All Documents
Current Document
Content is empty
If you don't find the content you expect, please try another search term
Content is empty
If you don't find the content you expect, please try another search term
Legal
Service Level Agreement（SLA）
Kingsoft Cloud GPU Elastic Physical Compute (GPU) Service Level Agreement
Last updated：2020-12-01 17:46:29
This Kingsoft Cloud International Website GPU Elastic Physical Compute (GPU) Service Level Agreement (“SLA”) applies to your purchase and use of the Kingsoft Cloud International Website GPU (“Service”) and your use of the Service is subjected to the terms and conditions of the Kingsoft Cloud International Website Agreement of Kingsoft Cloud Service between the relevant Kingsoft Cloud entity described in the Agreement of Kingsoft Cloud Service( “us”, or “we”) and you. This SLA only applies to your purchase and use of the Services for a fee, and shall not apply to any free Services or trial Services provided by us.
1.Definition
The GPU models to which this SLA applies include: Standard Compute (P3E);
Performance Compute (P3S); Standard Compute (P4E); Performance Compute (P4S); and Reasoning Compute (P3IE). The above models include sub-types after ".", for example, P4E.56F8 is a standard compute type (P4E).
Service Cycle: One Service Cycle equals to one calendar month. If the
service is used for less than one calendar month, the Service Cycle will be calculated as one calendar month.
Single Instance Service Cycle in Minutes: Calculated on the basis of the total number of days in a Service Cycle × 24 (hours) × 60 (minutes).
Single Region Service Cycle in Minutes: The total Instance Service Cycle in Minutes in all availability zones of a single region.
Instance Unavailability: The GPU instance is deemed unavailable if the disconnection between a GPU instance configured with access permitted rules and any IP address over TCP or UDP in the inbound and outbound directions lasts for more than 5 minutes.
Single Instance Service Downtime in Minutes: The total number of minutes in one Service Cycle during which time a single GPU instance is unavailable.
Single Region Multi-Zone Service Unavailability: Refers to the situation that, when GPU instances are deployed in at least two availability zones of a single region (hereinafter referred to as “single region multi-zone”), all GPU instances in an availability zone experience Instance Unavailability and meanwhile other GPU instances in any other availability zones of this region also experience Instance Unavailability (hereinafter referred to as “unavailable GPU instances in other zones of the same region”).
Single Region Multi-Zone Service Downtime in Minutes: Refers to the total number of minutes for all unavailable GPU instances in other zones of the same region when Single Region Multi-Zone Service Unavailability occurs in that region in one Service Cycle.
2.Service Availability Guarantee
2.1 The service availability of GPU shall be counted according to the service cycle based on two dimensions. The specific availability guarantee and calculation formula are as follows:
(1)Single GPU instance: The service availability of a single GPU instance in one Service Cycle should be no lower than 99.95%.
Service availability calculation formula: (Instance Service Cycle in Minutes – Instance Service Downtime in Minutes) / Instance Service Cycle in Minutes × 100%.
(2) Single Region Multi-Zone Service: The service availability of multi-zone in a single region in one Service Cycle should be no lower than 99.99%.
(3)Service availability calculation formula: The service availability of
multi-zone in a single region in one Service Cycle should be no lower than 99.99%.
Service availability calculation formula: (Single Region Service Cycle in Minutes - Single Region Multi-Zone Service Downtime in Minutes) / Single Region Service Cycle in Minutes × 100%.
2.2 If the GPU service does not reach the above-mentioned availability guarantee, the customer may be compensated according to this agreement, and the compensation scope does not include the unavailable service time caused by:
(1) System maintenance made by Kingsoft Cloud with a prior notification to the customer, including cutovers, maintenance, upgrades and simulated failure drills;
(2) Failures of any network, equipment that do not belong to Kingsoft Cloud and configuration adjustments;
(3) Hacker attacks on applications or data information of the customer;
(4) Loss or leakage of data, passwords, codes, and so on due to improper maintenance or confidentiality by the customer;
(5) Customer's self-upgrade of the operating system, including failure to observe the product usage documentation or usage suggestions of Kingsoft Cloud;
(6) Customer's applications or installations;
(7) Negligence of the customer or operation authorized by the customer;
(8) Service interruption caused by the server setting or user program installation conducted by the telecom operator(s) for resolving the decrease in the speed of accessing to the backbone line of the telecom operator(s) for any casual blocking of such backbone line;
(9) Service interruption caused by measures taken according to the regulatory requirements of the competent government department;
(10)Force majeure or accidents;
(11)Or other service unavailability not caused by Kingsoft Cloud.
3.Compensation Plan
If the availability of GPU service does not reach the service level in this agreement, Kingsoft Cloud will compensate customers for the unavailable time per GPU instance in a service cycle on 5-fold basis of the monthly service fee.
Kingsoft Cloud compensates by granting balance or vouchers, the granted balance or vouchers are used only to purchase GPU products. The compensation will not exceed the monthly service fee (excluding the deductible part by granting balance or vouchers) paid by the customer for this GPU instance in the current month.
Customers may apply for compensation for an instance that failed to achieve availability last month after the fifth business day of each month. The application for compensation should be raised in two (2) months after the month in which the GPU instance unavailability happened. No application for compensation will be accepted if the application exceeds the time limit.
4.Miscellaneous
Kingsoft Cloud reserves its rights to modify the clauses of this SLA. Kingsoft Cloud will notify the customer 30 days in advance by website publicity, SMS or email if there are any changes to SLA clauses. If customers do not agree with the modifications that Kingsoft Cloud made to SLA, they have rights to abandon the use of GPU service. Your continued use of the service shall be deemed as your acceptance of the amended SLA.
Pure Mode