Multi-stage feedback controller for prioritizing tenants for multi-tenant applications

• US 9,584,588 B2
• Filed: 08/21/2013
• Issued: 02/28/2017
• Est. Priority Date: 08/21/2013
• Status: Active Grant

• Current Assignee: SAP SE
• Original Assignee: SAP SE
• Inventors: Mehta, Arpit, Krebs, Rouven
• Primary Examiner(s): Bates, Kevin
• Assistant Examiner(s): Kim, Dae
• Application Number:
  US13/972,477
• Publication Number:
  US 20150058484A1
• Time in Patent Office: 1,287 Days
• US Class Current: 1/1
• CPC Class Codes: G06F 2209/5013 Request control
• CPC Class Codes: G06F 2209/5021 Priority
• CPC Class Codes: G06F 9/5072 Grid computing
• CPC Class Codes: H04L 41/5009 Determining service level performance parameters or violations of service level contracts, e.g. violations of agreed response time or mean time between failures [MTBF]
• CPC Class Codes: H04L 41/5025 by proactively reacting to service quality change, e.g. by reconfiguration after service quality degradation or upgrade
• CPC Class Codes: H04L 67/10 in which an application is distributed across nodes in the network software deployment G06F8/60; multiprogramming arrangements G06F9/46

Abstract

Methods, systems, and computer-readable storage media for determining weights for selecting requests from a plurality of tenant queues in a multi-tenant system receiving measured response time and measured throughput for each tenant in a set of tenants being considered in a current period, for each tenant in the set of tenants, determining a weight based on respective measured response times and respective measured throughput, the weight being determined based on one of a previous weight, an initialized weight, a modified proportional and integral (PI) control, and a volunteer weight increase, providing a set of weights that includes weight for each tenant in the set of tenants, and transmitting the set of weights to an access mechanism, the access mechanism selecting tenant requests for processing by a shared resource.

Claims

1. A computer-implemented method, for determining weights for selecting requests from a plurality of tenant queues in a multi-tenant system, the method being executed using one or more processors and comprising:

receiving, by the one or more processors, measured response time and measured throughput for each tenant in a set of tenants being considered in a current period;

for each tenant in the set of tenants, determining a weight based on respective measured response times per current period relative to a total of average response time for all tenants in the set of tenants and respective measured throughput provided as a throughput processed for each tenant in the set of tenants per current period, the weight being determined based on;

a previous weight, in response to absence of a response time violation, and

in response to occurrence of a response time violation, one of a modified proportional and integral (PI) control and a volunteer weight increase, the modified PI control being based on a relative error in a throughput rate to enhance a precision of the weight, and the volunteer weight increase being associated to the multi-tenant system being underutilized;

providing, by the one or more processors, a set of weights, the set of weights comprising the weight determined for each tenant in the set of tenants;

selecting, by an access mechanism, at least one tenant request based on the set of weights; and

transmitting, by the access mechanism, the at least one tenant request to an application server for processing of the at least one tenant request.

2. The method of claim 1, wherein for each tenant in the set of tenants, determining a weight based on respective measured response times and respective measured throughput comprises determining that a response time violation is not present for any tenant in the set of tenants and, in response, determining the weight for each tenant as an initialized weight.

3. The method of claim 2, wherein determining that a response time violation is not present is based on respective response time errors, each response time error being determined based on the measured response time and an agreed average response time for a respective tenant.

4. The method of claim 1, wherein for each tenant in the set of tenants, determining a weight based on respective measured response times and respective measured throughput comprises determining that a response time violation is not present for a particular tenant in the set of tenants and, in response, determining setting the weight for the particular tenant equal to a weight of a previous period.

5. The method of claim 1, wherein for each tenant in the set of tenants, determining a weight based on respective measured response times and respective measured throughput comprises determining whether a quota violation is present for a particular tenant in the set of tenants.

6. The method of claim 5, further comprising determining that the quota violation is present for the particular tenant and, in response, determining the weight for the particular tenant based on the volunteer weight increase.

7. The method of claim 5, further comprising determining that the quota violation is not present for the particular tenant and, in response, determining the weight for the particular tenant based on the modified PI control.

8. The method of claim 5, wherein determining whether the quota violation is present for the particular tenant in the set of tenants is performed in response to determining that a response time violation is present for the particular tenant.

9. The method of claim 5, wherein determining whether the quota violation is present for a particular tenant in the set of tenants is based on a respective quota error, the respective quota error being determined based on the measured throughput and an agreed throughput for the particular tenant.

10. The method of claim 1, wherein the measured response time and measured throughput are provided for a previous period.

11. The method of claim 1, wherein tenants in the set of tenants include tenants having non-empty request queues.

12. The method of claim 1, further comprising normalizing weights in the set of weights, such that a sum of the weights in the plurality of weights is equal to 1.

13. The method of claim 1, wherein the volunteer weight increase is determined based on a measured response time and a measured throughput for a respective tenant, and additional time that can be given to disruptive tenants without violating agreements of abiding tenants.

14. The method of claim 1, wherein, for the current period, a first weight of a first tenant is provided as a previous weight of the first tenant, a second weight of a second tenant is provided as the initialized weight, and a third weight of a third tenant is provided based on a weight determined using the modified PI control.

15. A non-transitory computer-readable storage medium coupled to one or more processors and having instructions stored thereon which, when executed by the one or more processors, cause the one or more processors to perform operations for determining weights for selecting requests from a plurality of tenant queues in a multi-tenant system, the operations comprising:

receiving measured response time and measured throughput for each tenant in a set of tenants being considered in a current period;

for each tenant in the set of tenants, determining a weight based on respective measured response times per current period relative to a total of average response time for all tenants in the set of tenants and respective measured throughput provided as a throughput processed for each tenant in the set of tenants per current period, the weight being determined based on;

a previous weight, in response to absence of a response time violation, and

in response to occurrence of a response time violation, one of a modified proportional and integral (PI) control and a volunteer weight increase, the modified PI control being based on a relative error in a throughput rate to enhance a precision of the weight, and the volunteer weight increase being associated to the multi-tenant system being underutilized;

providing a set of weights, the set of weights comprising the weight determined for each tenant in the set of tenants;

selecting, by an access mechanism, at least one tenant request based on the set of weights; and

transmitting, by the access mechanism, the at least one tenant request to an application server for processing of the at least one tenant request.

16. A system, comprising:

a computing device; and

a non-transitory computer-readable storage device coupled to the computing device and having instructions stored thereon which, when executed by the computing device, cause the computing device to perform operations for determining weights for selecting requests from a plurality of tenant queues in a multi-tenant system, the operations comprising;

receiving measured response time and measured throughput for each tenant in a set of tenants being considered in a current period;

for each tenant in the set of tenants, determining a weight based on respective measured response times per current period relative to a total of average response time for all tenants in the set of tenants and respective measured throughput provided as a throughput processed for each tenant in the set of tenants per current period, the weight being determined based on;

a previous weight, in response to absence of a response time violation, and

in response to occurrence of a response time violation, one of a modified proportional and integral (PI) control and a volunteer weight increase, the modified PI control being based on a relative error in a throughput rate to enhance a precision of the weight, and the volunteer weight increase being associated to the multi-tenant system being underutilized;

providing a set of weights, the set of weights comprising the weight determined for each tenant in the set of tenants;

selecting, by an access mechanism, at least one tenant request based on the set of weights; and

transmitting, by the access mechanism, the at least one tenant request to an application server for processing of the at least one tenant request.