JP2839538B2 - Disk cache simulation processing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Overview] A disk cache simulation processing method for simulating a disk cache and collecting cache characteristic data by using a main memory of the present system processing apparatus. In order to provide means for simulating the operation of the disk cache in a real environment, and to enable the acquisition of cache characteristic data such as the hit ratio, one track corresponds to one entry in the main memory. Based on the mapping table provided and the information on the access position when executing input / output
The position of the corresponding entry in the mapping table is calculated, and based on the information of the entry, an information collection unit that collects information on a cache memory hit or miss and an information collected by the information collection unit at the end of the simulation are analyzed. And an analysis output processing unit for outputting cache characteristic data.

[Industrial applications]

The present invention relates to a disk cache simulation processing method that simulates a disk cache and collects cache characteristic data by using a main memory of a main unit.

A disk cache provided with a cache memory in a disk control device such as a magnetic disk or an optical disk is used. In a disk cache, access data is staged in a cache memory, and for hit data, data is transferred from the cache memory to speed up input / output. The effect of performance improvement by introducing such a disk cache greatly affects the size of the cache memory and how to set the extent attribute, so that it is necessary to collect data for tuning.

[Conventional technology]

 FIG. 5 shows an example of the conventional system.

Conventionally, as a method of collecting cache characteristic data relating to a disk cache, a method of collecting data by a microprogram or the like in the disk controller 51 as shown in FIG. As described above, a method of obtaining cache characteristic data by software using a trace data analysis program 61 for analyzing input / output trace data 60 is used.

In the case of the method shown in FIG. 5A, when there is access from the main processing unit 50 to the disk storage device 57, the statistical information such as a hit or a mistake is collected in the disk control device 51 by the statistical information collection unit 56. Collect by

In order to manage the cache memory 55, the disk controller 51 has three management tables: an extent attribute table 52, a cache entry table 53, and an LRU management table 54. Extent attribute table
Reference numeral 52 denotes a table for storing extent attributes, and a cache entry table 53 indicates whether or not a copy of a record of a specified track address (CCHH) exists in the cache memory 55, and if so, the physical address thereof. It is a table to be stored. The LRU management table 54 is a table for selecting a copy of an extent to be evicted from the cache memory 55 when staging the accessed data by using an LRU (Least Recently Used) method.

In the method shown in FIG. 5B, during the execution of the online service 58, information relating to the access position is collected as trace data 60 by the I / O trace 59. Normally, the online service 58
After the completion of the process, the trace data analysis program 61 is operated to calculate a hit ratio when the disk cache is used. Then, the result is output to the printer 62.

[Problems to be solved by the invention]

In the method shown in FIG. 5A, the characteristic data of the cache memory can be obtained without the overhead on the main processing unit 50 side. However, this method requires that a disk controller 51 having a disk cache mechanism be introduced before simulation can be performed. Before the disk cache is introduced, it is necessary to determine how much the processing capacity is improved after the introduction. Can't estimate.

In the case of the method shown in FIG. 5 (b), it is possible to estimate a certain effect after the introduction even before the introduction of the disk cache, but the configuration is such that the data collection and the analysis processing are started completely separately. Therefore, there is a problem that the operability is poor and that it takes time from obtaining the trace data to obtaining the analysis result.

The present invention solves the above problems, and provides means for simulating the operation of a disk cache under a real environment before introducing the disk cache, thereby enabling acquisition of cache characteristic data such as a hit ratio. It is an object.

[Means for solving the problem]

 FIG. 1 shows a configuration example of the present invention.

In FIG. 1, reference numeral 10 denotes a processing unit including a CPU and a memory, 11 denotes an input / output request unit for requesting input / output, 12 denotes an input / output control unit in the operating system, and 13 denotes a channel command word (CCW). CCW trace part, 1
4 is an information collecting unit for collecting simulation information of the disk cache, 15 is a start / end processing unit for controlling start and end of simulation, 16 is an analysis output processing unit for analyzing simulation information and outputting the analysis result, 17 Is a main memory, 18 is a mapping table, 19 is a statistical information storage area for storing statistical information such as the number of hits for each extent, 20 is an operator console, 21 is a disk control device, 22 is a disk storage device such as a magnetic disk or an optical disk. , 23 represent a printer.

In the present invention, a mapping table 18 for simulating a disk cache operation is provided in the main memory 17. The mapping table 18 has an area in which one track corresponds to one entry for all tracks of the volume to be cached, and at least information about the extent attribute setting status, staging status, and LRU management of the cache memory is set in each entry. It is supposed to be.

When the start / end processing unit 15 is started by a simulation start command, the start / end processing unit 15 initializes the mapping table 18 and the statistical information storage area 19. Then, the I / O controller 12 is notified of the start of the disk cache simulation.

Thereafter, when there is an input / output request from the input / output request unit 11, the input / output control unit 12 activates the CCW tracing unit 13. The CCW tracing unit 13 analyzes the channel command word (CCW) in the channel program and notifies the information collecting unit 14 of the access path information and the record address (CCHHR) information.

The information collecting unit 14 calculates the position of the corresponding entry in the mapping table 18 from the access path information and the record address information, and based on the information of the entry, stores information on hits or misses in the cache memory in the statistical information storage area 19. Is performed.

The analysis output processing unit 16 analyzes the information collected in the statistical information storage area 19 by the information collection unit 14 at the end of the simulation, and outputs cache characteristic data to the printer 23.

[Action]

In order to simulate the same operation as the cache operation performed by the disk controller, an extent attribute table storing extent attributes, a cache entry table storing cache memory usage status, and order information of accessed data are managed. An LRU management table is required.

In the present invention, these three tables are realized by one of the mapping tables 18 and are handled in the main memory 17, so that the simulation processing can be sped up. As a result, the overhead for the simulation by the processing device 10 is reduced, and the simulation of the cache operation in a form close to that when the disk cache is actually introduced can be performed.

Further, an external storage device for logging trace data is not required, and the information collecting unit 14 and the analysis output processing unit 16 can immediately acquire cache characteristic data for tuning the disk cache.

As described above, the present embodiment differs from the conventional method in that cache characteristic data can be acquired without introducing a disk cache, and high-accuracy cache characteristic data can be easily acquired by a series of operations during online service.

〔Example〕

FIG. 2 is a configuration example of a mapping table according to the embodiment of the present invention, FIG. 3 is a table configuration example of a statistical information storage area according to the embodiment of the present invention, and FIG. 4 is a simulation of a simulation according to the embodiment of the present invention. 4 shows a processing flow.

The mapping table 18 shown in FIG. 1 is configured as shown in FIG. 2 in this embodiment. For all tracks of each volume to be cached, areas are allocated so that one track corresponds to a 4-byte entry.

In FIG. 2A, the table configuration is expressed two-dimensionally, but the track address is determined by the cylinder number and the track (head) number, whereby the address of each entry can be obtained. ing.

In the content of each entry, as shown in FIG. 2 (b), the first three bits are a flag indicating the setting status and the staging status of the extent attribute. 4th bit to 4th
The record number up to the 11th bit is the record number, and indicates from which record of the track the staging range is present. The twelfth and subsequent bits are an area for the LRU stack pointer.

In the present embodiment, a standard extent or a priority extent can be selected as the extent attribute. When the bit b0 is “1”, the standard extent attribute is set. When the bit b1 is “1”, the priority extent attribute is set. Indicates that If both bits b0 and b1 are "0",
The track is not subject to staging to the cache memory.

The standard extent attribute and the priority extent attribute are
This attribute is used when determining a track to be evicted from the cache memory when staging is required for a new track. Standard extents are subject to eviction before priority extents. However, the minimum guaranteed amount is determined for the number of staging standard extents, and when the minimum guaranteed amount is reached, even if the standard extent remains, the priority extent is subject to eviction.

Among the same extent attributes, the currently staged track that has not been accessed for the longest time by the LRU is the target of eviction from the cache memory. For the management, as shown in FIG. 2 (a), the priority LRU stack pointer 30 and
A queue terminal for the standard LRU stack pointer 31 is provided, and a queue for each entry is created in order from the most recently accessed one.

In this example, among the tracks having the attribute of the priority extent, the track that has been accessed most recently is the track T0, the next track is the track T1,..., And the track that has been accessed the oldest is the track Tn. Similarly, for the standard extent, a queue is created from the standard LRU stack pointer 31 from the track T11 to the track Tm.

Although not shown, a pointer to the last connected track Tn, Tm of the queue is also managed for each attribute in order to speed up the queuing process. In addition, in order to facilitate the process of flushing tracks from the cache memory, when initializing the mapping table 18, queues are created in advance for the number of tracks that can be staged.

The data collected by the disk cache simulation is the following data for each extent. The extent is a continuous area allocated to the file.

• Number of reads • Number of writes • Number of read hits, read miss circuits • Number of write hits, write misses In addition, data such as the amount of staging data, the amount of read data, and the amount of read hit data is collected. I do.

To calculate the optimal amount of cache memory, the LRU
The number of hits is counted for each queue depth.

For such information collection, a table as shown in FIG. 3 is used.

The table 40 shown in FIG. 3A is a table for collecting statistical information for each extent. For each extent, the file name, the I / O device number indicating the location of the volume, the volume serial number (VSN), and the range of the area in the volume (CC means cylinder number, HH means track number) Definition information is set in advance, and statistical information such as the number of hits is stored for each of these extents.

Tables 41 and 42 shown in FIG. 3B are tables for storing the number of hits according to the LRU depth for each extent attribute. For example, when the third track connected to the queue of the standard LRU is accessed, the value of the third position from the top in the table 41 is incremented by one.

When the table 41 is analyzed after the end of the simulation, for example, when the number of hits for the i-th and subsequent hits is 0, it is meaningless to define i or more as attributes of the standard extent. The same applies to the LRU of the priority extent in the table 42, and by analyzing these, it is possible to avoid preparing an unnecessarily large cache memory.

FIG. 4 shows the flow of simulation and information collection processing centered on the information collection unit 14 shown in FIG. Hereinafter, the processes (a) to (h) shown in FIG.
It will be described according to.

(A) Access path information such as an input / output device number indicating the location of a volume, and record address information (CC
When the HHR is notified, the position of the corresponding entry in the mapping table 18 is calculated based on the track address (CCHH) included in the record address information. In the record address, CC is the cylinder number, HH is the track (head) number, and R is the record number in the track. Also,
In the statistical information storage area 19, general extent-specific statistical information such as the number of times of reading and the number of times of writing is collected in accordance with the type of access.

(B) Check the corresponding entry in the mapping table 18, and
It is determined whether the attribute of the standard extent or the priority extent is set. If the extent attribute is not set, it will not be cached.
Return control to the I / O controller.

(C) If the extent attribute is set, the staging status of the entry is checked, and a hit or a miss is determined.

(D) If a hit occurs, the entry is positioned at the top of the LRU stack according to the attributes of the standard or priority extent, and the most recently accessed entry is displayed.

(E) Collect data on the extent, such as the number of hits. In addition, the number of hits for each depth is counted up in the table shown in FIG. 3B according to the entry's original position in the LRU queue. Then, control is returned to the input / output control unit.

(F) If the corresponding entry has not been staged, it is determined whether the staging of the standard extent attribute or the priority extent attribute that has already been staged is evicted from the cache memory for the staging. The priority extent attribute is selected only when the number of items staged with the standard extent attribute is equal to or less than a certain threshold, and otherwise, the standard extent attribute is selected.

(G) In the LRU stack of the attribute determined in the process (f), the track of the lowest entry in the staging state is evicted from the cache memory. In this eviction processing, the reconnection of the LRU queue and the setting of the staging state flag to "0" are performed.

(H) After setting the access track in the staging state and positioning it at the top of the LRU stack, data such as the number of misses relating to the extent is collected. afterwards,
Return control to the I / O controller and move on to access to the actual drive. The statistical information on the number of misses may be collected here, but since the number of misses is obtained from the difference between the number of accesses and the number of hits, the data need not be collected.

Through the processing described above, statistical information by simulation is collected. The analysis output processing unit 16 shown in FIG.
Analyzes the statistical information, calculates a hit ratio for each extent, and outputs the result to the printer 23 in the form of a graph or a table. Since this processing can be easily realized if necessary data is collected, detailed description of the embodiment will be omitted. In particular, the effect of the introduction of the disk cache greatly depends on the read hit ratio, so the cache memory size is determined based on the analysis results and the extent attributes are set so that the read hit ratio increases. Should be performed.

[Effects of the Invention] As described above, according to the present invention, it is possible to directly operate a disk cache simulation under a real environment such as an online service, and immediately obtain cache characteristic data. Even when there is no disk control device having a disk cache function, accurate cache characteristic data such as a hit rate can be obtained in the same manner as when a disk cache is actually introduced.

[Brief description of the drawings]

1 is a configuration example of the present invention, FIG. 2 is a configuration example of a mapping table according to the embodiment of the present invention, FIG. 3 is a table configuration example of a statistical information storage area according to the embodiment of the present invention, FIG. FIG. 5 shows a flow of a simulation process according to an embodiment of the present invention, and FIG. 5 shows an example of a conventional system. In the figure, 10 is a processing device, 11 is an input / output request unit, 12 is an input / output control unit, 13 is a CCW tracing unit, 14 is an information collection unit, 15 is a start / end processing unit, 16 is an analysis output processing unit, 17 Is a main memory, 18 is a mapping table, 19 is a statistical information storage area, 20 is an operator console, 21 is a disk control device, 22 is a disk storage device, 23
Represents a printer.

────────────────────────────────────────────────── (5) Continuation of the front page (51) Int.Cl. ⁶ Identification symbol FI G06F 12/08 G06F 12/08 E 12/12 12/12 A (56) References JP-A-58-144961 (JP, A) JP-A-58-29061 (JP, A) JP-A-61-62950 (JP, A) "28th (Early 1984) National Convention Lecture Paper Collection (I)", Information Processing Society of Japan March 13, 1984, p. 67−68 (58) Fields surveyed (Int.Cl. ⁶ , DB name) G06F 12/08-12/12 G06F 11/28-11/34 G06F 3/06

Claims (1)

(57) [Claims] In a disk cache simulation processing method in a data processing system for collecting cache characteristic data by simulating a disk cache, all tracks of a volume to be cached are stored in a main storage in which one track corresponds to one entry. A mapping table (18) that has an area in which at least information about the extent attribute setting status, staging status, and cache memory LRU management is set in each entry, The information collection unit (14) calculates the position of the corresponding entry in the mapping table from the information of the access position, and collects information on the hit or miss of the cache memory based on the information of the entry. If, during simulation ends, to analyze the information collected by the information collection unit, the disk cache simulation processing method, characterized in that an analysis output unit for outputting the cache characteristic data (16).