JPS63293653A - Memory multiplexing system - Google Patents

Abstract

PURPOSE:To reduce the cost for a memory multiplexing system by reading an auxiliary memory page if having the same contents as those of a real page into a real memory to continue the execution of an instruction of the real page when an error occurs while the instruction of the real page is executed. CONSTITUTION:A page table 240, an external page table 250 and a replacement bit 260 are used in addition to a real memory 210 and an auxiliary memory 220 in order to obtain a virtual memory 230 to which the logical reference is possible from a CPU 200. The table 240 shows the corresponding relation between both memories 210 and 220. The bit 260 shows the replacement of the real pages on the memory 210. Then, when a detector included in the CPU 200 detects the generation of a memory access fault at the time of executing the reference, the replacement and the instruction to the real pages, the pages of the memory 220 are read into the memory 210 in terms of the multiplexed pages with reference to both tables 240 and 250 and the bit 260.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a computer system having a virtual storage device,
The present invention relates to a method for duplicating not only general pageable areas but also non-pageable areas including a common area particularly suitable for improving the reliability of storage devices on an inexpensive auxiliary storage device.

[Conventional technology]

In conventional computer systems, in order to guarantee the reliability of storage devices, failures are detected and 1-bit errors are automatically corrected, as described in Japanese Patent Laid-Open No. 60-100253.
Some 2-bit errors are reported to the CPU via an interrupt. When a 2-bit error is detected, an attempt is made to recover from memory backup.

When there is a failure in the basic part of the system program, there is a possibility of system failure. For this reason, JP-A-60-100253
As described in the above issue, there is a method in which a plurality of combinations of CPUs and storage devices are prepared, and when an unrecoverable storage failure occurs, the CPU and storage device combinations are separated from the system and switched to another system.

[Problem that the invention seeks to solve]

In the above-mentioned conventional technology, real storage devices are multiplexed with real storage devices having the same configuration. Generally, real storage devices are more expensive than auxiliary storage devices, and having multiple storage devices is disadvantageous from an economic standpoint. On the other hand, with virtual storage, pages are held in auxiliary storage in addition to real storage, so if the real storage becomes unusable due to a failure, the pages in auxiliary storage can be read and used. be able to.

However, even in this method, since the non-pageable area is not targeted for paging, there is no data on the auxiliary storage device, and there is a problem in that the data is not duplicated.

[Means for Solving the Problems] In order to solve the above-mentioned problems, the present invention provides a means for duplication using an economical auxiliary storage device to improve reliability. In this case, a means is required to duplicate the non-pageable area containing important parts of the system program that could lead to system failure.

This periodically pages out unpageable areas and
This is achieved by duplicating it on an auxiliary storage device.

[Effect]

Even if data exists on auxiliary storage, if the corresponding page has been updated on main storage.

Data on the auxiliary storage device cannot be used as duplex data. In order to always have duplicated data, it is necessary to update the corresponding page on the auxiliary storage every time an update occurs on the main storage. However, updating the auxiliary storage device involves input/output operations, resulting in a significant drop in processing efficiency.Furthermore, the non-pageable area, which was not traditionally subject to paging, is also duplicated on the auxiliary storage device, which can further reduce efficiency. .

Therefore, the present invention adopts a method of periodically executing page-out processing without updating the auxiliary storage page every time the real page is updated. Although complete duplication is not achieved by this, a significant drop in efficiency can also be prevented. Since these two factors work in opposition to each other, it was decided that the user could select and specify the trade-off between increasing the degree of duplication to improve reliability and decreasing processing efficiency.

〔Example〕

An embodiment of the present invention will be described with reference to FIG. 1 and subsequent figures.

In FIG. 1, 100 is a central processing unit W1 (CPU), 101 is a memory fault detection device, 110 is a real storage device, and 120 is an auxiliary storage device. When a memory access request occurs while the CPU is operating by referring to, updating, and executing commands on the real storage device, a memory fault detection device within the CPU detects this and generates an interrupt (machine check interrupt). Processing for recovering from a memory failure by using duplicated pages on the auxiliary storage device is performed as an extension of this machine check interrupt.

FIG. 2 shows an overview of a virtual storage system using paging. In order to realize a virtual storage device 230 that can be logically referenced from the CPU 200, a real storage device rPt 210 that physically stores data, an auxiliary storage device 220, and a page table 240. It consists of an external page table 250 and an update bit 260 indicating that a real page on the real storage device has been updated. From information on these page table entries, external page table entries, and update bits, it is possible to know on which storage device the contents of the virtual page at that time are held. In the example in Figure 1, pages 0, 5, 6, and n are held on both the real storage device and the auxiliary storage device, and the update bit is off, so the real page is the same as the auxiliary storage page. Yes, it is duplicated. The fourth page holds data both on the real storage device and on the auxiliary storage device, but
Since the update bit is on, they do not match, and the page is not duplicated.Also, the first page is currently in use on the real storage device, but there is no data on the auxiliary storage device, so this It is also not duplicated. Since the second and third pages are not on the real storage device, memory failure cannot occur.In this way, the page table, external page table,
The update bit can identify whether a page is duplicated on the auxiliary storage device.

FIG. 3 shows the processing performed by the machine check interrupt handler when a memory failure occurs. Determine the above external page table entry and update bit (steps 301 and 30)
2) If the page is not duplicated on the auxiliary storage device, error recovery is impossible and a memory access error occurs. If the pages are duplicated in the auxiliary storage device, the faulty real page is separated from the virtual storage device and disabled for use, and then the instruction that caused the memory fault is re-executed. In this re-execution, a new real page is allocated from the virtual page due to a page fault, and instruction execution continues.

In the present invention, since duplication by the auxiliary storage device is performed also for the non-pageable area, page-out is performed periodically including the non-pageable area. FIG. 4 shows the procedure of this page-out processing. Page out is necessary when there is a real page but no virtual memory page, and when there is a real page and a virtual memory page but the update bit is on and they do not match (steps 403, 404). In this case, conventionally, page-out processing is not performed on non-pageable areas, and the actual page is made usable after page-out of pageable areas. It is necessary to process a page to be paged out and to not make the cusp page usable even if the page is paged out.

〔Effect of the invention〕

According to the present invention, duplication of storage devices can be realized at low cost. At this time, since the auxiliary storage device is used as the destination for duplication and is used using the paging function, there is no need to newly provide a storage device exclusively for duplication, which is efficient.

[Brief explanation of drawings]

Fig. 1 is an explanatory diagram of the positioning of the auxiliary storage device and the memory failure detection device, which are duplexed media in the present invention, Fig. 2 is a conceptual diagram of the virtual storage device and an explanatory diagram of duplication using the auxiliary storage device, and Fig. 3 is the memory FIG. 4 is a flow chart of the process of the machine check interrupt handler due to a failure. FIG. 4 is a flow chart of the process of duplicating the data into auxiliary storage. 100...Central processing unit, 101...Memory fault detection device, 110...Real storage device, 120...Auxiliary storage device, 200...Central processing unit, 210...Real storage device. 211.221... Duplicated page by auxiliary storage device, 220... Auxiliary storage device, 230... Virtual storage device, 240... Page table entry, 25
0: External page table entry, 260: Update bit.

Claims (1)

[Claims] 1. Consisting of a real storage device that can be directly referenced by a central processing unit (CPU) and an auxiliary storage device that cannot be directly referenced,
The real storage device and auxiliary storage device are divided into fixed-length areas (pages), the CPU executes on the real storage device, and when the real storage device is insufficient, the pages (real pages) of the real storage device in use can be used. In order to do this, in a computer with a virtual memory storage system, the updated real page is written to the auxiliary storage device and saved before it becomes a usable real page. When an error occurs during update/command execution, if there is a page on the auxiliary storage device (auxiliary storage page) with the same content as the real page, read it onto the real storage device and execute this page. A memory multiplexing method characterized by continuing computer operation by referencing, updating, and executing instructions on real pages. 2. The memory multiplexing method according to claim 1, which periodically pages out updated pages among non-pageable pages on the real storage device that do not have auxiliary storage pages. .