TWI661313B - System and method for allocation of cloud resources - Google Patents

Abstract

A cloud resource allocation system and method are used for multiple storage systems, which generate multiple attribute weights based on user selection, generate storage capacity indicators for each storage system and multiple attributes for each storage system based on real-time information of each storage system. Normalize the score, and then calculate the supply points for each storage system based on each attribute weight, storage capacity index of each storage system, and each attribute normalization score of each storage system. The selected storage system is distributed to the user.

Description

Cloud resource distribution system and method

This case relates to a cloud resource distribution technology, and more particularly to a cloud resource distribution system and method for a plurality of storage systems.

In the current development of the cloud field, only a few technologies mention how to provide a basic virtual machine for installing resource distribution system and method, but in the actual application of the cloud system, in addition to providing virtual machine services, there are other Cloud service requirements, such as: renting / moving / removing firewalls or installing / moving / removing load balancing services.

In the actual operation of cloud services, because the I / O speed of virtual machines in accessing flexible storage space is often a direct manifestation of cloud performance, how to make a good resource management configuration for a storage system cluster is a challenge.

This case provides a cloud resource allocation system for multiple storage systems, including: a computing module that generates multiple attribute weights and storage capacity indicators for each storage system based on the user's selection and real-time information for each storage system. And a plurality of attribute normalization scores of each of the storage systems, in accordance with each attribute weight, each storage capacity index of each storage system, and each Each attribute normalization score of the storage system calculates the installation points of each storage system; and the installation control module is based on the installation points of each storage system calculated by the calculation module, which will meet the user's The selected storage system is distributed to the user.

In one embodiment, the cloud resource distribution system further includes a resource assignment module, which updates real-time information of a storage system that matches the user's choice to a system database according to the supply points of each storage system.

In one embodiment, the real-time information includes total space capacity, remaining space capacity, space usage, read latency, average reads per second, write latency, average writes per second, speed, transfer volume, and / Or allotment status.

In an embodiment, the calculation module further includes: a weight element that generates the plurality of attribute weights according to the user's selection; a capture element that captures real-time information of each storage system; an indicator element that is based on each The real-time information of the storage system generates storage capacity indicators for each of the storage systems; the normalization element is based on the real-time information of each storage system to generate the plurality of attribute normalization scores of each of the storage systems; the integral element is based on the weight The attribute weights generated by the component, the storage capacity indicators of the storage system generated by the index component, and the attribute normalization scores of the storage system generated by the normalized component, calculate the Supply points; and ranking components, ranking the supply points of each storage system for the supply control module to allocate the storage system with the higher supply points to the user.

In an embodiment, the integration element is a product of the storage capacity index of each storage space multiplied by each attribute normalization score and each attribute weight. Product to obtain the supply points for each storage system. The storage capacity index is related to the remaining space capacity ratio points, the total space capacity points, and the user's previously allocated space capacity points. The plurality of attribute weights include a read delay attribute weight, a read speed attribute weight, a write delay attribute weight, a write speed attribute, and / or a transmission volume attribute weight, and the plurality of normalization scores include a read delay regular Normalization score, read speed normalization score, write latency normalization score, write speed normalization score, and / or transfer volume normalization score.

This case provides a method for allocating cloud resources for a plurality of storage systems, including: generating a plurality of attribute weights according to a user's selection; generating a storage capacity index of each storage system and each of the storages according to real-time information of each storage system; A plurality of attribute normalization scores of the system; according to each attribute weight, the storage capacity index of each storage system, and each attribute normalization score of each storage system, calculating the installation points of each storage system; and according to each The loading points of the storage system will be allocated to the user in accordance with the storage system selected by the user.

In an embodiment, according to each attribute weight, the storage capacity index of each storage system, and the attribute normalization score of each storage system, the step of calculating the installation points of each storage system includes: The storage capacity index of the storage space is multiplied by the product of each attribute normalization score and each attribute weight to obtain the installed points of each storage system.

In an embodiment, the step of allocating a storage system that meets the user's choice to the user according to the supply points of each storage system includes: ranking the supply points of each storage system to have Higher supply The point storage system is distributed to the user.

Therefore, in order to provide users with better cloud service quality, this case provides a weighted multi-dimensional attribute of virtual machine elastic storage space allocation technology for resource allocation of storage system clusters. In storage system clusters and users for elastic storage of virtual machines, Under the requirements of space characteristics, quickly find a suitable storage system to install flexible storage space. For example, a virtual machine's flexible storage space is dedicated to reading data and requires a high reading speed. However, if it is installed in a storage system that can only provide ordinary reading speed, it cannot meet user needs. In addition, this case is designed with a multi-dimensional attribute mechanism, which can use the characteristics of the storage space of the user's needs to apply the characteristics of the field and set the required parameters. At the same time, the monitoring module periodically and dynamically monitors the monitoring of various storage systems returned by the storage system. Index attributes and capacity conditions are calculated through a series of points, and the best points are used as the basis for selecting an appropriate storage system for installing flexible storage space. The point is calculated by multiplying the storage capacity index of the storage system by the attribute weight and the attribute normalization score of the storage system. The storage capacity index is calculated and converted from the remaining space capacity and total space capacity. The attribute weights can be adjusted dynamically, increasing the importance of the attributes of the storage system according to different needs. In order to avoid the different measurement units of attributes of different storage systems, which makes it impossible to compare the attributes of different storage systems, this case is to design a normalization algorithm to normalize the attributes of different storage systems, and then obtain the attribute normalization score of the storage system . Considering that the storage systems have different space sizes and access speeds, the flexible storage space used by users will also be different. This case allows users to allocate to the most appropriate storage system, thereby providing good I / O Access speed optimizes resource utilization.

100‧‧‧Distribution System

103‧‧‧Computation Module

104‧‧‧Resource Assignment Module

105‧‧‧control module

106‧‧‧System Database

107‧‧‧Storage System

108‧‧‧Storage System Cluster

109‧‧‧Monitoring Module

110‧‧‧ Flexible storage space

201‧‧‧Read latency

202‧‧‧write delay

203‧‧‧Space utilization

204‧‧‧Average writes per second

205‧‧‧ average reads per second

206‧‧‧speed

207‧‧‧Total space capacity

208‧‧‧Remaining space capacity

209‧‧‧Transmission

210‧‧‧Custom indicators

301‧‧‧weighting element

302‧‧‧Grab components

303‧‧‧pointer components

304‧‧‧ Normalized component

305‧‧‧Integrating element

306‧‧‧ Ranking components

Figure 1 is a schematic architecture diagram of a cloud resource distribution system according to the case; Figure 2 is a schematic diagram of an embodiment of a monitoring module of the cloud resource distribution system in the case; and Figure 3 is a cloud resource distribution system in the case One embodiment of the flexible storage space product package; FIG. 4 is the real-time information of the storage system of an embodiment of the cloud resource distribution system of the present case; FIG. 5 is an implementation of one of the computing modules of the cloud resource distribution system of the case Example diagram; Figure 6 is a schematic diagram of storage capacity index calculation of one embodiment of the cloud resource distribution system of the present case; Figure 7 is a schematic diagram of total space capacity point conversion of one embodiment of the cloud resource distribution system of the present case; Schematic diagram of conversion of remaining space capacity ratio integrals in one embodiment of the cloud resource distribution system of the present case; FIG. 9 is a schematic diagram of attribute weights of one embodiment of the cloud resource distribution system of the present case; A schematic diagram of the attribute regular drawing scores of an embodiment; and FIG. 11 is a schematic diagram of loading points for an embodiment of the cloud resource distribution system of the present case.

The following describes the implementation of this case through specific examples. Those skilled in the art can easily understand other advantages and effects of this case from the content disclosed herein. The structures, proportions, and sizes shown in the drawings in this specification are only used to match the content disclosed in the specification for the understanding and reading of those skilled in the art, and are not intended to limit the conditions that can be implemented in this case. Therefore, any modification, change or adjustment shall still fall within the scope covered by the technical content disclosed in this case, without affecting the efficacy and purpose that can be achieved in this case.

Refer to FIG. 1, which is a schematic architecture diagram of a cloud resource allocation system for a plurality of storage systems, which can be applied to public cloud or private cloud environments. Users can apply for a variety of cloud services in the user interface (not shown in the figure). When applying for a flexible storage space product, the page of the storage space product package will be displayed for the user to select the flexible storage space product that best suits their needs. As shown in Figure 3. Then, the order processing module (not shown in the figure) can convert user requirements into service orders and send them to the cloud resource distribution system (referred to as the distribution system in the figure) 100 to complete the setting. The cloud resource distribution system 100 includes a computing module 103, a resource assignment module 104, and a supply control module 105.

The calculation module 103 calculates the storage system that is most suitable for distribution by adding up the various indicators. The calculation module 103 can generate corresponding attribute weights, storage capacity indicators of each storage system 107, and a plurality of attribute normalization scores of each storage system 107 according to the user's selection and real-time information of each storage system 107, so as to Weight, storage capacity index of each storage system 107, and each attribute of each storage system 107 are positive The score is normalized, and the supply points of each storage system 107 are calculated.

The resource assignment module 104 can change the information and lease status of the storage system that meets the user's choice (that is, this time to be distributed) according to the supply points of each storage system 107, and then update it back to the system database 106.

The supply and installation control module 105 may allocate the storage system that meets the user's selection to the user based on the supply points of each storage system 107 calculated by the calculation module 103. The supply control module 105 will call the bottom order of the label to achieve the supply of the physical storage system.

The storage system cluster 108 is composed of a plurality of storage systems 107 with different space sizes and speed classes. The system database 106 mainly records the information and distribution status of the system 107 for processing during installation.

Referring to FIG. 2, the cloud resource distribution system 100 further includes a monitoring module 109, which mainly collects real-time information of each storage system 107 and writes it into the system database 106 through regular inquiry to achieve real-time monitoring of the status of the storage system. Each storage system 107 can be equipped with multiple flexible storage spaces 110, which can be distributed to users. Users can use the flexible storage space for application programs or program installation to expand the user's space.

Figure 3 is a flexible storage space product package, which is divided into four fields, namely the name, storage capacity, transmission speed and type. The flexible storage space product package is mainly for users to choose the required product specifications when applying for lease. The cloud distribution system will dynamically install the appropriate storage system and flexible storage space according to related settings to provide high-quality services. Hierarchical agreement. For example, the storage capacity ranges from 100GB to 2TB, the minimum is 100GB, the maximum is 2TB, and the interval is every 100GB. The transmission speed is divided into ordinary type, Standard type and high speed type, the type is divided into reading type and writing type.

Refer to Figure 4, which shows the real-time information of the storage system. The currently more commonly used are read latency 201, write latency 202, space usage 203, average writes per second 204, average reads per second 205, speed 206. The total space capacity 207, the remaining space capacity 208, the transmission volume 209, and / or the custom indicator 210.

Referring to FIG. 5, an embodiment of the calculation module 103 of the resource distribution system 100 includes a weight element 301, a grab element 302, an index element 303, a normalization element 304, an integration element 305, and a ranking element 306. The weighting element 301 converts the flexible storage space product package selected by the user into a plurality of attribute weights with different proportions. The fetching component 302 will go to the system database 106 to fetch real-time information, including read latency, write latency, space usage, average writes per second, average reads per second, speed, total space capacity, remaining space Capacity, transmission volume, distribution status, etc. The indicator component 303 can generate storage capacity indicators of each storage system according to real-time information of each storage system. The normalization element 304 generates a plurality of attribute normalization scores of each storage system according to the real-time information of each storage system. The integration element 305 calculates each storage system according to the attribute weights generated by the weight element 301, the storage capacity indicators of each storage system generated by the index element 303, and each attribute normalization score of each storage system generated by the normalization element 304. It's for loading points. The ranking element 306 ranks the supply points of each storage system, and the higher the score, the higher the priority will be allocated.

Refer to Figure 6 for the calculation of storage capacity indicators. The storage capacity index is related to the remaining space capacity proportional integral, total space capacity integral, and The user has previously allocated space capacity points. As shown in FIG. 6, α is a storage capacity index, and the calculation formula is (1 / (remaining space capacity ratio integral + total space capacity integral + user previously allocated space capacity integral)). The related parameters are described as follows: total space capacity is the total space capacity of the storage system; remaining space capacity is the remaining space capacity of the storage system; the remaining space capacity ratio is the remaining space capacity / total space capacity; the remaining space capacity ratio integral is the remaining space capacity Points after conversion by Fisher series; total space capacity points are points after conversion of total space capacity ratio by Fisher series; when users previously allocated space capacity credits when users applied for flexible storage space, they existed in the storage system. 100 points will be given when applying for rent, otherwise it will be zero points.

The total space capacity and remaining space capacity can be obtained through monitoring. The remaining space capacity ratio is calculated from the remaining space capacity / total capacity. The remaining space capacity ratio integral is the integral of the remaining space capacity ratio after the Fischer series conversion (see Figure 7). The total space capacity points are the points after the conversion of the total space capacity ratio to the Fischer series (see Figure 8). The user's previous allocation of space capacity points is mainly to disperse the flexible storage space allocated by the user to different storage systems. It is not desirable to be allocated on the same storage system. 100 points, which means that the user has previously applied for flexible storage space and allocated it on storage system 2. The main purpose is to spread the risk, and then if the user merges multiple different flexible storage spaces into the same logical partition in the future When the magnetic field is used, parallel computing can be used to allow multiple different flexible storage spaces to be read or written at the same time to accelerate performance.

For example, referring to Figure 7, the higher the total space capacity of the storage system, the greater the amount of flexible storage space that can be stored. When the storage system is damaged, the work that needs to be repaired or moved will depend on the space capacity. The larger the recovery time is, the larger the integral value of the total space capacity is, the higher the disaster recovery risk is. For example, referring to Figure 8, the lower the remaining space capacity ratio, the smaller the available space capacity, the greater the risk of the provision of flexible storage space; conversely, the higher the remaining space capacity ratio, the greater the available space capacity. More, it will be preferentially used to configure flexible storage space for installation. The conversion principles of Figures 7 and 8 are designed using the Fischer series, where the total space capacity field in Figure 7 is 5TB for one step, and the number of lists can increase as the total space capacity increases.

The calculation formula for installed points is as follows:

α is the storage capacity index, and the calculation method has been described above. P is the installation point of the storage system, W _i is the attribute weight of the i-th attribute of the storage system, and N _i is the attribute normalization score of the i-th attribute of the storage system. There are n attributes in total. The supply point P is calculated by multiplying the attribute weight of the i-th attribute of each storage system by the attribute normalization score of the i-th attribute of the storage system and multiplying by the storage capacity Index α, that is, .

The formula for calculating the attribute normalization score is as follows:

A _i is the value of the i-th attribute of the storage system, Average (A) is the average value of all the attributes of the storage system, and K is the attribute index of the storage system. 1. If the value of the storage system attribute is smaller, the better, then set K to -1. The attribute normalization score is calculated by dividing the value of the i-th attribute of the storage system by the average power of K of the attribute of the storage system, ie .

In addition, the calculation formula for the storage capacity index is (1 / (remaining space capacity ratio points + total space capacity points + user previously allocated space capacity points)). Take the column of storage system 1 in Figure 6 for illustration: the total space capacity is 20.8TB, and the remaining space capacity is 5TB. The total space capacity is 20.8TB. It can be seen from Figure 7 that it is located between 20-25TB, and the total space capacity points are 8 points (fifth line). In addition, the remaining space capacity ratio is (5 / 20.8) is 24%. It can be seen from Figure 8 that the position is from 20 to 30 in the middle (sixth line), and the remaining space capacity ratio integral is 13 points. The user's previous allocation of space capacity points is 0 points because the user has not been allocated before. This part of the information is recorded in the system database. Therefore, the total of the remaining space capacity ratio points, the total space capacity points, and the user's previously allocated space capacity points totals 21 points, while the storage capacity index α is 1/21.

The attribute normalization score of the storage system is mainly to convert the attribute values of different units, so that each attribute can be compared. For example, the Read Latency attribute is in milliseconds and the read speed is Read Speed. In terms of megabytes per second, it is meaningless to compare attributes without converting the absolute value of each attribute to a relative value. The formula for normalizing the attributes of the storage system is to calculate the difference between each value and the average, and use the multiple of the difference as the score for normalizing the attributes of the storage system. Figure 10 illustrates the algorithm of the attribute normalization score. There are storage systems 1 to 5. Take the two attributes of read latency (milliseconds) and read speed (how many megabytes per second) as examples. In the read delay attribute, the average value of storage system 1 to 5 is 3000. Because the value of the read delay attribute is as small as possible, the value of the storage system attribute index K is -1, so the read of storage system 1 Taking the delay normalization score as the calculation method is , The value is 3. In the read speed attribute, the average value of the storage system 1 to 5 is 300, because the larger the value of the read speed attribute, the better, the value of the storage system attribute index K is 1, so the read delay of the storage system 1 The normalized score is calculated as With a value of 0.67.

The attribute weight of the storage system is in a one-to-one relationship with the product package setting. One product package corresponds to one storage system attribute. Figure 9 shows an example of product package weights. Each product package has different attributes.

When calculating the storage system installation points, it is necessary to obtain the storage capacity index α, the attribute weight W of all attributes of the storage system, and the attribute normalization score N of all attributes of the storage system. Figure 11 is an example of storage system installation credits. In this example, the storage system 5 with the highest installation credits is selected as the installation target. Among them, the storage capacity indicator α is the storage capacity indicator in Figure 6. The calculation formula, wherein the weight source of all attributes of the storage system is an example of the weight of the elastic storage space product package in FIG. 9, and the source of the attribute normalization score of all attributes of the storage system is an example of the algorithm of the attribute normalization score of FIG.

According to the above description of Figures 1 to 11, it can be seen that this case also provides a method for allocating cloud resources for a plurality of storage systems, which includes: generating a plurality of attribute weights according to a user's selection; The storage capacity index of each storage system and the plurality of attribute normalization scores of each storage system; calculated based on each attribute weight, the storage capacity index of each storage system, and each attribute normalization score of each storage system, and calculate Installation points for each of the storage systems; and according to the installation points for each of the storage systems, a storage system that meets the user's choice is allocated to the user. In addition, according to each attribute weight, the storage capacity index of each storage system, and each attribute normalization score of each storage system, the step of calculating the loading points of each storage system includes: The storage capacity index is multiplied by the product of each attribute normalization score and each attribute weight to obtain the supply points for each storage system. In addition, according to the supply points of each storage system, the step of allocating a storage system that meets the user's selection to the user includes: ranking the supply points of each storage system to have a higher supply The point storage system is distributed to the user.

Therefore, this case can be described as a virtual machine flexible storage space allocation technology with weighted multi-dimensional attributes. The storage system information fed back by monitoring systems such as user requirements, resource capacity status, and performance indicators is regarded as multi-dimensional. Attributes. These attributes are normalized to obtain a normalized score, and then the supply points are calculated based on the weight of each attribute, which is used as a basis for selecting a storage system for installing virtual machine flexible storage space resources.

The above-mentioned embodiments only exemplify the effectiveness of this case, and are not intended to limit this case. Anyone familiar with this technique can modify and change the above implementations without departing from the spirit and scope of this case. Therefore, the scope of protection of the rights in this case should be as listed in the scope of patent application mentioned later.

Claims (10)

A cloud resource distribution system for a plurality of storage systems includes a calculation module that generates a plurality of attribute weights, storage capacity indicators of each storage system, and each according to a user's selection and real-time information of each storage system. A plurality of attribute normalization scores of the storage system to calculate the supply points for each storage system based on each attribute weight, each storage capacity index of each storage system, and each attribute normalization score of each storage system; and The installation of the control module is based on the supply points of each of the storage systems calculated by the calculation module, and allocates the storage system that meets the user's choice to the user. The cloud resource allocation system described in item 1 of the patent application scope further includes a resource assignment module, which updates real-time information of a storage system that meets the user's choice to a system based on the supply points of each storage system. database. The cloud resource allocation system described in item 1 of the scope of patent application, wherein the real-time information includes total space capacity, remaining space capacity, space usage, read latency, average reads per second, write latency, Average number of writes per second, speed, transfer volume, or distribution status. According to the cloud resource allocation system described in item 1 of the patent application scope, wherein the calculation module further includes: a weight element, which generates the plurality of attribute weights according to the user's selection; and a capture element, which captures each of the The real-time information of the storage system; the index component generates the storage capacity index of each storage system according to the real-time information of each storage system; the normalized component generates the plurality of storage systems according to the real-time information of each storage system Attribute normalization score; the integral element is based on each attribute weight generated by the weight element, the storage capacity index of each storage system generated by the index element, and each of the storage system generated by the normalization element. The attribute normalizes the score to calculate the supply points for each storage system; and the ranking element is to rank the supply points for each storage system so that the supply control module will have a higher storage point for storage The system is distributed to the user. The cloud resource allocation system described in item 1 of the scope of the patent application, wherein the calculation module multiplies the storage capacity index of each storage space by the product of each attribute normalization score and each attribute weight to Earn installation points for each storage system. The cloud resource allocation system described in item 1 of the scope of patent application, wherein the storage capacity index is related to the remaining space capacity ratio points, the total space capacity points, and the user's previously allocated space capacity points. The cloud resource allocation system according to item 1 of the scope of patent application, wherein the plurality of attribute weights include a read delay attribute weight, a read speed attribute weight, a write delay attribute weight, a write speed attribute, or a transmission amount. Attribute weight, and wherein the plurality of normalization scores include a read delay normalization score, a read speed normalization score, a write delay normalization score, a write speed normalization score, or a transmission volume normalization score. A method for allocating cloud resources for a plurality of storage systems includes: generating a plurality of attribute weights according to a user's selection; generating a storage capacity index of each storage system according to real-time information of each storage system; A plurality of attribute normalization scores; based on each attribute weight, the storage capacity index of each storage system, and each attribute normalization score of each storage system, calculating the installed points of each storage system; and according to each of the storages The installation points of the system will be allocated to the user with a storage system that meets the user's choice. The method for allocating cloud resources as described in item 8 of the scope of the patent application, wherein each storage is calculated according to each attribute weight, each storage capacity index of each storage system, and each attribute normalization score of each storage system. The step of installing points of the system includes: multiplying the storage capacity index of each of the storage spaces by the product of each attribute normalization score and each of the attribute weights to obtain the points of installation of each storage system. The method for allocating cloud resources as described in item 8 of the scope of patent application, wherein the step of allocating a storage system that meets the user's choice to the user according to the supply points of each storage system includes: The storage system installation points are ranked to allocate the storage system with a higher installation point to the user.