For guidance on configuring Auto-scaling CUs, see Best practices for Auto-scaling CUs.

    Auto-scaling CUs require an active Subscription resource package.

    Pay-as-you-go

    Choose Pay-as-you-go for projects with unstable workloads or variable storage requirements. Compute, storage, and data transfer all draw from shared pools — no reserved resources and no upfront commitment required.

    Pay-as-you-go Off-Peak Edition

    Choose Pay-as-you-go Off-Peak Edition for latency-insensitive batch jobs that require the lowest possible cost. Typical workloads include processing large volumes of low-value data such as user behavior logs or system logs.

    This edition shares the compute pool with Pay-as-you-go and is isolated from Subscription resources. Jobs submitted here do not contend with Subscription workloads for capacity.

    How Subscription, Elastically reserved CUs, and Auto-scaling CUs work together

    Elastically reserved CUs and Auto-scaling CUs both extend a Subscription package — neither can be used independently. A typical production setup layers all three:

    • Subscription provides the baseline reserved compute capacity for production jobs.

    • Elastically reserved CUs add capacity when you anticipate peak load — you control when to scale out and in.

    • Auto-scaling CUs handle unexpected spikes automatically, scaling up when the system detects workload fluctuations and scaling down when load subsides.

    Have experienced technical staff plan the required storage and computing resources for your project based on your business needs, then choose the appropriate resource package by layering Elastically reserved CUs or Auto-scaling CUs based on your peak load patterns.

    What's next