JPS6125176B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides a virtual computer system, especially when a plurality of virtual computers operate on one real computer.
Regarding the above-mentioned virtual computers, the present invention relates to a virtual computer system in which a plurality of high-performance virtual computers are realized by matching the real address space of the virtual computer created by a virtual computer control program (CP) with the actual real address space.

Generally, in a virtual computer system, a plurality of virtual computers are converted into one real computer and processed. An operating system (OS) runs on a virtual computer as if it were occupying a real computer. The operating system converts and processes the virtual address into a pseudo real address of the virtual machine, but this real address must be further translated into an actual real address on the real computer. Hereinafter, the actual real address space will be referred to as the primary address space, the pseudo real address space on the virtual computer will be referred to as the secondary address space, and the virtual address space created by the operating system running on the virtual computer will be referred to as the tertiary address space.

Conventionally, when mapping a tertiary address space to a primary address space, the following address conversion has been performed. FIG. 1 shows an example of a conventional address translation method. In the figure, 1 is the tertiary address space,
2 is the secondary address space, 3 is the primary address space,
4 is an address translation table created by the operating system running on the virtual machine, 5 is an address translation table created by the virtual machine control program (CP), and 6 is a tertiary address space created by the virtual machine control program. An address translation table that maps to the next address space,
7 represents the prefix area of the virtual machine. The operating system running on the virtual machine creates an address translation table 4 for mapping the tertiary address space 1 to the secondary address space 2. On the other hand, the virtual computer control program is 2
An address translation table 5 for mapping the next address space 2 to the primary address space 3 is created, and the main memory of the virtual machine is virtualized using a paging technique. When a program on a virtual machine operates under dynamic address translation, paging must be performed in two stages, so the virtual machine control program newly creates address translation table 4 and address translation table 5. are merged to create an address translation table 6 that directly maps the tertiary address space 1 to the primary address space 3. Therefore, the overhead of the virtual machine control program required for creating and changing the address conversion table 6 is large. Also, operating system CCW
(Channel Command Word) cannot be executed as is, so CCW translation processing is required for address conversion, and the overhead of the virtual machine control program is also large.

In order to solve the above-mentioned drawbacks, the following partially improved virtual computer system was considered. Second
The figure shows an example of an improved conventional address translation scheme. In the figure, numerals 1 to 5 and 7 correspond to those in FIG. 1, and 8 represents the prefix area of the actual computer. With the exception of the prefix area 7, the real main memory is provided as it is as the main memory of the virtual machine. That is, except for page 0, secondary address spaces 2 and 1
Match the address with the next address space 3. As long as you change only the entry on the 0th page of the address translation table 4 to point as shown in Figure 2, when a program on a virtual machine runs with dynamic address translation on, this address translation table 4 can be used as is. Therefore, the address conversion table 6 shown in FIG. 1 becomes unnecessary. Therefore, the overhead of the virtual computer control program required to operate the address translation table 6 as explained in FIG. 1 can be reduced, and since the CCW prepared by the operating system can be executed as is, Control program overhead can also be reduced.

However, in this method, in a high-performance virtual machine such as the one described above, the secondary address space is a space with a series of addresses, and this space must be allocated from the beginning of the primary address space. Only one can be achieved,
When a plurality of virtual machines exist, the other virtual machines require the address translation table 6 shown in FIG. 1, resulting in overhead. The increase in overhead in virtual computer systems is a major concern, and there has been a strong desire to reduce it.

An object of the present invention is to provide a plurality of high-performance virtual machines on one virtual machine system. Therefore, the virtual computer system of the present invention is a system that operates a plurality of virtual computers, and has a primary address space that is a real address space of the implemented main memory, and a virtual computer that operates on a virtual computer corresponding to each virtual computer. Virtual machine address control is performed by mapping a secondary address space, which appears as a real address space from the operating system (OS), and a tertiary address space, which is the virtual address space of the OS, in multiple stages. In the virtual computer system, the above-mentioned contiguous area has a size that is the sum of the system resident area allocated from address 0 to address l in the secondary address space and the paging target area provided after address l. means for extracting from address n in the primary address space; and an address conversion table for mapping the tertiary address space to the secondary address space, converting the area from address 0 to l in the secondary address space into the primary address space. The means corresponds to addresses from n to (n+l) in the above secondary address space, and the area to be paged in the secondary address space is made to correspond to addresses after (n+l) in the primary address space, and becomes the paging target. means for matching an address in the secondary address space of a region with a corresponding address in the primary address space, and mapping from the tertiary address space to the primary address space; It is characterized in that it is configured to perform the mapping using an address conversion table that maps to the secondary address space. This will be explained below with reference to the drawings.

FIG. 3 shows the configuration of an embodiment of the present invention. In the figure,
Symbols 1 to 4, 7, and 8 correspond to FIGS. 1 and 2, and 9 represents an operating system in which the virtual address of the operating system running on the virtual machine matches the address on the secondary address space 2. Resident areas 10, 10', and 10'' each represent a virtual computer, 11 a real computer, 12 a domain, 13 an unimplemented memory space on the virtual computer 10, and 14 a paging area of the operating system.

In FIG. 3, from an arbitrary address, for example, address n, in the primary address space 3 consisting of the main memory implemented, a continuous An area 12 (hereinafter referred to as "domain") is cut out and reserved for the virtual machine 10. On the other hand, the secondary address space 2 of the virtual machine 10 is divided into an area 9 where the operating system resides and is controlled so that the virtual address and the real address on the virtual machine match, and an area for paging of the operating system. If the former area 9 is from address 0 to l address, for example, then the latter area 14 is assigned an address starting from address (n+l) on the secondary address space 2. That is,
From the operating system's perspective, the area 13 from address l to address (n+l) in the secondary address space 2 is handled as if no memory was mounted therein. Of course, the resident area 9 of the operating system also includes the prefix area 7. This area 9 is an area that is not subject to paging even if the operating system running on the virtual machine 10 employs, for example, a multiple virtual storage system. Therefore, once the entries from address 0 to l in the address translation table 4 created by the operating system are updated so that they start from address n and end at address (n+l), address to l address is address conversion table 4
As a result, the address n, which is the beginning of the domain 12 extracted from the primary address space 3, is mapped to the address (n+l), and there is no need to change the address translation table 4 thereafter. Furthermore, the area 14 that is the target of paging by the operating system starting from address (n+l) in secondary address space 2 is directly associated with the area starting from address (n+l) in domain 12 of primary address space 3. Assigned. Therefore, paging area 1 of this operating system
Regarding 4, the address in secondary address space 2 and the address in primary address space 3 completely match, and the mapping from tertiary address space 1 to primary address space 3 is based on address conversion table 4. You will be able to do it directly. From the above, when a program runs on the virtual machine 10 with dynamic address translation turned on, the address translation table 4 created by the operating system can be used as is, and the tertiary address space 1 to the primary address space There is no need to create and use a new address translation table for mapping to 3. In addition, for input/output requests sent from the virtual machine 10, the real address on the virtual machine and the real address on the real computer match, so there is no need for translation processing for CCW address conversion, and the request can be executed as is. be able to. Since the domain 12 can be carved out from any address on the primary address space 3, multiple domains can be prepared in one system corresponding to each virtual machine, and multiple high-performance virtual machines can be operated simultaneously. , can be made to work.

As explained above, according to the present invention, a plurality of high-performance virtual machines can be realized with one virtual machine system. In fact, while the overhead due to the virtual machine control program of a conventional virtual machine was about 50%, the present invention has made it possible to keep the overhead to about 30% for any virtual machine.

[Brief explanation of the drawing]

FIGS. 1 and 2 are explanatory diagrams of examples of conventional address translation methods, and FIG. 3 shows the configuration of an embodiment of the present invention. In the figure, 1 is a tertiary address space, 2 is a secondary address space, 3 is a primary address space, 4 to 6 are address translation tables, 7 and 8 are prefix areas, 9 is an operating system resident area,
10, 10', and 10'' are virtual computers, 11 is a real computer, 12 is a domain, 13 is an unimplemented space, and 14 is a paging area.

Claims (1)

[Scope of Claims] 1. A system that operates a plurality of virtual machines, including a primary address space that is a real address space of the implemented main memory, and an operating system that runs on the virtual machine for each virtual machine. A virtual computer that performs address control of the virtual computer by mapping in multiple stages a secondary address space that appears to be the real address space from the system (OS) and a tertiary address space that is the virtual address space of the OS. In the system, a contiguous area with a size that is at least the sum of the system resident area allocated from address 0 to l address in the secondary address space and the paging target area provided after address l is located at the primary address address above. means for extracting from address n in the space; and an address conversion table for mapping the tertiary address space to the secondary address space, converting the area from address 0 to l in the secondary address space to address n in the primary address space. to address (n+l); and means for associating the paging target area in the secondary address space with the address (n+l) and subsequent addresses in the primary address space, and the above two of the paging target area. and means for matching an address in the secondary address space with a corresponding address in the primary address space, and the mapping from the tertiary address space to the primary address space is performed by A virtual computer system characterized in that it is configured to perform mapping using an address conversion table.