Quality of service aware captive aggregation with true datacenter testing
View Patent ↗Technologies are generally described for efficient datacenter management. In some examples, client jobs on a datacenter are subcontracted out to another datacenter so that the subcontracted jobs can be pulled back to the original datacenter when capacity becomes available again. A captive aggregator module combining network message, command management, data analysis on QoS effects of traffic relay, and strategic management decides whether it makes sense to keep tasks captively aggregated. A middleware system for testing true performance of an application or surrogate and optimizing among multiple datacenters based on true performance evaluates candidate datacenters prior to subcontracting jobs from the original datacenter. The middleware deploys the application to multiple candidate clouds to perform substantially similar tasks on resources that claim to be roughly equivalent in price/performance. The middleware receives data on actual cost and resource consumption, analyzes differences, and redistributes actual work tasks to take advantage of differences.
1. A method to manage Quality of Service (QoS) aware captive aggregation with datacenter testing through middleware, the method comprising:
receiving a job at a datacenter, wherein the received job exceeds a capacity of the datacenter;
determining a suitable subcontract datacenter for at least one of a plurality of existing jobs at the datacenter by communicating with the middleware, the middleware configured to:
deploy a reference task to a plurality of candidate datacenters to perform identical tasks on resources, wherein the reference task is a surrogate task with a plurality of calibration points, and the surrogate task is configured to deploy in a model of an actual task application with similar levels of communication between blocks and computation between actions; and
receive data on an actual cost and resource consumption for performing the reference task;
forwarding the at least one existing job to the suitable subcontract datacenter based on the received data; and
upon receiving results of the forwarded job, providing the results to an originating client, wherein a QoS for the forwarded job is maintained at a substantially same level as if the job was performed at the datacenter.
2. The method according to claim 1 , further comprising
enabling the originating client for the forwarded job to exchange data directly with the suitable subcontract datacenter.
3. The method according to claim 1 , further comprising
determining suitability of a subcontract datacenter based on a pricing structure; and
pulling back the forwarded job prior to its completion at the suitable subcontract datacenter based on a pricing structure associated with the forwarded job.
4. The method according to claim 1 , wherein determining suitability of the subcontract datacenter further includes:
analyzing differences between the actual cost and resource consumption for performing the reference task at the datacenter and at a candidate datacenter; and
forwarding the job based on analysis results.
5. The method according to claim 4 , wherein the candidate datacenters are selected based on having substantially similar price and performance characteristics.
6. The method according to claim 4 , further comprising
generating a plurality of comparison metrics based on the analysis results; and
optimizing a task assignment function and assigning tasks based on the comparison metrics, wherein each job includes one or more tasks.
7. The method according to claim 1 , further comprising
recording timestamps at the calibration points.
8. An apparatus to manage Quality of Service (QoS) aware captive aggregation with datacenter testing through middleware, the apparatus comprising:
a memory;
a processor coupled to the memory, wherein the processor is configured to execute a captive aggregation module in conjunction with a datacenter management application, the captive aggregation module configured to:
receive a job at a datacenter, wherein the received job exceeds a capacity of the datacenter;
determine a suitable subcontract datacenter for at least one of a plurality of existing jobs at the datacenter by communicating with the middleware, the middleware configured to:
deploy a reference task to a plurality of candidate datacenters to perform identical tasks on resources, wherein the reference task is a surrogate task with a plurality of calibration points, and the surrogate task is configured to deploy in a model of an actual task application with similar levels of communication between blocks and computation between actions; and
receive data on an actual cost and resource consumption for performing the reference task;
forward the at least one existing job to the suitable subcontract datacenter based on the received data; and
upon receiving results of the forwarded job, provide the results to an originating client, wherein a QoS for the forwarded job is maintained at a substantially same level as if the job was performed at the datacenter.
9. The apparatus according to claim 8 , wherein the middleware is further configured to:
analyze differences between the actual cost and resource consumption for performing the reference task at the datacenter and at a candidate datacenter; and
enable the captive aggregation module to forward the job based on analysis results.
10. The apparatus according to claim 8 , wherein the reference tasks are employed to calibrate relative demand and performance in a predefined condition of a plurality of inputs.
11. The apparatus according to claim 8 , wherein the middleware is further configured to
distribute the reference task to secondary datacenters.
12. The apparatus according to claim 8 , wherein the middleware is further configured to
dynamically modify an interval of monitoring the candidate datacenters.
13. The apparatus according to claim 8 , wherein the middleware is further configured to
select the reference task based on a usage of resource diversity.
14. A datacenter to manage Quality of Service (QoS) aware captive aggregation with datacenter testing through middleware, the datacenter comprising:
one or more servers configured to perform datacenter management tasks, wherein at least one server, as a part of the datacenter management tasks, is configured to execute a captive aggregation module, the captive aggregation module configured to:
receive a job at a datacenter, wherein the received job exceeds a capacity of the datacenter;
determine a suitable subcontract datacenter for at least one of a plurality of existing jobs at the datacenter by communicating with a middleware module, the middleware module configured to:
deploy a reference task to a plurality of candidate datacenters to perform identical tasks on resources, wherein the reference task is a surrogate task with a plurality of calibration points, and the surrogate task is configured to deploy in a model of an actual task application with similar levels of communication between blocks and computation between actions; and
receive data on an actual cost and resource consumption for performing the reference task;
forward the at least one existing job to the suitable subcontract datacenter based on the received data; and
upon receiving results of the forwarded job, provide the results to an originating client, wherein a QoS for the forwarded job is maintained at a substantially same level as if the job was performed at the datacenter.
15. The datacenter according to claim 14 , wherein the captive aggregation module is further configured to
determine suitability of a subcontract datacenter based on a pricing structure; and
pull back the forwarded job prior to its completion at the suitable subcontract datacenter based on a pricing structure associated with the forwarded job.
16. A datacenter to manage Quality of Service (QoS) aware captive aggregation with datacenter testing through middleware, the datacenter comprising:
one or more servers configured to perform datacenter management tasks, wherein at least one server, as a part of the datacenter management tasks, is configured to execute a middleware module, the middleware module configured to:
deploy a reference task to a plurality of candidate datacenters to perform identical tasks on resources, wherein the reference task is a surrogate task with a plurality of calibration points, and the surrogate task is configured to deploy in a model of an actual task application with similar levels of communication between blocks and computation between actions;
receive data on an actual cost and resource consumption for performing the reference task;
analyze differences between the actual cost and resource consumption for performing the reference task at a datacenter with excess jobs and at a candidate datacenter; and
enable a captive aggregation module of the datacenter to forward the job based on analysis results.
17. The datacenter according to claim 16 , wherein the candidate datacenters are selected based on having substantially similar price and performance characteristics.
18. The datacenter according to claim 16 , wherein the middleware module is further configured to
generate a plurality of comparison metrics based on the analysis results.
19. The datacenter according to claim 18 , wherein the middleware module is further configured to
optimize a task assignment function and enable assignment of tasks based on the comparison metrics, wherein each job includes one or more tasks.
20. The datacenter according to claim 16 , wherein the reference task is employed to calibrate relative demand and performance in a predefined condition of a plurality of inputs.
21. The datacenter according to claim 16 , wherein the middleware module is further configured to
distribute the reference task to secondary datacenters.