Microsoft Corporation
Real time multi-tenant workload tracking and auto throttling

Abstract:

Technologies are disclosed for real-time workload tracking and throttling within a multi-tenant service. Multi-tenant services receive requests from computing devices associated with different tenants. While processing requests, the multi-tenant service itself sends requests to an underlying resource, such as a database. Requests from computing device associated with an overactive tenant may cause the multi-tenant service to overwhelm the underlying resource. The overwhelmed underlying resource may not know which tenant a request received by the underlying resource is associated with, and so the underlying resource is unable to only throttle requests originating from computing devices associated with the overactive tenant. Instead, the underlying resource throttles all requests from the multi-tenant service. To avoid this result, the multi-tenant service tracks utilization of the underling resource associated with each tenant, and throttles requests received from overactive tenants before the underlying resource becomes overwhelmed and throttles all requests from the multi-tenant service.