TWI516952B - Adaptive fuzzy rule controlling system for software defined storage system for controlling performance parameter - Google Patents

Description

Adaptive for controlling performance parameters of software-defined storage systems Fuzzy rule control system

The present invention relates to an adaptive fuzzy rule control system for software defined storage, and more particularly to an adaptive fuzzy rule control system for software defined storage to control specific performance parameters. This particular performance parameter is required by the service level agreement.

Cloud services have grown very popular in the last decade. Cloud services are based on cloud computing and provide related services or products without increasing the burden on the client. Cloud computing involves a large number of computer hosts that are connected to each other via a communication network, such as the Internet. It relies on the sharing of resources to achieve consistency and economic size. The concept of cloud computing is an infrastructure formed by a combination of network infrastructure and resource sharing services. Among all the shared services, memory and storage devices are definitely the two most demanding projects. This is because some popular applications, such as video streaming, require huge amounts of data storage. When the cloud service works, the memory and Storage device management is very important to ensure that customers maintain normal service quality.

For example, servers used to provide cloud services typically manage or connect to several hard disks. The client uses the server and the data is read from or written to the hard disk. The delay in response time due to the limitations of the hard disk system can cause problems in service requirements. Under the operation of a normal hard disk system, the delay time is usually generated when the access speed required by the application surface (ie, the workload) exceeds that provided by the hard disk system. Thus, the maximum amount of load that a hard disk system can provide is often the bottleneck in the operation of the entire cloud service system. In other words, the number of input/output operations per second of the hard disk system cannot meet the external requirements. In response to this problem, it is necessary to remove or reduce the workload to achieve and improve the performance of the server. In practice, part of the workload can be shared by other servers (if any) or hard disks, and these servers or hard disks will be automatically or manually added to support existing hard disks. Regardless of which of the above methods is used to solve the problem, the added cost is a large number of hard disks that are reserved in advance due to unpredictable work conditions, and the necessary power consumption for additional hardware devices. From an economic perspective, it is really not worth doing. However, since the system's service level protocol usually specifies the shortest delay time or the minimum number of input/output operations per second, it must be achieved. For operators that maintain cloud services with limited funds, how to reduce costs is an important issue.

It is worth noting that the workload of the server (hard disk system) can more or less predict the evolution in the future based on historical records, and the development trend of the demand for cloud service workload is predictable. Therefore, the workload can be achieved with minimal effort by reconfiguring the hard disk in the hard disk system. However, a machine cannot learn how and when to reconfigure a hard disk. In many cases, this work is done by authorized personnel, depending on the status of time or on a fixed schedule, and the effect may not be very good.

Another requirement for growing as fast as cloud services is software definition storage. Software definition storage refers to computer data storage technology that can separate storage hardware from the software that manages the storage infrastructure. Under Software Definition Storage, you can launch some functional options such as deduplication, replication, thin provisioning, snapshots, and backups to provide policy management. With software-defined storage technology, there are several previous cases that provide solutions to the above problems. A method for reconfiguring a storage system is disclosed, for example, in U.S. Patent No. 20,130,297,907. The method comprises two main steps: receiving user demand information of the storage device, and automatically generating a function setting of the storage device and a device setting file for the storage device by using the user's demand information; and setting the function using the function device; To automatically reconfigure the storage device as one or more logical devices with independent behavioral characteristics. The content of this publication points to a new way to reconfigure storage devices by software definition storage concepts. The method and system according to the disclosure also allows the user to dynamically adjust one or more The configuration of the logic devices to more flexibly meet the user's needs information. However, this publication does not disclose how these functional settings are generated. At the same time, those function settings cannot be changed according to changes in different applications (ie, workload).

Accordingly, the present invention discloses a new system for implementing a software defined storage configuration to address the above problems. It utilizes an adaptive fuzzy rule control and operates without human intervention. By applying the present invention, the configuration of a storage device that satisfies any workload can be dynamically calculated that the reconfiguration of the storage device can be completed at a particular point in time in the future.

As mentioned previously, the prior art does not provide a means to dynamically reconfigure the storage device in response to changes in different applications (ie, workload), so a new method and system is needed to meet this need.

According to an aspect of the present invention, an adaptive fuzzy rule control system for a software defined storage system for controlling performance parameters in a storage node includes: a flow monitoring module for obtaining an observation value of performance parameters in the storage node An adaptive neuro-fuzzy inference module for learning a dynamic relationship between a plurality of storage device configurations and performance parameters over a period of time in the storage node, and outputting a fuzzy rule according to the dynamic relationship Establishing; a traffic prediction module for providing a predicted value of the performance parameter at a specific time point in the future; and a fuzzy rule control module for using the storage node at the specific time point in the future According to fuzzy rules and the prediction The value arranges the configuration of the storage device so that a particular value of a particular performance parameter can be reached at that particular point in time in the future. The storage node is operated by a software defined storage software.

According to the concept of the present case, the performance parameter includes the number of input and output operations per second, the delay time, or the throughput. The adaptive neuro-fuzzy inference module generates a plurality of membership functions, each of which is used to define a degree of performance parameters or a configuration of the storage device at a particular level. The fuzzy rule joins the membership function of the performance parameter for a particular level to the membership function of the configuration of the storage device for another particular level. The degree to which a fuzzy inference method is defined by the membership function is used to obtain the configuration of the storage device with at least a given performance parameter. The fuzzy inference method is the Mamdani inference method or the Sugeno inference method. The adaptive neuro-fuzzy inference module further checks a difference between a particular value of the particular performance parameter and an observed value of the particular performance parameter.

According to the present invention, if the difference value exceeds a tolerance value, the adaptive neuro-fuzzy inference module learns a new fuzzy rule and membership function. This period of time ranges from tens of seconds to the entire history period. The observation is not continuously recorded during this period of time. The learning dynamic relationship is achieved by a neural network algorithm. The specific value is a service level agreement or a quality of service requirement. The storage device is a hard disk, a solid state hard disk, a random access memory, or a hybrid combination thereof. The configuration is a percentage of different types of storage devices or a single type of storage device A fixed amount in use. The flow monitoring module, the adaptive neuro-fuzzy inference module, the traffic prediction module, or the fuzzy rule control module is a hardware or a software executed on at least one of the storage nodes.

The present invention utilizes an adaptive fuzzy rule control and operates without human intervention. A storage device read configuration that satisfies any workload can dynamically calculate that the reconfiguration of the storage device can be completed at a particular point in time in the future.

10‧‧‧Adaptive fuzzy rule control system

100‧‧‧ storage node

102‧‧‧Management Server

104‧‧‧hard disk

106‧‧‧ Solid State Drive

120‧‧‧Flow Monitoring Module

140‧‧‧Adaptive Neuro-Fuzzy Inference Module

160‧‧‧Traffic prediction module

180‧‧‧Fuzzy Rule Control Module

Figure 1 illustrates a block diagram of an adaptive fuzzy rule control system in accordance with an embodiment of the present invention.

Figure 2 shows the architecture of a storage node.

Figure 3 illustrates the membership function for the number of input and output operations per second.

Figure 4 illustrates the membership function for the flux.

Figure 5 illustrates the membership function for the percentage of solid state hard disks.

The sixth chart column is a fuzzy rule used by the adaptive fuzzy rule control system.

Figure 7 shows an area used to calculate the percentage of solid state hard disk when the number of input/output operations per second is 70,000 and the throughput is 7 GB/s.

Figure 8 depicts this area in four sub-areas.

The ninth chart column is based on the sub-area for the calculation step of the solid-state hard disk percentage.

The invention will be more specifically described by reference to the following embodiments.

Please refer to Figures 1 to 9. An embodiment of the invention is disclosed herein. 1 is a block diagram of an adaptive fuzzy rule control system 10. System 10 can be used in a network to define a storage system for a software that controls performance parameters in an acceptable range. In this embodiment, the performance parameters include the number of input and output operations per second, the delay time, and the throughput. The network can be the internet. In this way, the storage node 100 can be a database server, manages a large number of storage devices and provides a client cloud service, and can also be a file server or a mail server, and has a dedicated storage device. The network may also be used in a regional network of a laboratory, or a wide area network for a multinational enterprise, and the present invention does not limit the application of the storage node 100. However, storage node 100 must be a software definition store. In other words, the hardware (storage device) of the storage node 100 should be separable from the software managing the storage node 100. The storage node 100 is operated by a software defined by the software definition storage. Therefore, the reconfiguration of the storage device in the storage node 100 can be implemented by a separate software or hardware.

Please see Figure 2. Figure 2 shows the architecture of the storage node 100. The storage node 100 includes a management server 102, 10 hard disks 104, and 10 solid state disks 106. The management server 102 can receive instructions to reconfigure the hard disk 104 and the solid state disk 106. The different configurations of the storage node 100, that is, the percentage of the hard disk 104 and the solid state disk 106 used, can work differently. Maintain a certain delay time. The solid state hard disk 106 has a faster storage speed than the hard disk 104. However, at the same capacity, the price of the solid state drive 106 is much more expensive than the hard disk 104. Normally, at the same cost, the hard disk 104 has a storage capacity that is about ten times that of the solid state disk 106. For such a storage node 100, since the life cycle of the solid state hard disk 106 will drop very fast, it is uneconomical to provide a service in which the solid state hard disk 106 is standby, and when the solid state hard disk 106 is used, the storage capacity Will become a problem. When the configuration of the storage node 100 includes some hard disks 104 and solid state disks 106, the storage node 100 can still meet the requirements of Service Level Agreement or Quality of Service requirements as long as the delay time can meet the requirements of Service Level Agreement or Quality of Service requirements. Run smoothly and avoid the aforementioned problems.

The adaptive fuzzy rule control system 10 includes a flow monitoring module 120, an adaptive neuro-fuzzy inference module 140, a flow prediction module 160, and a fuzzy rule control module 180. The flow monitoring module 120 is used to obtain an observation of the performance parameters of the storage node 100. The adaptive neuro-fuzzy inference module 140 can learn a dynamic relationship between a plurality of storage device configurations and performance parameters in the storage node 100, and output a fuzzy rule, and the fuzzy rule is established according to the dynamic relationship. of. Here, the specific value of the delay time refers to the value required in the service level agreement or the quality of service requirement, which is the longest delay time of the storage node 100, and should be implemented in the service provided under normal use (except for the storage node 100). At boot time or when a large workload occurs). For the present embodiment, the specific value of the delay time is 2 seconds. Any specific value is possible, The invention is not limited. In addition, the period of time ranges from tens of seconds to the entire history period. Thus, the adaptive neuro-fuzzy inference module 140 can have sufficient data to learn and analyze. In practice, observations may not be continuously recorded during this period of time. This means that the adaptive neuro-fuzzy inference module 140 will learn the dynamic relationship with data from different time periods. Learning dynamic relationships can be achieved in a number of ways, and in this embodiment a neural network algorithm is applied.

The adaptive neuro-fuzzy inference module 140 generates a number of membership functions. The membership function is used to define the extent of a performance parameter or the configuration of the storage device at a particular level. A fuzzy rule joins a membership function for a performance parameter at a particular level to a membership function for the configuration of the storage device at another particular level. Please refer to Figures 3 to 5, which are used to describe the membership function and the fuzzy rule in this embodiment.

Figure 3 illustrates the membership function for the number of input and output operations per second. According to fuzzy logic, variables have a true value between 0 and 1. Fuzzy logic has been extended to the notion of dealing with partial truth values, where the true value may be between full truth and complete error. Therefore, when the number of input/output operations per second is greater than or equal to 0 but less than 20,000, the membership function describing the degree of input/output operations per second to the low level is 1.0; when the number of input/output operations per second is greater than or equal to 20,000 but less than 40,000 When the membership function linearly drops from 1.0 to 0; and when the number of input/output operations per second is greater than or equal to 40,000 but less than or equal to 100,000, the membership function is zero. Similarly, when the number of input/output operations per second is greater than or equal to 0 but less than 20,000, the input/output operation per second is described. The membership function is 0 in the middle level; when the number of input/output operations per second is greater than or equal to 20,000 but less than 50,000, the membership function linearly rises from 0 to 1.0; when the number of input and output operations per second is greater than or equal to 50,000 However, when less than 80,000, the membership function linearly drops from 1.0 to 0; and when the number of input/output operations per second is greater than or equal to 80,000 but less than or equal to 100,000, the membership function is zero. When the number of input/output operations per second is greater than or equal to 0 but less than 60,000, the membership function describing the degree of input/output operations per second to the high level is 0; when the number of input/output operations per second is greater than or equal to 60,000 but less than 80,000 The membership function linearly rises from 0 to 1.0; and when the number of input/output operations per second is greater than or equal to 80,000 but less than or equal to 100,000, the membership function is 1.0.

Figure 4 illustrates the membership function for the flux. When the liquidity is greater than or equal to 0 GB/s but less than 2 GB/s, the membership function describing the degree of liquidity at a low level is 1.0; when the liquidity is greater than or equal to 2 GB/s but less than 4 GB/s, the membership function is 1.0 linearly drops to 0; and when the throughput is greater than or equal to 4 GB/s but less than or equal to 10 GB/s, the membership function is zero. Similarly, when the throughput is greater than or equal to 0 GB/s but less than 2 GB/s, the membership function describing the degree of liquidity at the intermediate level is 0; when the throughput is greater than or equal to 2 GB/s but less than 4 GB/s, The membership function linearly rises from 0 to 1.0; when the throughput is greater than or equal to 4 GB/s but less than 6 GB/s, the membership function is 1.0; when the throughput is greater than or equal to 6 GB/s but less than 8 GB/s, the membership is The function linearly drops from 1.0 to 0; and when the throughput is greater than or equal to 8 GB/s but less than or equal to 10 GB/s, This membership function is 0. When the flux is greater than or equal to 0 GB/s but less than 6 GB/s, the membership function describing the degree of flux at a high level is 0; when the flux is greater than or equal to 6 GB/s but less than 9 GB/s, the membership function is 0 linearly rises to 1.0; and when the throughput is greater than or equal to 9 GB/s but less than or equal to 10 GB/s, the membership function is 1.0.

Figure 5 illustrates the membership function for the percentage of solid state hard disks in the storage node 100 (other portions are hard disks). When the percentage of the solid state hard disk is greater than or equal to 0% but less than 20%, the membership function describing the percentage of the solid state hard disk at the low level is 1.0; when the percentage of the solid state hard disk is greater than or equal to 20% but less than 40% The membership function linearly drops from 1.0 to 0; and when the percentage of solid state hard disks is greater than or equal to 40% but less than or equal to 100%, the membership function is zero. Similarly, when the percentage of the solid state hard disk is greater than or equal to 0% but less than 20%, the membership function describing the percentage of the solid state hard disk at the middle level is 0; when the percentage of the solid state hard disk is greater than or equal to 20% but less than At 40%, the membership function rises linearly from 0 to 1.0. When the percentage of solid state hard disk is greater than or equal to 40% but less than 60%, the membership function is 1.0; when the percentage of solid state hard disk is greater than or equal to 60% but When less than 80%, the membership function linearly drops from 1.0 to 0; and when the percentage of solid state hard disks is greater than or equal to 80% but less than or equal to 100%, the example number function is zero. When the percentage of the solid state hard disk is greater than or equal to 0% but less than 60%, the membership function describing the percentage of the solid state hard disk at the high level is 0; when the percentage of the solid state hard disk is greater than or equal to 60% but less than 80% , The membership function linearly rises from zero to 1.0; and when the percentage of solid state hard disks is greater than or equal to 80% but less than or equal to 100%, the example number function is 1.0.

The fuzzy rule joins the membership function for the performance parameters at a particular level to the membership function for the configuration of the storage device at another particular level, as shown in Figure 6. Each rule is connected to the membership function of the number of input and output operations per second and the membership function of the liquidity at a specific level, to the membership function of the percentage of solid state hard disk at a specific level. For example, rule 5 is the percentage of solid state hard drives used in the middle level, which is connected to the number of input and output operations per second and the middle level. Rule 5 means that if the membership function for the number of input/output operations per second is a medium level and the membership function for the throughput is a medium level, the membership function of the corresponding application is also of the medium level. Similarly, the correspondence of all other fuzzy rules can be found in Figure 6. The application of fuzzy rules will be explained later.

The traffic prediction module 160 is configured to provide predicted values of performance parameters at a particular point in time in the future. For example, the traffic prediction module 160 can predict the number of input/output operations per second, the delay time, and the throughput after ten minutes based on the historical data analyzed. The performance parameters are then provided to the fuzzy rule control module 180. The fuzzy rule control module 180 receives these values and can perform related processing after ten minutes. Of course, the traffic prediction module 160 can continuously predict and provide the number of input and output operations and throughput per second. The present invention does not limit the method or apparatus used by the traffic prediction module 160 to provide predictive values for performance parameters.

The fuzzy rule control module 180 is configured to arrange the configuration of the storage device in the storage node 100 according to the fuzzy rule and the predicted value at a specific time point in the future. Thus, a particular value of a particular performance parameter can be achieved at that particular point in time in the future. To explain how the adaptive fuzzy rule control system 10 operates, an example in this embodiment is explained below.

When the traffic prediction module 160 predicts that the number of input and output operations per second and the throughput may be 70,000 and 7 GB/s after ten minutes, it will send the data to the adaptive neuro-fuzzy inference module 140. The degree to which the fuzzy inference method is defined by the membership function is used to obtain the configuration of the storage device with at least a given performance parameter. There are many inferences, such as Mamdani inference or Sugeno inference, that can be used in the field of fuzzy logic, and the invention does not limit which one is used. In the present embodiment, the Mamdani inference method is taken as an explanation. Please refer to Figure 3 again. When the number of input and output operations per second is 70,000, the two points A and B can be found in the middle and high level membership functions. Thus, the degree of correspondence is 0.33 and 0.5, respectively. Similarly, in Fig. 4, when the throughput is 7 GB/s, the two points C and D can be found in the middle and high level membership functions, respectively. Therefore, the degree of correspondence is 0.5 and 0.33, respectively.

Rule 5, Rule 6, Rule 8 and Rule 9 are applied to the above four points. For example, for rule 5, if the membership function for the number of input/output operations per second is at the medium level, and the membership function of the liquidity is at the medium level, the membership function for the percentage of the solid state hard disk is at the medium level. . The percentage of solid state hard disk is used for the number of input and output operations per second and the amount of liquidity The smallest is 0.33. Similarly, in rule 6, the degree for the solid state hard disk percentage is 0.33; in rule 8, the degree for the solid state hard disk percentage is 0.5; in rule 9, the degree for the solid state hard disk percentage is 0.33 . With the Mamdani inference method, the control rule from the above fuzzy rule can be obtained from the calculation of the position of the interlaced line area centroid from the origin of the horizontal axis in Fig. 7. In order to simplify the calculation and description, the calculation of the centroid position first cuts the sub-area of the area of the staggered line into four simple shapes, that is, A1, A2, A3, and A4 in Fig. 8. The calculation table in the adaptive neuro-fuzzy inference module 140 is shown in FIG. The output control rule is 66% solid state hard disk (34% hard disk) for the percentage of solid state hard disk after ten minutes. After the fuzzy rule control module 180 receives the percentage of the solid state hard disk, it can arrange the configuration of the storage node 100 after ten minutes.

It should be noted that the membership function and related fuzzy rules can be set according to the operating experience of the software definition storage system or the dynamic relationship. That is, the membership function and associated fuzzy rules come from a source that can derive the optimal delay time control. The membership functions used are not linear in different intervals, and other types of relationships can be used as long as the control rules derive better control of the delay time. Of course, the adaptive neuro-fuzzy inference module 140 can check for a difference between the specific value of the delay time and the observed value of the delay time from the flow monitoring module 120. Once the difference value exceeds a tolerance value, the adaptive neuro-fuzzy inference module 140 learns the new fuzzy rule and membership function. For example, if an allowed delay time for the storage node 100 is 2 seconds, the tolerance value is 1 second, and when the difference value is 2 seconds, the actual delay time is 4 seconds. This cannot be a service level association Accepted by the council. The current fuzzy rules and membership functions are not applicable to the new state of the storage node 100, and the new fuzzy rules and membership functions must be established again from learning and analyzing the new state of the storage node 100. In the present embodiment, two performance parameters (flow amount and number of input/output operations per second) are connected to the configuration of the storage device (percentage of solid state hard disks). In practice, one, two, or all three performance parameters can be connected to the configuration of the storage device.

It should be emphasized that the storage device is not limited to hard disks and solid state disks, random access memory can be used, mixed combination of hard disk and random access memory, or solid state hard disk and random access memory. The hybrid combination is also applicable. The configuration in this embodiment is the percentage of different types of storage devices in use, and it can also be a fixed number of uses in a single type of storage device (for example, the storage node only contains solid state drives, and the reconfiguration is initiated by standby). In the solid state hard disk to achieve). Most importantly, the traffic monitoring module 120, the adaptive neuro-fuzzy inference module 140, the traffic prediction module 160, or the fuzzy rule control module 180 are hardware or on at least one processor in the storage node 100. The software that is executed.

Although the present invention has been disclosed in the above embodiments, it is not intended to limit the invention, and those skilled in the art can make some modifications and refinements without departing from the spirit and scope of the invention. The scope of the invention is defined by the scope of the appended claims.

10‧‧‧Adaptive fuzzy rule control system

100‧‧‧ storage node

120‧‧‧Flow Monitoring Module

140‧‧‧Adaptive Neuro-Fuzzy Inference Module

160‧‧‧Traffic prediction module

180‧‧‧Fuzzy Rule Control Module

Claims (13)

An adaptive fuzzy rule control system for a software defined storage system for controlling performance parameters in a storage node, comprising: a flow monitoring module for obtaining an observation value of performance parameters in the storage node; an adaptive neural fuzzy inference module a group for learning a dynamic relationship between a plurality of storage device configurations and performance parameters over a period of time in the storage node, and outputting a fuzzy rule, the fuzzy rule being established according to the dynamic relationship; a traffic prediction module, For providing a predicted value of the performance parameter at a specific time point in the future; and a fuzzy rule control module for arranging the predicted value according to the fuzzy rule in the storage node at the specific time point in the future Reconfiguration of the storage device such that a particular value of a particular performance parameter can be achieved at a particular point in time in the future, wherein the storage node is operated by a software-defined storage software; wherein the performance parameter includes the number of input and output operations per second, Delay time, or throughput; and the percentage of the configuration for different types of storage devices A fixed number of single type storage devices in use. The adaptive fuzzy rule control system according to claim 1, wherein the adaptive neural fuzzy inference module generates a plurality of membership functions, each A membership function is used to define the extent of a performance parameter or the configuration of the storage device at a particular level. The adaptive fuzzy rule control system according to claim 2, wherein the fuzzy rule is connected to a membership function of a performance parameter at a specific level to a configuration of the storage device for another specific level. Membership function. The adaptive fuzzy rule control system of claim 3, wherein a fuzzy inference method is used to determine the configuration of the storage device with at least a given performance parameter by the extent defined by the membership function. The adaptive fuzzy rule control system according to claim 4, wherein the fuzzy inference method is a Mamdani inference method or a Sugeno inference method. The adaptive fuzzy rule control system according to claim 1, wherein the adaptive neuro-fuzzy inference module further checks a difference between a specific value of the specific performance parameter and an observed value of the specific performance parameter. . The adaptive fuzzy rule control system according to claim 6, wherein the adaptive neuro-fuzzy inference module learns a new fuzzy rule and a membership function if the difference value exceeds a tolerance value. The adaptive fuzzy rule control system of claim 1, wherein the period of time ranges from tens of seconds to the entire history period. The adaptive fuzzy rule control system of claim 1, wherein the observation value is not continuously recorded during the period of time. The adaptive fuzzy rule control system according to claim 1, wherein the learning dynamic relationship is achieved by a neural network algorithm. The adaptive fuzzy rule control system of claim 1, wherein the specific value is a service level agreement or a quality of service requirement. The adaptive fuzzy rule control system of claim 1, wherein the storage device is a hard disk, a solid state hard disk, a random access memory, or a mixture thereof. The adaptive fuzzy rule control system according to claim 1, wherein the flow monitoring module, the adaptive neuro-fuzzy inference module, the flow prediction module, or the fuzzy rule control module are hardware or Software executed on at least one of the storage nodes.