HK1251377B - Method for determining data in cache memory of cloud storage architecture and cloud storage system using the same - Google Patents

Description

Cloud storage device system and method for determining data in cache of its architecture

Technical Field

The present invention relates to the field of cloud storage devices, and in particular to a cloud storage device system and a method for determining data in a high-speed cache of the cloud storage device system.

Background Art

Cloud service systems typically strive to provide their services to clients as quickly as possible to respond to their requests. This goal is easily achieved when the number of clients is small. However, if the number of clients increases significantly, response times will inevitably vary, subject to the constraints of the cloud service system's hardware architecture and network traffic, but they should remain within a reasonable range. On the other hand, if a cloud service competes commercially with other cloud services, regardless of its limitations, the cloud service system should technically respond to client requests in the shortest possible time using limited resources. This is a common issue faced by many cloud system developers, and everyone is eager for a suitable solution.

In a traditional work environment, as shown in Figure 1, numerous client computers 1 are connected to a server 4 via the Internet 3. Server 4 primarily processes client requests and may perform complex computations or simply access stored data. In the latter case, stored data may be stored in cache 5 or auxiliary memory 6. The number of caches 5 or auxiliary memory 6 is not limited to one; it can be any number required by the cloud service. Server 4, cache 5, and auxiliary memory 6 form the architecture of the cloud service system. Cache 5 may be dynamic random access memory (DRAM) or static random access memory (SRAM). Auxiliary memory 6 may be a solid-state drive (SSD), a hard disk drive (HDD), a writable digital versatile disc (DVD), or even magnetic tape. The physical difference between cache 5 and auxiliary memory 6 lies in their ability to store data during power outages. In cache 5, data is temporarily stored when needed and disappears when power is lost. However, no matter whether it is powered on or not, the auxiliary memory 6 can store data permanently. The cache 5 has the advantage of fast access to data, but has the disadvantages of volatility, high price and small storage space.

As described above, it's clear that determining the appropriate data to store in cache 5 is crucial for ensuring that hot data (more frequently accessed) required by most requests is quickly accessible, while cold data (less frequently accessed) is served at a tolerably slow rate. This improves cloud service performance. On average, the response time for all client computer requests should fall within an acceptable range. Currently, many conventional algorithms are used to determine which data should be cached (stored in cache 5). For example, the Least Recently Used (LRU) algorithm, the Most Recently Used (MRU) algorithm, the Pseudo-LRU (PLRU) algorithm, the Segmented LRU (SLRU) algorithm, the 2-way set associative algorithm, the Least-Frequently Used (LFU) algorithm, the Low Inter-reference Recent Set (LIRS) algorithm, and others. These algorithms are based on the recency and frequency characteristics of the data being analyzed, and their results are independent of other data (i.e., lack data-related characteristics). These algorithms are categorized as "data-dependent" algorithms, using the original cache data (results from the aforementioned traditional caching algorithms) as target data to obtain "data-dependent" data for caching. This means that the new cache data is somewhat related to the original cache data (new cache data has a higher probability of appearing with the original cache data). These algorithms have been found to be effective for certain workload patterns. However, because they all calculate data occurring in relative time periods, rather than absolute time periods, this leads to a phenomenon: data selected by all algorithms for caching in the first time period (e.g., the first eight hours) may not necessarily be accessed in the second time period (e.g., the eight hours after the first eight hours). This is understandable, as almost all data access is absolute time-dependent or frequency-dependent. For example, computer startups between 8:55 AM and 9:05 AM every morning, meetings at 2:00 PM every Wednesday, biweekly payroll, and inventory checks performed on the last day of each month. Therefore, the timestamp itself is an important and independent factor in considering cached data. However, currently, no solution considers caching based on timestamps.

Summary of the Invention

In view of this, it is necessary to address the problem that traditional technologies do not consider cache storage solutions based on timestamps, and to provide a cloud storage device system that analyzes which data should be cached based on time-related data accessed within a period of time, thereby improving the performance of the cloud storage device system and a method for determining the data in the cache of its architecture.

The present invention provides a method for determining data in a cache of a cloud storage device architecture, the method comprising the steps of:

A. Recording the cache contents of the cloud storage system over a period of time, where each transaction includes the recording time, or the recording time and the cache data accessed during that period of time;

B. Specify a specific time in the future;

C. Calculating a time-related confidence score for each cached data from the processed content based on a reference period;

D. rank the confidence levels associated with time; and

E. providing cache data with higher time-related confidence in the cache, and removing cache data with lower time-related confidence in the cache when the cache is exhausted before the specific time in the future.

In one embodiment, step E can be replaced by step E':

E', providing cache data with a higher time-related confidence and data calculated from at least one other cache algorithm to the cache to exhaust the cache before the specified time in the future, wherein there is a fixed ratio between the cache data with a higher time-related confidence and the data calculated from the other cache algorithm.

In one embodiment, the fixed ratio is calculated based on the amount of data or the space occupied by the data.

In one embodiment, the specific time includes a specific minute of an hour, a specific hour of a day, a specific day of a week, a specific day of a month, a specific day of a quarter, a specific day of a year, a specific week of a month, a specific week of a quarter, a specific week of a year, or a specific month of a year.

In one embodiment, the processing contents are recorded periodically with a time span between two consecutively recorded processing contents.

In one embodiment, the reference period includes a specific minute in an hour, a specific hour in a day, or a specific day in a year.

In one embodiment, the time-related confidence level is calculated by the following steps:

C1. Calculate a first quantity, where the first quantity is the quantity that occurred during the past period of time in the reference time period;

C2. Calculate a second number, the second number being the number of the reference time periods when the target cache data is accessed; and

C3. Divide the second amount by the first amount.

In one embodiment, the cache algorithm includes a Least Recently Used (LRU) algorithm, a Most Recently Used (MRU) algorithm, a Pseudo-LRU (PLRU) algorithm, a Random Replacement (RR) algorithm, a Segmented LRU (SLRU) algorithm, a 2-way set associative algorithm, a Least-Frequently Used (LFU) algorithm, a Low Inter-reference Recent Set (LIRS) algorithm, an Adaptive Replacement Cache (ARC) algorithm, a Clock with Adaptive Replacement (CAR) algorithm, a Multi Queue (MQ) algorithm, or a data-related algorithm using the result from step D as the target data.

In one embodiment, the data type includes an object, a block, or a file.

The present invention also provides a cloud storage device system, which includes:

Host, used to access data;

a cache memory connected to the host computer for temporarily storing cache data for quick access;

a transaction content recorder configured or installed in the cache, connected to the host to record transaction contents of the cache over a past period of time, wherein each transaction content includes a recorded time, or a recorded time and cache data accessed during the past period of time, receives a specific time in the future specified by the host, calculates a time-related confidence for each cache data from the transaction content based on a reference time period, sorts the time-related confidence, and provides cache data with higher time-related confidence in the cache, and removes cache data with lower time-related confidence from the cache when the cache is exhausted before the specific time in the future;

A plurality of auxiliary memories are connected to the host and used for distributing and storing data for access.

In one embodiment, the cloud storage device system may also include:

Host, used to access data;

a cache memory connected to the host computer for temporarily storing cache data for quick access;

a transaction content recorder configured or installed in the cache, connected to the host to record transaction contents of the cache over a past period of time, wherein each transaction content includes a recorded time, or a recorded time and cache data accessed during the past period of time, receives a specific time in the future specified by the host, calculates a time-related confidence score for each cache data from the transaction content based on a reference time period, sorts the time-related confidence scores, and provides cache data with higher time-related confidence scores and data calculated from at least one other cache algorithm to the cache to deplete cache usage before the specific time in the future, wherein there is a fixed ratio between cache data with higher time-related confidence scores and data calculated from at least one other cache algorithm; and

A plurality of auxiliary memories are connected to the host and used for distributing and storing data for access.

In one embodiment, the fixed ratio is calculated based on the amount of data or the space occupied by the data.

In one embodiment, the specific time includes a specific minute of an hour, a specific hour of a day, a specific day of a week, a specific day of a month, a specific day of a quarter, a specific day of a year, a specific week of a month, a specific week of a quarter, a specific week of a year, or a specific month of a year.

In one embodiment, the processing contents are recorded periodically with a time span between two consecutively recorded processing contents.

In one embodiment, the reference period includes a specific minute in an hour, a specific hour in a day, or a specific day in a year.

In one embodiment, the time-related confidence level is calculated by the following steps:

C1. Calculate a first quantity, where the first quantity is the quantity that occurred during the past period of time in the reference time period;

C2. Calculate a second number, the second number being the number of the reference time periods when the target cache data is accessed; and

C3. Divide the second number by the first number.

In one embodiment, the cache algorithm includes an LRU algorithm, an MRU algorithm, a PLRU algorithm, an RR algorithm, an SLRU algorithm, a 2-way set associative algorithm, an LFU algorithm, a LIRS algorithm, an ARC algorithm, a CAR algorithm, an MQ algorithm, or a data-related algorithm that uses data generated by a content recorder as target data.

In one embodiment, the data type includes an object, a block, or a file.

The beneficial effects of the present invention include at least:

In the aforementioned cloud storage system and method for determining cache data within its architecture, cached data is time-dependent. Therefore, when the next relevant time arrives, the data is most likely to be accessed. Prior to that relevant time, the data can be cached, improving the performance of the cloud storage system. This is unattainable with traditional caching algorithms.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of a data access architecture in a conventional embodiment;

FIG2 is a schematic diagram of the structure of a cloud storage device system in one embodiment;

FIG3 is a window for processing content records in one embodiment;

FIG4 is a flow chart illustrating a method for determining data in a cache of a cloud storage device architecture according to one embodiment;

FIG5 is a schematic diagram of a table showing time-related confidence scores calculated for all cached data in one embodiment;

FIG. 6 is a schematic diagram showing a list of time-related confidence scores calculated for all cached data in another embodiment.

DETAILED DESCRIPTION

The present invention will be described in more detail with reference to the following embodiments.

To make the objectives, technical solutions, and advantages of the present invention more clearly understood, the following describes in further detail the cloud storage device system and the method for determining cache data within its architecture, in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are intended only to illustrate the present invention and are not intended to limit the present invention.

FIG2 illustrates an idealized architecture for practicing one embodiment of the present invention. A cloud storage system 10 includes a host 101, a cache 102, a content recorder 103, and multiple auxiliary memories 104. Cloud storage system 10 supports data storage for cloud services and may be partially installed in a server 100, as shown in FIG2 . Server 100 is hardware used to receive requests from client devices, such as personal computers 301, tablet computers 302, and smartphones 303, or other remote devices connected via the Internet 200. After executing these requests, server 100 transmits corresponding responses back to the client devices. Each component is described in detail below.

The primary function of host 101 is to respond to requests from client devices and perform data access. In practice, host 101 may be a controller within server 100. In other embodiments, if the central processing unit (CPU) of server 100 performs the same functions as the controller, host 101 may refer to the CPU, or even to server 100 itself. The definition of host 101 is not determined by its form factor, but rather by its function. Furthermore, host 101 may have other functions, such as caching hot data for storage in cache 102, but this is outside the scope of this invention.

The cache 102 is connected to the host 101 and can temporarily store cache data for fast access. In practice, the cache 102 can be any hardware that provides high-speed access to data. For example, the cache 102 can be SRAM (Static Random Access Memory). The cache 102 can be an independent module for a large cloud storage device system, and some architectures can embed the independent module into the host 101 (CPU). Like caches in other cloud storage device systems, there is a default cache algorithm to determine which data should be cached and stored in the cache 102. In one embodiment, a parallel mechanism is provided to operate in conjunction with the existing cache algorithm for a specific purpose or time. In fact, the cache mechanism can also dominate and replace the cache data determined by the original cache algorithm.

The processing content recorder 103 is an essential component of the cloud storage device system 10. In this embodiment, it is a hardware module and is configured in the cache 102. In other embodiments, the processing content recorder 103 may be software installed in the cache 102 or the controller of the host 101. In this embodiment, the processing content recorder 103 is connected to the host 101, and many of its functions are features of the present invention: recording the processing content of the cache 102 over a past period of time, wherein each processing content includes a recording time, or a recording time and cache data accessed during the past period of time, receiving a specific time in the future specified by the host 101, calculating a time-related confidence for each cache data from the processing content based on a reference time period, sorting these time-related confidences, and providing cache data with higher time-related confidence to the cache 102, and when the cache 102 is exhausted before the specific time in the future, removing the cache data with lower time-related confidence in the cache 102 (or providing cache data with higher time-related confidence and data calculated from at least one other cache algorithm to the cache 102 to exhaust the use of the cache 102 before the specific time in the future). These functions will be described later in conjunction with the method proposed by the present invention. It is important to emphasize that the term "time-dependent confidence" used in this invention is similar to the confidence defined in relational rules. This time-dependent confidence further extends to a confidence value calculated by taking a specific time or period as the target and obtaining the probability that one or more data items have been accessed in the past history.

Auxiliary memory 104 is also connected to host 101 and can store data in a distributed manner for access upon request. Unlike cache 102, auxiliary memory 104 has a slower input/output speed, resulting in slower access to any data stored therein in response to access requests. Frequently accessed data in auxiliary memory 104 is copied and stored in cache 102 for caching. In practice, auxiliary memory 104 can be an SSD (Solid State Drive), an HDD (Hard Disk Drive), a writable DVD (Digital Versatile Disc), or even a magnetic tape. The configuration of auxiliary memory 104 depends on the purpose of the cloud storage system 10 or the workload running on it. In this embodiment, there are three auxiliary memories 104. In reality, a cloud storage system may have hundreds to thousands of auxiliary memories, or even more.

It should be noted that certain definitions used in the present invention require prior explanation. Please see Figure 3, which shows a window for processing content records, used to monitor how data in cache 102 has been accessed in the past. This window includes a TID (Processing Content ID, 0001 to 0024) column, a Cache Data column (D01 to D18), a Reference Period column (H00 to H07), and a Recording Time column. H00 indicates a recording time between 00:00 and 01:00, H01 indicates a recording time between 01:00 and 02:00, and so on. A "1" in the TID and Cache Data fields indicates that the corresponding cache data has been accessed before the "Current" recording time and the "Last" recording time. A "1" in the TID and Reference Period fields indicates that the recording time in different time periods is quantified for processing content. Processing content is a record of cache data accesses during the past period. In this embodiment, records (processing content) from the past eight hours are analyzed. For better illustration, each processing content has a corresponding TID for identification. The processing content recorder 103 periodically records these processing contents, using the time span between two consecutively recorded processing contents. In this embodiment, each processing content is recorded 20 minutes after the previous processing content is recorded, with a time span of 20 minutes. In practice, the recording time does not necessarily need to fall exactly on the predetermined schedule. For example, the recording time may fall at 00:30:18, 00:50:17, etc., not exactly at the 15th second, but within a range. This is because some large data may be being accessed, or the processing content recorder 103 may be waiting for a response from the remotely connected cache 102. A more proactive approach is to randomly select the time span, which also falls within the scope of the present invention.

It should be noted that the number of processed contents is large, potentially thousands or more. For example, a three-month recording with a 10-minute time span is used, with 24 processed contents as a specific example. With more processed contents in the processed content recorder 103, data demand at a specific time in the future can be more accurately predicted. Of course, not all data cached in the cache 102 will be accessed within a certain period of time. As shown in Figure 3, processed content 0015 does not record any accessed data; it only records the time, 04:50:05.

Before disclosing the details of how the data in cache 102 is determined by the method of the present invention via cloud storage system 10, let's first look at the cache data. Although there are 18 cached data entries, the number of cached data entries may be greater than 18 depending on the capacity of cache 102. These 18 cached data entries were obtained at 07:50:05 by the method of the present invention and/or other cache algorithms used by cloud storage system 10. Because if certain data is accessed too frequently, the process content recorder 103 may add new data to cache 102 from one of the auxiliary memories 104, and the cached data used for analysis may also change. Other data may have been cached before 03:50:05 but was later removed because it was not requested or "expected to be accessed."

Figure 3 illustrates the characteristics of cache data. Cache data D01 is frequently accessed in the first three hours and the last hour. Cache data D02 is accessed on average every 20 minutes. Cache data D03 is accessed on average every two 20 minutes. Cache data D04 is accessed on average between 00:10:05 and 00:30:05, 02:50:05 and 03:10:05, and 05:30:05 and 05:50:05. Cache data D05 is accessed between 00:30:05 and 00:50:05 and 06:10:05 and 06:30:05. Cache data D06 is accessed only between 05:30:05 and 05:50:05. Cache data D07 is accessed evenly between 00:30:05 and 01:10:05, 03:10:05 and 03:50:05, and 06:10:05 and 06:50:05. Cache data D08 is accessed only between 07:10:05 and 07:30:05 and may be the latest data added after 07:10:05 in anticipation of demand. Cache data D09 is accessed most frequently in almost every period except 04:30:05 and 04:50:05. Cache data D10 is accessed randomly. There is no record of cache data D11 being accessed. Cache data D12 is accessed evenly within 40 minutes of every 20-minute interval. Cache data D13 is accessed randomly. Cache data D14 is accessed frequently between 00:50:05 and 04:30:05. Cache data D15 is accessed frequently between 02:50:05 and 06:50:05, except between 04:30:05 and 04:50:05. Cache data D16 has similar access requirements to cache data D01. Cache data D17 and D18 are both accessed evenly, but cache data D17 has more requests between 03:50:05 and 04:30:05, and cache data D18 has more requests between 01:50:05 and 03:10:05.

The main purpose of the present invention is to predict the data requested at a specific time in the future based on historical information, and to provide the corresponding data to the cache 102 before the specific time in the future arrives. A method for determining the data in the cache 102 of the cloud storage device system 10 has several steps. Please see Figure 4, which is a flow chart of the method proposed by the present invention. As described above, the method is executed by the processing content recorder 103. First, S01, the processing content of the cache 102 of the cloud storage device system 10 in the past period of time is recorded. Each processing content includes a recording time (processing content ID 0015), or a recording time and cache data accessed within the past period of time (8 hours in the example). Then S02, a specific time in the future is specified. The cache 102 receives the specific time in the future from the host 101. In this embodiment, the specific time in the future can be any time or period in the future. For example, it can be a specific minute of the hour (for each hour), a specific hour of the day (for each day), a specific day of the week (for each week), a specific day of the month (for each month), a specific day of the quarter (for each quarter), a specific day of the year (for each year), a specific week of the month (for each month), a specific week of the quarter (for each quarter), a specific week of the year (for each year), or a specific month of the year (for each year). In this embodiment, the processing content is used to determine which data should be cached before 00:00:00 (H00) on other days.

The third step, S03, is to calculate a time-related confidence score for each cached data item from the processed content based on a reference time period. The reference time period refers to "within a specific minute of an hour" (H00, every 20 minutes of the first hour of each day). In other examples, the reference time period may be "within a specific hour of a day" or "within a specific day of a year," depending on the number of time span records. In a specific example, the reference time period may be "within a specific sub-time unit within a primary time unit," for example, within a 24-hour day. The time-related confidence score can be calculated by the following steps: A. calculating a first number, which is the number of times the reference time period occurred within the past time period; B. calculating a second number, which is the number of times the target cached data was accessed within the reference time period; and C. dividing the second number by the first number. In this embodiment, FIG5 shows a list of the calculated time-related confidence scores for all data items. If the future specific time is the first minute of 8:00 AM, and the reference time period refers to all 20 minutes within the past 8 hours, the results are shown in FIG6. 5 and 6 , based on different situations, each cache data has a different calculated time-related confidence relative to other cache data.

Next, S04, these time-related confidences are sorted. The results of the example are also shown in Figures 5 and 6, respectively. Finally, S05, cache data with higher time-related confidences are provided to cache 102, and when cache 102 is exhausted before a specific time in the future, cache data with lower time-related confidences are removed from cache 102. Figure 6 is used as an example for illustration. Before 00:00 on other days, perhaps at 11:59:59 PM, all data except D11 is stored as new cache data in cache 102 for access requests after 00:00. The reason for D11's removal is that cache 102 does not have enough space to store 18 data items, and D11 has a lower time-related confidence than the other data items. The 18 cached storage files are used for analysis because one or more cached data items have been removed from the cloud storage device system 10 due to low hit rates or other factors, as well as the addition of new data (D08). The total number of cached data used is 18. The newly cached data in cache 102 is the data most likely to be requested after 08:00, which is calculated based on the confidence associated with time. It should be noted that the data or cached data types mentioned above can be objects, blocks, or files.

In other embodiments, the last step (S05) may be different, meaning that the processing content recorder 103 functions differently than in the previous embodiment. The modified step involves providing cache data with a higher time-related confidence and data calculated from at least one other cache algorithm to cache 102, so as to exhaust cache 102 before a specific time in the future. A fixed ratio exists between cache data with a higher time-related confidence and data calculated from other cache algorithms, calculated based on the amount of data or the space occupied by the data. Returning to Figure 6 , if cache 102 is configured to cache 20 data items, and the ratio used for caching data as described in the present invention is 60%, with the remaining 40% being data calculated from other cache algorithms, the cache data obtained by this method is D01, D02, D03, D07, D09, D10, D12, D13, D14, D15, D16, and D18, a total of 12 data items. The remaining data is provided by the aforementioned cache algorithms. If some of the same cache data is requested by both parties, the data with a lower priority, as determined by the present invention or another cache algorithm, may be used instead. This is not a limitation of the present invention. Of course, in most cases, cache 102 is designed to cache data based on its capacity, not the amount of data. In the example above, 60% of cache 102's capacity should be reserved for the data determined by the present invention, while the remaining 40% should be reserved for data requested by at least one existing cache algorithm. The aforementioned cache algorithms include, but are not limited to, the Least Recently Used (LRU) algorithm, the Most Recently Used (MRU) algorithm, the Pseudo-LRU (PLRU) algorithm, the Random Replacement (RR) algorithm, the Segmented LRU (SLRU) algorithm, the 2-way set associative algorithm, the Least-Frequently Used (LFU) algorithm, the Low Inter-reference Recent Set (LIRS) algorithm, the Adaptive Replacement Cache (ARC) algorithm, the Clock with Adaptive Replacement (CAR) algorithm, the Multi Queue (MQ) algorithm, or the data-dependent algorithms defined in the background of the invention. It should be noted that if a data-dependent algorithm is applied, the target data should use the result of the operation of the present invention. This means that the cache data with a higher ranking obtained in step S04 is then input into the data-dependent algorithm as the target data to obtain the result of the data-dependent algorithm. In the cloud storage device system 10, the target data is generated by processing the content recorder 103 for use by the data-dependent algorithm. Data-related algorithms may also be executed using the process content recorder 103 .

The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be determined by the appended claims.

【Explanation of symbols】

1 Client computer; 3 Internet; 4 Server; 5 Cache; 6 Auxiliary memory; 10 Cloud storage system; 100 Server; 101 Host; 102 Cache; 103 Processing content recorder; 104 Auxiliary memory; 200 Internet; 301 Personal computer; 302 Tablet computer.

Claims (16)

1. A method for determining data in the cache of a cloud storage device architecture, characterized in that the method includes the following steps: A. Record the processing content of the high-speed cache of the cloud storage device system in the past period of time, wherein each processing content includes the recording time, or the recording time and the cached data accessed in the past period of time, wherein the recording time is the time when the cached data is accessed. B. Specify a particular time in the future; C. Based on a reference time period, calculate a time-related confidence level for each cached data from the processed content, wherein the reference time period is a specific sub-time unit selected in the main time unit; D. Rank the time-related confidence levels; and E. Before the specified future time, add cached data with high time-related confidence to the cache, and when the cache is exhausted before the specified future time, remove cached data with low time-related confidence from the cache. The time-related confidence level is calculated using the following steps: C1. Calculate the first quantity, which is the number of times the reference time period occurred in the past period. C2. Calculate the second quantity, which is the quantity of the reference time period when the target cache data is accessed; and C3. Divide the second quantity by the first quantity. 2. The method as described in claim 1, wherein the specific time includes a specific minute in an hour, a specific hour in a day, a specific day in a week, a specific day in a month, a specific day in a quarter, a specific day in a year, a specific week in a month, a specific week in a quarter, a specific week in a year, or a specific month in a year. 3. The method as described in claim 1, wherein step A periodically records the processed content at regular intervals. 4. The method as claimed in claim 1, wherein the reference time period includes a specific minute within an hour, a specific hour within a day, or a specific day within a year. 5. The method as described in claim 1, wherein the data type includes objects, blocks, or files. 6. A method for determining data in a cache of a cloud storage device architecture, characterized in that the method includes the steps of: A. Record the processing content of the high-speed cache of the cloud storage device system in the past period of time, wherein each processing content includes the recording time, or the recording time and the cached data accessed in the past period of time, wherein the recording time is the time when the cached data is accessed. B. Specify a particular time in the future; C. Based on a reference time period, calculate a time-related confidence level for each cached data from the processed content, wherein the reference time period is a specific sub-time unit selected in the main time unit; D. Rank the time-related confidence levels; and E. Before the specified future time, cached data with high time-related confidence and data calculated from at least one other caching algorithm are added to the cache for use by the cache at the specified future time, wherein a fixed ratio exists between the cached data with high time-related confidence and the data calculated from other caching algorithms. The time-related confidence level is calculated using the following steps: C1. Calculate the first quantity, which is the number of times the reference time period occurred in the past period. C2. Calculate the second quantity, wherein the second quantity is the quantity of the reference time period when the target cache data is accessed; and C3. Divide the second quantity by the first quantity. 7. The method as described in claim 6, wherein the fixed ratio is calculated based on the amount of data or the space occupied by the data. 8. The method as described in claim 6, wherein the cache algorithm includes LRU algorithm, MRU algorithm, PLRU algorithm, RR algorithm, SLRU algorithm, 2-way set associative algorithm, LFU algorithm, LIRS algorithm, ARC algorithm, CAR algorithm, MQ algorithm, or a data-related algorithm that uses the result from step D as the target data. 9. A cloud storage device system, characterized in that the system comprises: The host is used to access data; A high-speed cache, connected to the host, is used to temporarily store cached data for fast access; A content recorder, configured or installed to the cache, connects to the host to record cached content processed over a past period, wherein each piece of content includes a recording time, or a recording time and cached data accessed during that past period, receiving a specific future time specified by the host, calculating a time-related confidence score for each cached data from the processed content based on a reference time period, sorting the time-related confidence scores, and adding cached data with higher time-related confidence scores to the cache before the specific future time, and removing cached data with lower time-related confidence scores from the cache when the cache is exhausted before the specific future time, wherein the recording time is the time when the cached data is accessed, and the reference time period is a specific sub-time unit selected in the main time unit; and Multiple auxiliary memory modules, connected to the host, are used to distribute and store data for access. The time-related confidence level is calculated using the following steps: C1. Calculate the first quantity, which is the number of times the reference time period occurred in the past period. C2. Calculate the second quantity, wherein the second quantity is the quantity of the reference time period when the target cache data is accessed; and C3. Divide the second quantity by the first quantity. 10. The cloud storage device system as claimed in claim 9, characterized in that there exists a fixed ratio between cached data with a high time-related confidence level and data calculated from other caching algorithms, the fixed ratio being calculated based on the amount of data or the space occupied by the data. 11. The cloud storage device system as described in claim 9, wherein the specific time includes a specific minute in an hour, a specific hour in a day, a specific day in a week, a specific day in a month, a specific day in a quarter, a specific day in a year, a specific week in a month, a specific week in a quarter, a specific week in a year, or a specific month in a year. 12. The cloud storage device system as described in claim 9, wherein the processing content recorder periodically records the processing content at intervals of a certain time span. 13. The cloud storage device system as claimed in claim 9, wherein the reference time period includes a specific minute within an hour, a specific hour within a day, or a specific day within a year. 14. A cloud storage device system, characterized in that the system comprises: The host is used to access data; A high-speed cache, connected to the host, is used to temporarily store cached data for fast access; A content recorder, configured or installed to the cache, is connected to the host to record cached content processed over a past period, wherein each piece of content includes a recording time, or a recording time and cached data accessed during that past period; receiving a host specifying a future specific time; calculating a time-related confidence score for each cached data from the processed content based on a reference time period; sorting the time-related confidence scores; and before the future specific time, adding cached data with higher time-related confidence scores and data calculated from at least one other caching algorithm to the cache for use by the cache at the future specific time, wherein a fixed ratio exists between the cached data with higher time-related confidence scores and the data calculated from other caching algorithms, wherein the recording time is the time when the cached data is accessed, and the reference time period is a specific sub-time unit selected in the main time unit; and Multiple auxiliary memory modules, connected to the host, are used to distribute and store data for access. The time-related confidence level is calculated using the following steps: C1. Calculate the first quantity, which is the number of times the reference time period occurred in the past period. C2. Calculate the second quantity, wherein the second quantity is the quantity of the reference time period when the target cache data is accessed; and C3. Divide the second quantity by the first quantity. 15. The cloud storage device system as described in claim 14, wherein the caching algorithm includes LRU algorithm, MRU algorithm, PLRU algorithm, RR algorithm, SLRU algorithm, 2-way set associative algorithm, LFU algorithm, LIRS algorithm, ARC algorithm, CAR algorithm, MQ algorithm, or a data-related algorithm that uses data generated by the content recorder as target data. 16. The cloud storage device system as described in claim 14, wherein the data type includes objects, blocks, or files.