Automated capacity provisioning method using historical performance data
View Patent ↗The method may include collecting performance data relating to processing nodes of a computer system which provide services via one or more applications, analyzing the performance data to generate an operational profile characterizing resource usage of the processing nodes, receiving a set of attributes characterizing expected performance goals in which the services are expected to be provided, and generating at least one provisioning policy based on an analysis of the operational profile in conjunction with the set of attributes. The at least one provisioning policy may specify a condition for re-allocating resources associated with at least one processing node in a manner that satisfies the performance goals of the set of attributes. The method may further include re-allocating, during runtime, the resources associated with the at least one processing node when the condition of the at least one provisioning policy is determined as satisfied.
1. A system for automated capacity provisioning, the system comprising:
an assessment module configured to cause at least one processor to analyze performance data of a computer system to implement an application with fluctuating demand, and generate a system resource usage profile based on the performance data, the system resource usage profile characterizing resource usage of computer resources of the computer system over time;
a policy generating module configured to cause the at least one processor to automatically generate one or more provisioning policies based on an analysis of the system resource usage profile and one or more service level agreements (SLAs), the policy generating module configured to cause the at least one processor to calculate a number of servers required for each of a plurality of time slots, to merge adjacent time slots having same calculated number of servers, and to calculate transactional weights to control arrival rate of transactions to the servers for one or more of the time slots; and
a load balancer configured to cause the at least one processor to scale the computer resources based on the one or more provisioning polices including adjusting an allocation of servers to the number of servers required for a particular time slot of the plurality of time slots,
wherein the load balancer is configured to cause the at least one processor to distribute arriving transactions to the servers using the calculated transactional weights.
2. The system of claim 1 , wherein the policy generating module includes a predictive module configured to cause the at least one processor to apply historical trending and predictive analysis to predict the number of servers at a given time.
3. The system of claim 1 , wherein the policy generating module includes a what-if module configured to cause the at least one processor to generate what-if scenarios to produce different combinations of workloads and resources to determine resulting performance in each of the combinations.
4. The system of claim 3 , wherein the what-if module is configured to cause the at least one processor to determine a smallest number of servers required to meet a corresponding SLA at a given workload level.
5. The system of claim 1 , further comprising:
a migrating module configured to cause the at least one processor to generate one or more provisioning policies to consolidate available servers between partitions of the computer system by estimating a number of homogeneous servers needed to support the application before migration.
6. The system of claim 5 , wherein the migrating module is configured to cause the at least one processor to calculate a maximum normalized unitization total for all available servers over a time period, and determine whether a server overhead is greater than a partition overhead, and if the server overhead is not greater than the partition overhead, the migrating module is configured to cause the at least one processor to calculate a number of required servers.
7. The system of claim 1 , wherein the policy generating module is configured to cause the at least one processor to receive an updated system resource usage profile, and automatically modify the one or more provisional policies based on a subsequent analysis of the updated system resource usage profile in conjunction with the one or more SLAs.
8. The system of claim 1 , wherein the policy generating module is configured to cause the at least one processor to provide a user interface configured to allow a user to set up dynamic rules for adjusting the scaling of the computer resources.
9. A non-transitory computer-readable medium storing instructions, when executed by at least one processor, are configured to perform the following operations:
analyze performance data of a computer system to implement an application with fluctuating demand;
generate a system resource usage profile based on the performance data, the system resource usage profile characterizing resource usage of computer resources of the computer system over time;
automatically generate one or more provisioning policies based on an analysis of the system resource usage profile and one or more service level agreements (SLAs) including calculate a number of servers required for each of a plurality of time slots, merge adjacent time slots having same calculated number of servers, and calculate transactional weights to control arrival rate of transactions to the servers for one or more of the time slots; and
scale the computer resources based on the one or more provisioning polices including adjust an allocation of servers to the number of servers required for a particular time slot of the plurality of time slots and distribute arriving transactions to the servers using the calculated transactional weights.
10. The non-transitory computer-readable medium of claim 9 , wherein the automatically generate one or more provisioning policies further includes generate what-if scenarios to produce different combinations of workloads and resources to determine resulting performance in each of the combinations.
11. The non-transitory computer-readable medium of claim 9 , wherein the automatically generate one or more provisioning policies further includes generate one or more provisioning policies to consolidate available servers between partitions of the computer system by estimating a number of homogeneous servers needed to support the application before migration.
12. The non-transitory computer-readable medium of claim 9 , further comprising:
receive an updated system resource usage profile; and
automatically modify the one or more provisional policies based on a subsequent analysis of the updated system resource usage profile in conjunction with the one or more SLAs.
13. A method for automatically allocating resources in a computer system, the method comprising:
analyzing performance data of a computer system to implement an application with fluctuating demand;
generating a system resource usage profile based on the performance data, the system resource usage profile characterizing resource usage of computer resources of the computer system over time;
automatically generating one or more provisioning policies based on an analysis of the system resource usage profile and one or more service level agreements (SLAs) including calculating a number of servers required for each of a plurality of time slots, merging adjacent time slots having same number of servers, and calculating transactional weights to control arrival rate of transactions to the servers for one or more of the time slots; and
scaling the computer resources based on the one or more provisioning polices including adjusting an allocation of servers to the number of servers required for a particular time slot of the plurality of time slots and distributing arriving transactions to the servers using the calculated transactional weights.
14. The method of claim 13 , further comprising:
applying historical trending and predictive analysis to predict the number of servers at a given time.
15. The method of claim 13 , further comprising:
generating what-if scenarios to produce different combinations of workloads and resources to determine resulting performance in each of the combinations; and
determining a smallest number of servers required to meet a corresponding SLA at a given workload level.