Backend resource costs for online service offerings

• US 10,411,969 B2
• Filed: 10/03/2016
• Issued: 09/10/2019
• Est. Priority Date: 10/03/2016
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Srivastava, Shantanu, Sampedro, Andres Cuitlahuac Paz, Kothari, Atin Atul, Briggs, Andre Fayeofori
• Primary Examiner(s): Katsikis, Kostas J
• Application Number:
  US15/284,299
• Publication Number:
  US 20180097705A1
• Time in Patent Office: 1,072 Days
• CPC Class Codes: H04L 41/142 using statistical or mathematical methods
• CPC Class Codes: H04L 41/147 for predicting network behaviour
• CPC Class Codes: H04L 41/5019 Ensuring fulfilment of SLA
• CPC Class Codes: H04L 43/08 Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
• CPC Class Codes: H04L 43/0852 Delays
• CPC Class Codes: H04L 43/0882 Utilisation of link capacity
• CPC Class Codes: H04L 43/55 Testing of service level quality, e.g. simulating service usage
• 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

Representative embodiments disclose mechanisms to calculate resource costs (e.g., processing, memory, storage and/or monetary costs) to provide an online service. The service is offered by a service provider that provides a plurality of different online services. To identify systems involved in providing a particular service, each request is associated with a unique identifier. The identifier is passed between systems and logged. Thus a trace of the unique identifier yields all systems involved in providing the service. A maximum operational load, calculated from operational data, is used to calculate a cost per request. A statistical sampling of the logged data is retrieved and the cost for each request in the sample is calculated based on a sum of the costs of the systems used to service the request. The cost of each system is the cost per request times the number of times the system is used to fill the request.

Claims

1. A method for calculating resource costs of online service offerings, comprising:

extracting, from a logging database, a statistical sampling comprising a plurality of records, each record representing a request for a service of a plurality of online services, each record comprising a request identifier and a plurality of systems utilized to fulfill the request, each identifier assigned so that each system involved in filling the request uses the same request identifier for the request;

for each record in the statistical sampling, performing operations comprising;

identifying a list of systems comprising the plurality of systems utilized to fulfill the request based on the associated identifier; and

calculating a backend resource cost for the record based on the backend resource cost for each system in the list of systems;

calculating at least one statistical metric using the backend resource cost for at least a subset of the records in the statistical sampling;

predicting a maximum operational load for the service based on data collected during normal operation of the service; and

calculating a backend resource cost per request by dividing a cost by the maximum operational load.

2. The method of claim 1, further comprising categorizing each request by at least one dimension, the at least one dimension comprising one or more of:

the service;

a region representing a source of the request;

a traffic type; and

a language; and

calculating the backend resource cost along at least one selected dimension.

3. The method of claim 1, wherein calculating the backend resource cost for the record comprises performing operations comprising:

for each system in the list of systems;

retrieving the backend resource cost per request;

identifying a number of times the system is used in filling the request;

calculating a system cost for the record by multiplying the backend resource cost per request and the number of times the system is used in filling the request; and

accumulating the system cost.

4. The method of claim 1, wherein the at least one statistical metric comprises at least one of:

a mean backend resource cost;

a medial backend resource cost;

a standard deviation of the backend resource cost; and

a probability density function of the backend resource cost.

5. The method of claim 1, wherein predicting a maximum operational load for the service comprises:

collecting data describing a load and at least one of either a latency or a utilization during normal operation of the service;

using regression to fit a curve to the collected data, the curve representing at least one of either the latency or the utilization for the load;

using the fitted curve, predicting a load value at which at least one of either the latency of the utilization will exceed a service level agreement; and

calculating the maximum operational load as being equal to or less than the load value.

6. The method of claim 5, wherein the maximum operational load is calculated by subtracting a business contingency load from the load value.

7. The method of claim 1, wherein the backend resource cost comprises at least one of:

a monetary cost;

a CPU utilization cost;

a memory utilization cost;

an I/O utilization cost;

a bandwidth utilization cost; and

a storage utilization cost.

8. A computing system comprising:

a processor and executable instructions accessible on a machine-readable medium that, when executed, cause the system to perform operations comprising;

extracting a statistical sampling comprising a plurality of records from a logging database, each record comprising a request identifier associated with a request and a list of systems utilized in filling the request, each identifier assigned so that each system involved in filling the request uses the same request identifier for the request;

for each record in the statistical sampling, performing operations comprising;

identifying the list of systems utilized to fulfill the request associated with the request identifier in the record; and

calculating a backend resource cost for the record based on the backend resource cost for each system in the list of systems; and

calculating at least one statistical metric using the backend resource cost for at least a subset of the records in the statistical sampling;

predicting a maximum operational load based on data collected during normal operation of the service; and

calculating a backend resource cost per request by dividing a cost by the maximum operational load.

9. The system of claim 8, wherein the operations further comprise categorizing each request by at least one dimension, the at least one dimension comprising one or more of:

a service;

a region representing a source of the request;

a traffic type; and

a language; and

calculating the backend resource cost along at least one selected dimension.

10. The system of claim 8, wherein the backend resource cost comprises at least one of:

a monetary cost;

a CPU utilization cost;

a memory utilization cost;

an I/O utilization cost;

a bandwidth utilization cost; and

a storage utilization cost.

11. The system of claim 8, wherein calculating the backend resource cost for the record comprises performing operations comprising:

for each system in the list of systems;

retrieving the backend resource cost per request;

identifying a number of times the system is used in filling the request;

calculating a system cost for the record by multiplying the backend resource cost per request and the number of times the system is used in filling the request; and

accumulating the system cost.

12. The system of claim 8, wherein the at least one statistical metric comprises at least one of:

a mean backend resource cost;

a medial backend resource cost;

a standard deviation of the backend resource cost; and

a probability density function of the backend resource cost.

13. The system of claim 8, wherein

predicting a maximum operational load for comprises:

collecting data describing a load and at least one of either a latency or a utilization during normal operation of the service;

using regression to fit a curve to the collected data, the curve representing at least one of either the latency or the utilization for the load;

using the fitted curve, predicting a load value at which at least one of either the latency of the utilization will exceed a service level agreement; and

calculating the maximum operational load as being equal to or less than the load value.

14. The system of claim 13, wherein the maximum operational load is calculated by subtracting a business contingency load from the load value.

15. A machine-readable medium having executable instructions encoded thereon, which, when executed by at least one processor of a machine, cause the machine to perform operations comprising:

collecting data describing a load and at least one of either a latency or a utilization during normal operation of an online service;

using regression to fit a curve to the collected data, the curve representing at least one of either the latency or the utilization for the load;

using the fitted curve, predicting a load value at which at least one of either the latency of the utilization will exceed a service level agreement;

calculating a maximum operational load as being equal to or less than the load value;

calculating a cost per request by dividing a cost by the maximum operational load;

extracting, from a logging database, a statistical sampling comprising a plurality of records, each record representing a request for a service of a plurality of online services, each record comprising a request identifier and a plurality of systems utilized to fulfill the request;

selecting at least one dimension to calculate a backend resource cost across;

for each record in the statistical sampling, performing operations comprising;

tracing the request identifier and identifying a list of systems utilized in filling the request associated with the request identifier; and

calculating the backend resource cost for the record based on the backed resource cost for each system in the list of systems;

accumulating the backend resource cost along the selected at least one dimension; and

calculating at least one statistical metric using the backend resource cost for at least a subset of the records in the statistical sampling.

16. The machine-readable medium of claim 15, wherein calculating the backend resource cost for the record comprises performing operations comprising:

for each system in the list of systems;

retrieving the backend resource cost per request;

identifying a number of times the system is used in filling the request;

calculating a system cost for the record by multiplying the backend resource cost per request and the number of times the system is used in filling the request; and

accumulating the system cost.

17. The machine-readable medium of claim 15, wherein the backend resource cost comprises at least one of:

a monetary cost;

a CPU utilization cost;

a memory utilization cost;

an I/O utilization cost;

a bandwidth utilization cost; and

a storage utilization cost.

18. The machine-readable medium of claim 15, wherein the selected at least one dimension comprises one or more of:

the service;

a region representing a source of the request;

a traffic type;

a user type; and

a language.