WO2019086786A1

WO2019086786A1 - Method of communication between processes, corresponding computer program and computing installation

Info

Publication number: WO2019086786A1
Application number: PCT/FR2018/052657
Authority: WO
Inventors: Zoltan Menyhart
Original assignee: Bull Sas
Priority date: 2017-10-30
Filing date: 2018-10-26
Publication date: 2019-05-09
Also published as: FR3073061B1; EP3704582A1; FR3073061A1

Abstract

This method of communication between first and second processes currently under execution, in a computing installation comprising a first memory (M1) allocated to the first process for its execution, comprises the following steps: a file system (SF1, SF2) of the computing installation (100) creates a file (F1); the file system (SF1, SF2) receives a reading request in respect of data in the file (F1); in response to the reading request, the file system (SF1, SF2) retrieves the data so that the second process can access same. The file (F1) is a virtual file associated with a part (ZM1) of the first memory (M1) and the file system (SF1, SF2) retrieves the data from the part (ZM1) of the first memory (M1).

Description

PROCESS COMMUNICATION METHOD, COMPUTER PROGRAM AND CORRESPONDING COMPUTER INSTALLATION

The present invention relates to a method of communication between processes, a computer program and a corresponding computer installation.

The invention applies more particularly to a method of communication between first and second processes running in a computer installation comprising a first memory allocated to the first process for its execution, the method comprising the following steps:

a file system of the computer installation creates a file,

the file system receives a request to read data from the file, and

- In response to the read request, the file system retrieves the data for the second process to access.

The file is stored in a mass memory and mapped both in the memory allocated to the first process and in a memory allocated to the second process. Thus, when the first process wishes to communicate data to the second process, the first process writes to its mapped file in memory, which is reflected on the file stored in the mass memory. The second process then attempts to read from its memory-mapped file, which triggers, if the data has not yet been copied to the memory-mapped file, a page fault detected by a memory management unit. The latter then requests the file system to read from the file the missing page, which is copied to the file mapped in memory of the second process, for the latter to access.

The previous method has the disadvantage of requiring a mass memory to store the file.

It may thus be desirable to provide a method that makes it possible to overcome at least some of the aforementioned problems and constraints.

The subject of the invention is therefore a method of communication between first and second processes running in a computer installation comprising a first memory allocated to the first process for its execution, the method comprising the following steps:

a file system of the computer installation creates a file, the file system receives a request to read data from the file,

in response to the read request, the file system retrieves the data for the second process to access it,

the method being characterized in that the file is a virtual file associated with a portion of the first memory, and in that the file system retrieves the data from the portion of the first memory.

Thus, since the file is virtual and associated with a portion of the memory allocated to the first process, this file does not require mass memory.

Optionally, the computer installation comprises at least first and second computers, on which the first and second processes respectively execute, and the file system is a distributed file system so that the file is accessible from each of the first and second computers.

Optionally also, a second memory of the computer installation is allocated to the second process for its execution, the method further comprising the following steps:

the file is mapped in the second memory so as to obtain a file mapped in memory,

the second process attempts to access data from the file mapped in memory, these data having not yet been copied to the file mapped in memory, the step of reception by the file system of the read request being carried out in response to this access attempt,

- After the data recovery step by the file system, the recovered data is copied to the mapped file in memory for the second process to access.

Optionally also, the file is associated with permissions, the method further comprising, before the step of recovering data by the file system:

- checking that the permissions associated with the file allow the second process to read it.

Also optionally, the method further includes creating a directory path after which the file is placed. Optionally also, the first computer belongs to a computer group of the computer installation and one of the directories of the path corresponds to this group.

The invention also relates to a computer program downloadable from a communication network and / or recorded on a computer readable medium and / or executable by a processor, characterized in that it comprises instructions for the execution of steps of a method according to the invention, when said program is executed on a computer.

The invention also relates to a computer installation in which first and second processes are intended to execute, comprising:

a first memory intended to be allocated to the first process for its execution,

a file system designed to create a file, to receive a request to read data in the file, and, in response to the read request, to retrieve the data so that the second process can access it,

the computer installation being characterized in that the file is a virtual file associated with a portion of the first memory, and in that the file system is adapted to recover the data from the portion of the first memory.

The invention will be better understood with the aid of the description which follows, given solely by way of example and with reference to the appended drawings in which:

FIG. 1 diagrammatically represents the general structure of a computer installation, according to one embodiment of the invention, FIG. 2 illustrates the successive steps of a process of communication between processes, according to an embodiment of the invention. invention, and

FIG. 3 is an example of a tree of a file system used in the method of FIG. 2.

With reference to FIG. 1, a computer installation 100 embodying the invention will now be described.

The electrical installation 100 comprises several computers, including two computers 102 _1; 102 ₂ . The electrical installation 100 further comprises a network 104 by which the computers, and in particular the two computers 102 _1; 102 ₂ , are connected to each other. Each computer 102 _1; 102 ₂ comprises a network interface 104 _1; 104 ₂ by which the first computer 102 _1; respectively the second computer 102 ₂ , is connected to the network 104.

In addition, each computer 102 _1; 102 ₂ comprises a central processing unit 106 _1; 106 ₂ and a main memory 108 _1; 108 ₂ . The central processing unit 106 _1; 106 ₂ comprises for example a processor and the main memory 108 _1; 108 ₂ comprises for example a volatile memory, for example RAM type (of the English "Random Access Memory").

In addition, each computer 102 _1; 102 ₂ comprises a mass storage unit 1 1 C, 1 10 ₂ non-volatile, such as a hard disk.

In general, the central processing unit 106 _1; 106 ₂ of each computer 102 _1; 102 ₂ is for executing computer program instructions. Each running computer program, called process, is allocated a memory for its execution. This memory extends at least on the main memory 108i, 108 ₂ and consists of virtual addresses, that is to say do not correspond to physical addresses. This makes it possible in particular to use disjoint pieces of main memory 108i, 108 ₂ , or even pieces of mass memory (such as storage unit 1 10i, 1 10 ₂ ), while the process thinks it has access to a block continuous main memory. Generally, virtual memory is made up of areas of the same size, called pages.

Each computer 102 _1; 102 _{2 further} comprises a memory management unit 1 12 _1; 1 12 ₂ designed in particular to convert the virtual addresses that processes try to access into physical addresses.

In the example described, the computers 102 _1; 102 ₂ each execute an operating system SE1, SE2. They also run a file system SF1, SF2 managing a tree A directories and files from a root directory ("root" in English). In general, a file system is a collection of mechanisms and rules for storing and retrieving data. In the example described, the file system SF1, SF2 is a distributed file system comprising a part executed on each computer of the computer installation 100, and in particular a part SF1 executed by the computer 102 _1; and a portion SF2 executed by the computer 102 ₂ . The file system SF1, SF2 is designed to implement synchronization mechanisms between its parts SF1, SF2 through the network 104, in order to present the same tree A on each computer 102 _1; 102 ₂ . In the example described, the file system SF1, SF2 is shown as distinct from the operating systems SE1, SE2. However, the file system may also be entangled in the SE1, SE2 operating systems, so that it can be difficult to distinguish them. In this case, the operating systems SE1, SE2 themselves can be considered as forming the file system SF1, SF2. In addition, the tree A can be placed in another file system tree (such as a "real" file system, ie files stored in a non-volatile mass memory, such as the hard disks 1 10 ₁ , 1 10 ₂ ), that is to say that the root directory of the tree A corresponds to a directory of this other tree.

In addition, in the computer 102 _1; a memory M1 is allocated to a process P1 for its execution. This memory M1 has a memory zone ZM1 that the process P1 is designed to share with other processes. The memory zone ZM1 is a part of the memory M1 having, in its extent, the same characteristics (for example, of which only read-only access is allowed) and / or comprising data forming one or more computer objects (for example , a stack ("stack" in English), a file mapped in memory from a hard disk or an array of information managed together).

In addition, in the computer 102 ₂ , a memory M2 is allocated to a process

P2 for its execution.

So that the process P1 can share data with the process P2, the tree A comprises a file F1, which is a virtual file associated with the memory zone ZM1, instead of being associated with data on a mass memory as for a classic file. The file F1 is further mapped in the memory M2, so as to obtain a file mapped in memory F1 ^* .

The computer installation 100 further comprises a load balancer 1 14 designed to distribute the processes, such as the processes P1, P2, between the different computers, such as the computers 102 _1; 102 ₂ of the computer installation 100 and to migrate the processes from one computer to another depending on the load variations on the computers. In the example described, the load balancer 1 14 is a computer program executed by the computer 102

With reference to FIG. 2, a communication method 200 between the first and second processes P1, P2 will now be described. During a step 202, in order to make the memory zone ZM1 available, the process P1 requests the file system SF1, SF2 to create a virtual F1 file and its association with the memory zone ZM1.

During a step 204, the file system SF1, SF2 creates the virtual F1 file and associates it with the memory zone ZM1.

For this purpose, in the example described, a directory path beginning with the root directory is created in the tree A of the file system SF1, SF2, and the file F1 is placed at the end of this path. Of course if all or part of the path directories already exist, they are reused.

For example, the path is: Root_Directory

/ Program_Group / Program_Group / Program

/ Process_group / Sub_Process_Group / Process / Instance / Domain / Sub_Domain, and the F1 virtual file is placed in the /Domain_directory directory. Of course, the names of the directories listed above are generic names, which are in practice replaced by common names. In addition, in other embodiments, one or more of previous directories may be omitted. For example, directories defining subgroups could be omitted when the processes listed by tree A are few.

Each of the directories of the path is associated with a unique universal identifier (UUID), the unique term to be understood in the sense of "very likely uniqueness".

The names of the directories / Program_group

/ Sub_Group_of_Program / Program are for example defined by an administrator. Each directory / program represents a program, that is, a set of instructions that is not running. The directories / Program_Group / SubGroup_Group are used to classify programs.

The names of the directories / Process_group and / Sub_Process_Group are for example defined by the load balancer 1 14 and correspond respectively to groups of computers and subgroups of computers to which belongs the computer running the process. Thus, in the example described, the file F1 is placed in the directories / Process_group and / Sub_Process_Group respectively corresponding to the computer group and the subgroup of computers to which belongs to the computer ^ 02 ^. Preferably, these directories therefore change each time the process is migrated from one computer to another of the computer installation 100 by the load balancer 1 14.

The names of the directories / Process and / Instance are for example defined by the SE1 operating system of the computer on which the 102 _Ï P1 process is running.

For example, the names of the / Domain and / Sub_Domain directories are defined by the P1 process. The directory / Domain represents a logical organization of the memory areas and the directory / Sub_Domain is used when the memory areas are numerous and must therefore be classified into several levels of hierarchy.

Thus, the tree A allows to group similar elements and to organize them in a logical hierarchy.

The name of the file F1, which is the memory area ZM1, is for example defined by the SE1 operating system of the computer 102 _Ï on which the P1 process is running.

The file F1 and the directories of the path are associated with respective permissions, allowing to control which process can access it. These permissions are for example UNIX type, that is to say that each file / directory belongs to a user (called owner) belonging to a group of users and that read, write and execute permissions are defined for the user. owner, for users in his group and for other users. Of course more detailed permissions may be used instead, for example of the type of access control list ("Access Control List"). In addition, preferably, the permissions of an object (directory / file) of the tree A are at least as restrictive as the permissions of the parent object.

During a step 206, the process P2 requests the operating system SE2 to map the file F1 in its memory M2.

During a step 208, the file system SF1, SF2 verifies that the permissions associated with the file F1 allow the P2 process to access it for reading.

If this is the case, during a step 210, the operating system SE2 maps the file F1 in the memory M2 so as to obtain a file mapped in memory F1 ^* . During a step 212, the second process P2 attempts to access data D of the file mapped in memory F1 ^* .

This access attempt goes through the memory management unit 1 12 ₂ which detects, during a step 214, whether the data D has already been copied to the file mapped in memory F1 ^* .

If it is the case, during a step 216, the memory management unit 1 12 ₂ provides the physical address corresponding to the virtual address of the data D so that the process P2 accesses it. Otherwise, during a step 218, the memory management unit 12 ₂ detects a page fault and requests the file system SF1, SF2 to provide it with the missing page containing the data D from the file F1. In addition, the SF1, SF2 file system receives this request.

In response to the read request, during a step 220, the file system SF1, SF2 retrieves the data D (in the example described, the missing page), in the memory part associated with this file F1, that is to say in the example described from the memory zone ZM1, so that the second process P2 can access it.

During a step 222, the file system SF1, SF2 supplies the data D (the missing page) to the memory management unit 1 12 ₂ .

During a step 224, the memory management unit 12 ₂ copies the data D (the missing page) into the file mapped in memory F1 ^* and supplies the process with the physical address of these data D.

During a step 226, the second process P2 accesses the data D in the file mapped in memory F1 ^* .

Steps 202 to 226 are executed each time the process P2 attempts to access data from the file mapped to memory F1 ^* .

Furthermore, when the P1 process is stopped, the SF1 file system, SF2 is preferably designed to transform the virtual file F1 _REE file F1 real i present on a mass storage device, preferably non-volatile, such as one of the hard disks 1 10 _1; 1 10 ₂ . The exchange mechanism ("swap" in English) can for example be used to transfer the data of the memory zone ZM1 in the file F1 _rée |. Thus, the data of the memory zone ZM1 remains accessible by other processes, even when the process P1 no longer exists. The method 200 can then be used again, except that in step 220, the file system SF1, SF2 retrieves the data D (in the example described, the missing page), in the file F1 _rée i present in a memory massive. In addition, the computer installation 100 preferably comprises a coherence mechanism designed to ensure consistency between the data D present in the memory zone ZM1 and their copy in the file mapped in memory F1 ^* . This consistency can be done for the memory zone ZM1 as a whole. Alternatively, the memory zone ZM1 can be divided into coherence units of predefined size, for example 4 kb. Each coherence unit can then be located in the memory zone ZM1 by an offset ("offset" in English), for example taken from the beginning of the memory zone ZM1. In the case where coherency units are used, coherence can be performed on each coherency unit. In this case, a virtual file can be created for each coherency unit and these virtual files are placed in a directory representing the memory area to which the coherency units belong. In the example described, the virtual file F1 would be associated with one of the coherency units and placed in a directory / Memory_zone placed in the / Domain directory or the / Sub_Domain directory if the latter is used.

Thus, in general, each virtual file is associated with a part of the memory M1 of the first process P1, for example the memory zone ZM1 or else a coherence unit of the memory zone ZM1.

With reference to FIG. 3, an example tree A is illustrated. The references used are the following:

302: root directory; 304 - 308: Application Group Directories; 310 - 312: application subgroup directories; 314 - 318: application directories; 320: process group directory; 322: process subgroup directory; 324 - 330: process directories; 332 - 336: Instance Directories; 338-340: domain directories; 342: subdomain directory; 344 - 348: virtual files respectively associated with portions of the virtual memory of the instance 336 of the process 328.

Directories 302, 306, 312, 316, 320, 322, 328, 336, 338, 342 form the path after which the virtual files 344, 346, 348 are placed. The virtual files 344, 346, 348 are for example respectively associated with three memory areas of the memory of the instance 336 of the process 328.

It is clear that a computer installation 100 and a method 200 such as those described above make it possible to transmit data from the process P1 to the process P2. Of course, a similar method can be implemented to allow the transmission of data from the P2 process to the process P1, using a virtual file F2 associated with a part ZM2 of the memory M2 and mapped in the memory M1 of the first process P1.

Note also that the invention is not limited to the embodiments described above. It will be apparent to those skilled in the art that various modifications can be made to the embodiments described above, in the light of the teaching just disclosed.

In particular, although the operating systems SE1, SE2, the file system SF1, SF2, and the load balancer 1 14 have been described as computer programs, the invention is not limited to case. Indeed, the functions associated with the instructions of these computer programs could be at least partly micro programmed or micro wired in dedicated integrated circuits. Thus, alternatively, each of the operating systems SE1, SE2, the file system SF1, SF2, and the load distributor 1 14 could be replaced by an electronic device composed only of digital circuits (without a computer program) for performing the same functions.

Similarly, all or part of the functions of memory management units ^' \ ^' \ 2, 1 12 ₂ can be realized by computer programs comprising instructions for carrying out their functions.

In the following claims, the terms used are not to be construed as limiting the claims to the embodiments set forth in this specification, but should be interpreted to include all the equivalents that the claims are intended to cover because of their formulation and whose prediction is within the reach of those skilled in the art by applying his general knowledge to the implementation of the teaching that has just been disclosed to him.

Claims

1. Method (200) of communication between first and second processes running in a computer installation (100) having a first memory (M1) allocated to the first process for its execution, said first memory (M1) being composed of addresses virtual devices to be converted to physical addresses by a memory management unit (112), the method (200) comprising the steps of:

a file system (SF1, SF2) of the computer installation (100) creates (204) a file (F1),

the file system (SF1, SF2) receives (218) a request to read data from the file (F1),

in response to the read request, the file system (SF1, SF2) retrieves (220) the data for the second process to access it,

the method (200) being characterized in that the file (F1) is a virtual file associated with a part (ZM1) of the first memory (M1), and in that the file system (SF1, SF2) retrieves the data from the part (ZM1) of the first memory (M1) leaving the file (F1) virtual.

The method (200) of claim 1, wherein the computing facility (100) includes at least first and second computers (102 _1; 102 ₂ ), on which the first and second processes respectively execute, and in which wherein the file system (SF1, SF2) is a distributed file system such that the file (F1) is accessible from each of the first and second computers (102 _1; 102 ₂ ).

3. Method (200) according to claim 1 or 2, wherein a second memory (M2) of the computer installation (100) is allocated to the second process for its execution, the second memory (M2) being composed of virtual addresses. intended to be converted to physical addresses by a memory management unit (12 ₂ ), the method (200) further comprising the steps of:

the file (F1) is mapped (210) in the second memory (M2) so as to obtain a file mapped in memory (F1 ^* ), the second process attempts to access (212) data from the file mapped in memory ( F1 ^* ), these data have not yet been copied to the file mapped in memory (F1 ^* ), the reception step (218) by the file system (SF1, SF2) of the read request being performed in response to this access attempt,

after the data recovery step (220) by the file system (SF1, SF2), the recovered data is copied (224) to the memory mapped file (F1 ^* ) for the second process to access.

4. Method (200) according to any one of claims 1 to 3, wherein the file (F1) is associated with permissions, the method (200) further comprising, before the step of recovering (220) data by the file system (SF1. SF2):

checking that the permissions associated with the file (F1) allow the second process (P2) to access it for reading.

5. Method (200) according to any one of claims 1 to 4, further comprising the creation of a directory path at the end of which the file (F1) is placed.

Method (200) according to claims 2 and 5 taken together, wherein the first computer (102 ^ belongs to a computer group of the computer installation (100) and in which one of the directories of the path corresponds to that group .

7. Computer program downloadable from a communication network and / or recorded on a computer-readable medium and / or executable by a processor, characterized in that it comprises instructions for executing the steps of a method ( 200) according to any one of claims 1 to 6, when said program is executed on a computer.

A computer installation (100) in which first and second processes are intended to execute, comprising:

a first memory (M1) intended to be allocated to the first process for its execution, this first memory (M1) being composed of virtual addresses intended to be converted into physical addresses by a memory management unit (1 12 ^, a system of files (SF1, SF2) designed to create a file (F1), receive a request to read data in the file (F1), and, in response to the read request, retrieve the data so that the second process can to access, the computer installation (100) being characterized in that the file (F1) is a virtual file associated with a part (ZM1) of the first memory (M1), and in that the file system (SF1, SF2) is designed to recover the data from the part (ZM1) of the first memory (M1) leaving the file (F1) virtual.