MaxCompute will update its Service-Level Agreement (SLA) on September 1, 2023 (UTC+8). To help you better understand how service availability and compensation amounts are calculated under the new SLA, this topic provides case examples for different service unavailability scenarios.
Background information
Alibaba Cloud's cloud-native big data computing service, MaxCompute, will update its Service-Level Agreement (SLA) on September 1, 2023 (UTC+8). The updated SLA calculates service availability and compensation separately for each service type. It currently applies to the following services:
MaxCompute subscription-based compute service
MaxCompute pay-as-you-go Standard Edition compute service
MaxCompute pay-as-you-go off-peak compute service
MaxCompute subscription-based Data Transmission Service
MaxCompute free Data Transmission Service
For more information, see Cloud-Native Big Data Computing Service MaxCompute Service Level Agreement.
Case 1: Subscription-Based Compute Service Unavailability
Background
An Alibaba Cloud account (UID: 1315XXXX4628) purchased 600 CUs of reserved MaxCompute subscription-based compute resources in the China (Shanghai) region on August 5, 2023, for a two-month term with an 85% business discount, paying 153,000 CNY upfront.
After the purchase, an Alibaba Cloud order was created with instance ID: fe5df-XXXX-XXXX-5d8e. MaxCompute created a top-level subscription quota named default_subscription_Quota_p for this account.
To meet business needs, the customer configured 100 CUs of elastic reserved compute resources on this top-level quota using the MaxCompute console, active daily from 03:00:00-03:59:59.
The customer created two MaxCompute projects, ProjectA and ProjectB, and set the default compute quota for both projects to the sub-quota QuotaA under default_subscription_Quota_p. All daily compute jobs in ProjectA and ProjectB used the QuotaA quota.
Service Unavailability Incident
From 02:00:00 to 02:59:59 on September 3, 2023, ProjectA sent 10 valid compute service requests per minute at 02:03, 02:04, 02:05, 02:06, and every minute from 02:12 to 02:52. All requests returned the error code ODPS-0010000:System internal error, which caused downstream development node delays and business impact. The customer filed a compensation claim with Alibaba Cloud MaxCompute on October 2, 2023.
Compute Service Availability Calculation
A backend log review confirmed that all these requests were failed compute service requests, not duplicates, and that no other compute requests used this subscription quota during the period. These requests used the subscription-based QuotaA. Therefore, this incident qualifies as subscription-based compute service unavailability.
Per the SLA, availability is measured in 5-minute intervals. If the error rate exceeds 5% in any 5-minute interval, that interval is considered unavailable. During 02:00:00-02:59:59, failures at 02:03 and 02:04 fell into the 02:00:00-02:04:59 interval with a 100% error rate. Failures at 02:05 and 02:06 fell into the 02:05:00-02:09:59 interval with a 100% error rate. This pattern continued, resulting in 11 five-minute intervals with 100% error rates, which are counted as 11 unavailable time segments.
The MaxCompute subscription-based compute service availability is calculated per top-level subscription quota (Quota) as follows:
Service availability = ((Total minutes in service period − Number of unavailable segments × 5) / Total minutes in service period) × 100% = ((30 × 24 × 60 − 11 × 5) / (30 × 24 × 60)) × 100% = 99.87%.
Compensation Amount Calculation
The compensation policy for the MaxCompute subscription-based compute service is as follows:
Service availability | Voucher compensation amount |
99.9% > Service availability ≥ 99.0% | 15% of monthly service fee |
99.0% > Service availability ≥ 95.0% | 30% of monthly service fee |
95.0% > Service availability | 100% of monthly service fee |
With a service availability of 99.87%, the first compensation tier applies. The customer is entitled to 15% of the monthly service fee.
When the MaxCompute subscription-based compute service is unavailable, the monthly service fee is calculated as follows: Monthly service fee (subscription-based compute service) = Reserved compute resource fee + Elastic reserved compute resource fee + Project service fee. The monthly service fee is broken down as follows:
The reserved compute resource fee is the portion of the prepaid bill allocated to September (UID: 1315XXXX4628, instance ID: fe5df-XXXX-XXXX-5d8e, product detail: Big Data Computing Service MaxCompute (monthly)) =
153,000 CNY / 2 = 76,500 CNY.The elastic reserved compute resource fee is the billed amount (UID: 1315XXXX4628, instance ID: default_subscription_Quota_p;cn-shanghai, product detail: Big Data Computing Service MaxCompute (monthly)) =
1,593.75 CNY.The project service fee is the billed amount (UID: 1315XXXX4628, instance ID: ProjectA;cn-shanghai, product detail: Big Data Computing Service MaxCompute (monthly)) = 500 CNY.
The total monthly service fee (subscription-based compute service) is 78,593.75 CNY. The compensation amount is 11,789.0625 CNY (78,593.75 × 15%).
Case 2: Pay-as-You-Go Standard Edition Compute Service Unavailability
Background
An Alibaba Cloud account (UID: 1315XXXX4628) activated the pay-as-you-go Standard Edition service in the China (Hangzhou) region on August 5, 2023. MaxCompute created a sub pay-as-you-go Standard Edition quota named default_pay-as-you-go_Quota for this account.
The customer created the MaxCompute project ProjectC and set its default compute quota to default_pay-as-you-go_Quota. All daily compute jobs in ProjectC used the default_pay-as-you-go_Quota quota.
Service Unavailability Incident
From 02:00:00 to 02:59:59 on September 3, 2023, ProjectC sent 10 valid compute service requests per minute at 02:03, 02:04, 02:05, 02:06, and every minute from 02:12 to 02:52. All requests returned the error code: ODPS-0010000:System internal error, which caused downstream development node delays and business impact. The customer filed a compensation claim with Alibaba Cloud MaxCompute on October 2, 2023.
Compute Service Availability Calculation
A backend log review confirmed that all these requests were failed compute service requests, not duplicates, and used the default_pay-as-you-go_Quota. Therefore, this incident qualifies as pay-as-you-go Standard Edition compute service unavailability.
Per the SLA, availability is measured in 5-minute intervals. If the error rate exceeds 5% in any 5-minute interval, that interval is considered unavailable. During 02:00:00-02:59:59, failures at 02:03 and 02:04 fell into the 02:00:00-02:04:59 interval with a 100% error rate. Failures at 02:05 and 02:06 fell into the 02:05:00-02:09:59 interval with a 100% error rate. This pattern continued, resulting in 11 five-minute intervals with 100% error rates, which are counted as 11 unavailable time segments.
The MaxCompute pay-as-you-go Standard Edition compute service availability is calculated per project (Project) as follows:
Service availability = ((Total minutes in service period − Number of unavailable segments × 5) / Total minutes in service period) × 100% = ((30 × 24 × 60 − 11 × 5) / (30 × 24 × 60)) × 100% = 99.87%.
Compensation Amount Calculation
The compensation policy for the MaxCompute pay-as-you-go Standard Edition compute service is as follows:
Service availability | Voucher compensation amount |
99.9% > Service availability ≥ 99.0% | 15% of monthly service fee |
99.0% > Service availability ≥ 95.0% | 30% of monthly service fee |
95.0% > Service availability | 100% of monthly service fee |
With a service availability of 99.87%, the first compensation tier applies. The customer is entitled to 15% of the monthly service fee.
When the MaxCompute pay-as-you-go Standard Edition compute service is unavailable, the monthly service fee is the total amount that the customer paid in a calendar month for the affected MaxCompute project (UID: 1315XXXX4628, instance ID: ProjectC;cn-hangzhou) under the product detail Big Data Computing Service MaxCompute (pay-as-you-go). Assuming the customer's billed amount was 8,965.23 CNY, the compensation amount is 1,344.7845 CNY (8965.23 × 15%).
Case 3: Pay-as-You-Go Off-Peak Compute Service Unavailability
Background
An Alibaba Cloud account (UID: 1315XXXX4628) activated the pay-as-you-go off-peak service in the China (Hangzhou) region on August 5, 2023. MaxCompute created a sub pay-as-you-go off-peak quota named os_SpotQuota for this account.
The customer created the MaxCompute project ProjectD and set its default compute quota to os_SpotQuota. All daily compute jobs in ProjectD used the os_SpotQuota quota.
Service Unavailability Incident
From 02:00:00 to 03:59:59 on September 3, 2023, ProjectD sent 10 valid compute service requests per minute during 02:12:00-02:52:59 and 03:08:00-03:48:59. All requests returned the error code: ODPS-0010000:System internal error, which caused downstream development node delays and business impact. The customer filed a compensation claim with Alibaba Cloud MaxCompute on October 2, 2023.
Compute Service Availability Calculation
A backend log review confirmed that all these requests were failed compute service requests, not duplicates, and used the os_SpotQuota. Therefore, this incident qualifies as pay-as-you-go off-peak compute service unavailability.
Per the SLA, availability is measured in 5-minute intervals. If the error rate exceeds 5% in any 5-minute interval, that interval is considered unavailable. During 02:00:00-03:59:59, failures at 02:12, 02:13, and 02:14 fell into the 02:10:00-02:14:59 interval with a 100% error rate. This pattern continued, resulting in 18 five-minute intervals with 100% error rates, which are counted as 18 unavailable time segments.
The MaxCompute pay-as-you-go off-peak compute service availability is calculated per project (Project) as follows:
Service availability = ((Total minutes in service period − Number of unavailable segments × 5) / Total minutes in service period) × 100% = ((30 × 24 × 60 − 18 × 5) / (30 × 24 × 60)) × 100% = 99.79%.
Compensation Amount Calculation
The compensation policy for the MaxCompute pay-as-you-go off-peak compute service is as follows:
Service availability | Voucher compensation amount |
99.0% > Service availability ≥ 95.0% | 15% of monthly service fee |
95.0% > Service availability ≥ 90.0% | 30% of monthly service fee |
90.0% > Service availability | 100% of monthly service fee |
The customer's service availability is 99.79%, which exceeds the promised availability (99.0%) for the pay-as-you-go off-peak compute service. Therefore, no compensation is required.
If service availability falls below 99.0%, the monthly service fee is the total amount that the customer paid in a calendar month for the affected MaxCompute project under the product detail Big Data Computing Service MaxCompute (pay-as-you-go off-peak).
Case 4: Subscription-Based Data Transmission Service Unavailability
Background
An Alibaba Cloud account (UID: 1315XXXX4628) purchased a 600-Slot dedicated resource group for the MaxCompute Data Transmission Service (subscription-based) in the China (Shanghai) region on August 5, 2023, for a two-month term with an 85% business discount, paying 153,000 CNY upfront.
After the purchase, an Alibaba Cloud order was created with instance ID: he5df-XXXX-XXXX-5d8f. MaxCompute created a dedicated resource group named Tunnel_QuotaA for this account.
The customer created two MaxCompute projects, ProjectA and ProjectB, and ran data transmission tasks from ProjectA.
Service Unavailability Incident
From 02:00:00 to 02:59:59 on September 3, 2023, ProjectA sent 10 valid data transmission service requests per minute at 02:03, 02:04, 02:05, 02:06, and every minute from 02:12 to 02:52. All requests returned HTTP status 500, which caused business impact. The customer filed a compensation claim with Alibaba Cloud MaxCompute on October 2, 2023.
Calculating service availability
A backend log review confirmed that all these requests were failed data transmission service requests, not duplicates, and that no other data transmission requests occurred in the region during the period. Because the customer had purchased a subscription-based dedicated resource group in the affected region, this incident qualifies as subscription-based Data Transmission Service unavailability.
Per the SLA, availability is measured in 5-minute intervals. If the error rate exceeds 5% in any 5-minute interval, that interval is considered unavailable. During 02:00:00-02:59:59, failures at 02:03 and 02:04 fell into the 02:00:00-02:04:59 interval with a 100% error rate. Failures at 02:05 and 02:06 fell into the 02:05:00-02:09:59 interval with a 100% error rate. This pattern continued, resulting in 11 five-minute intervals with 100% error rates, which are counted as 11 unavailable time segments.
The MaxCompute subscription-based Data Transmission Service availability is calculated per dedicated resource group as follows:
Service availability = ((Total minutes in service period − Number of unavailable segments × 5) / Total minutes in service period) × 100% = ((30 × 24 × 60 − 11 × 5) / (30 × 24 × 60)) × 100% = 99.87%.
Compensation Amount Calculation
The compensation policy for the MaxCompute subscription-based Data Transmission Service is as follows:
Service availability | Voucher compensation amount |
99.9% > Service availability ≥ 99.0% | 15% of monthly service fee |
99.0% > Service availability ≥ 95.0% | 30% of monthly service fee |
95.0% > Service availability | 100% of monthly service fee |
The customer's service availability is 99.87%, which falls into the first compensation tier. The customer is entitled to 15% of the monthly service fee.
In this case, the customer's monthly service fee is the portion of the prepaid Data Transmission Service (subscription-based) bill allocated to September (UID: 1315XXXX4628, instance ID: he5df-XXXX-XXXX-5d8f, product detail: Big Data Computing Service MaxCompute (monthly)) = 153,000 CNY / 2 = 76,500 CNY.
The compensation amount is 11,475 CNY (76,500 × 15%).
Case 5: Free Data Transmission Service Unavailability
Background
An Alibaba Cloud account (UID: 1315XXXX4628) activated the MaxCompute service in the China (Hangzhou) region on August 5, 2023. The account holder created the MaxCompute project ProjectC and ran data transmission tasks using the shared resource group for the Data Transmission Service. The customer's business discount is 0.85.
Service Unavailability Incident
From 02:00:00-09:59:59 on September 3, 2023, ProjectC sent 10 valid data transmission service requests every 5 minutes. All requests returned HTTP status 500, which caused business impact. The customer filed a compensation claim with Alibaba Cloud MaxCompute on October 2, 2023.
Compute service availability
A backend log review confirmed that all these requests were failed data transmission service requests, not duplicates, and that no other data transmission requests occurred outside this window. Because the customer did not purchase a subscription-based dedicated resource group in the affected region, this incident qualifies as free Data Transmission Service unavailability.
Per the SLA, availability is measured in 5-minute intervals. If the error rate exceeds 5% in any 5-minute interval, that interval is considered unavailable. During 02:00:00-09:59:59, there were 96 unavailable time segments.
The MaxCompute free Data Transmission Service availability is calculated per project (Project) as follows:
Service availability = ((Total minutes in service period − Number of unavailable segments × 5) / Total minutes in service period) × 100% = ((30 × 24 × 60 − 96 × 5) / (30 × 24 × 60)) × 100% = 98.89%.
Compensation Amount Calculation
The compensation policy for the MaxCompute free Data Transmission Service is as follows:
Service availability | Voucher compensation amount |
99.9% > Service availability ≥ 99.0% | 15% of monthly service fee |
99.0% > Service availability ≥ 95.0% | 30% of monthly service fee |
95.0% > Service availability | 100% of monthly service fee |
With a service availability of 98.89%, the first compensation tier applies. The customer is entitled to 15% of the monthly service fee.
When the MaxCompute free Data Transmission Service is unavailable, the monthly service fee is the lower of the converted monthly service fee for free Data Transmission Service or the monthly service fee of the affected project.
The converted monthly service fee for free Data Transmission Service is calculated as
Default maximum concurrency in the affected region × Public cloud unit price of dedicated Data Transmission Service resource group (150 CNY/month) × Customer business discount = 300 × 150 × 0.85 = 38,250 CNY. For information about the default maximum concurrency per region, see Data Transmission Service Shared Resource Group Description.The monthly service fee of the affected project is the total amount that the customer paid in a calendar month for the affected MaxCompute project (UID: 1315XXXX4628, instance ID: ProjectC;cn-hangzhou). Assuming the customer's billed amount was
8,965.23 CNY.
The lower value is 8,965.23 CNY. The compensation amount is 1,344.7845 CNY (8,965.23 CNY × 15%).
If other service types also experience unavailability in the same calendar month, for example, the pay-as-you-go Standard Edition service in the preceding case, no separate compensation will be issued for the free Data Transmission Service.