Method for optimizing utilization of workload-consumed resources for time-inflexible workloads

#The present invention relates to a method for optimizing the utilization of workload-consumed resource for time-inflexible workloads. More particularly, the present invention relates to a method for optimizing the utilization of workload-consumed resources for time-inflexible workloads by combining some of the time-inflexible workloads according to standard deviation to reduce waste of the resources.

#Nowadays, the advent of cloud-based computing architectures has created new requirements for rapid and scalable deployment of resources for web stores, media outlets, and other on-line sites or services. In general, a cloud-based architecture determines deployment of a set of resources, such as power, number of CPUs (Central Processing Units), size of storages, operating systems, bandwidth, . . . etc. Those resources can be combined or allied together for a specific workload or a group of workloads. The workload may be an email service, operation of an ERP (Enterprise Resource Planning) system, data archiving, a video streaming service, or other services mentioned above. An optimized utilization of resources for these workloads in a data center or a cloud storage system can reduce waste of resources without violating the requirement of workload performance in the SLA (Service Level Agreement). Furthermore, cost of power consumption, maintenance and new hardware supplement can be lessened.

#Optimized utilization of the resources is not an easy task. This is because all the workloads are changing with time. Dynamic deployment of the resources is not able to catch up with the speed of changes. More resources for the surprising peak requirements from the operating workloads may be a safe way but not an economic one. Meanwhile, some hardware devices, e.g. HDDs (Hard Drive Disks) or SSDs (Solid State Disks) are consumables. Periodical or from-time-to-time replacement of the worn hardware devices makes it more difficult to predict the real capacities of the resources. Therefore, optimized utilization of the resources for workloads applied is a big concern for operating a data center or a cloud storage system.

#FIG. 1 is a flow chart of a method according to the present invention.

#FIG. 2 shows a record of IOPS that a cloud E-mail service consumed during a week.

#FIG. 3 tabularizes processes for finding out a combination of workloads out of 9 workloads running on a cloud system.

#The present invention will now be described more specifically with reference to the following embodiments.

#The present invention is to provide methods to optimize utilization of workload-consumed resources for time-inflexible workloads running in a storage system or a cloud platform. Furthermore, the methods are able to combine the time-inflexible workloads according to some of their characteristics for a given resource in the storage system or the cloud platform. Therefore, the workload-consumed resource refers to any hardware or priority in processes the storage system or the cloud platform offers the workloads. The workload-consumed resource, for example, may be power, storage capacity, bandwidth, CPU (Central Processing Unit) cores, IOPS (Input/Output Per Second), throughput, latency consumed by the workload. In the embodiments below, workload-consumed resources used are not limited to as they are. Any suitable workload-consumed resource can be used instead in the embodiments as those in the art know it is a common application. The workload-consumed resource profile may be obtained by recording the amount of the workload-consumed resource consumed by a specific workload. It can also come from predicting a trend of the amount of the workload-consumed resource consumed by the specific workload based on historical records. Here, any suitable methods, algorithms, or modules that provide such predictions can be applied. The data of the trend are best obtained by a storage traffic modeling system provided by the same inventor in U.S. patent application Ser. No. 14/290,533. Co-reference of the same techniques can be acquired from the application. Of course, the workload-consumed resource profile may also be obtained by combining a portion of a predicted trend and a portion of historical records of the of amount of the workload-consumed resource consumed by the specific workload. It is not limited by the present invention.

#Please refer to FIG. 1. It is a flow chart of a method according to the present invention. The first step of the method is to provide a number of workload-consumed resource profiles and calculate a standard deviation for each workload-consumed resource profile (S01). According to the present invention, each workload-consumed resource profile is a record of amount of a workload-consumed resource consumed by a specific time-inflexible workload over a cycle. In order to illustrate what a so-called workload-consumed resource profile is and how the standard deviation is calculated, please see FIG. 2. The example shows a record of IOPS that a cloud E-mail service (workload) consumed during a week. The curve is the workload-consumed resource profile. From all the historical records, the curve may somehow look similar every week. A peak value happens at noon on Tuesdays and a valley occurs at midnight on Wednesdays. A trend shows that Sundays, Thursdays and Fridays almost have relative peaks for IOPS. Similar curves can be obtained every week but with different extreme values. Since E-mails should be transmitted and received in time, the workload-consumed resource of IOPS therefore should be fulfilled without any delay, and it fits the purpose of present invention perfectly. A comparison is a data archiving service which is not a time-inflexible workload because the service can be processed later in a time of low traffic flow. One week can be deemed as a cycle. The standard deviation can be obtained by below formula: