JP3906594B2 - CPU capacity adjustment method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a method of effectively using CPU resources of a computer system, and more particularly to a method of effectively using CPU resources that are wasted in a virtual computer system.
[0002]
[Prior art]
Conventionally, when a computer user constructs a virtual computer system composed of a plurality of virtual machines on an introduced computer (physical computer), the CPU capacity required for each virtual machine is totaled to obtain a total CPU capacity. A product model that satisfies the total CPU capability is determined from among the product models that are set stepwise depending on the size of the CPU capability. Then, the ratio (service rate) at which each virtual machine is serviced is specified according to the CPU capability of the determined product model. A function for designating a service rate for each virtual machine is disclosed in JP-A-9-81401.
[0003]
[Problems to be solved by the invention]
That is, the total CPU capacity required by the user is smaller than the CPU capacity of the introduced computer, and the difference CPU capacity is wasted.
[0004]
An object of the present invention is to allocate this wasted CPU capacity to a virtual machine that operates for monitoring or maintenance of the entire system and effectively uses it for operation management.
[0005]
[Means for Solving the Problems]
According to the present invention, a virtual computer used by a user on a physical computer (hereinafter referred to as a user-defined virtual computer) and a virtual computer for system maintenance (hereinafter referred to as a maintenance virtual computer) are controlled by the system control means . In the virtual machine system operating below, the virtual machine system has a configuration information storage means for storing the configuration information of each virtual machine and a ratio value to the CPU capacity of the entire system allocated to the user-defined virtual machine (hereinafter referred to as CPU capacity coefficient). CPU capability storage means for storing the user-defined virtual machine configuration information includes a user-defined service rate, and the system control means stores the user-defined virtual machine in the configuration information storage means. Register, register the maintenance virtual machine in the configuration information storage means, and store the user-defined virtual machine and the storage virtual machine. Perform initial processing of the use virtual machine, the virtual machine the maintenance activated based the CPU capacity of the entire system to the rest of the percentage value obtained by subtracting the CPU power coefficients stored in the CPU power storage unit, the user-defined virtual The user-defined virtual machine is activated based on a service rate obtained by multiplying the user-defined service rate of the computer by the CPU capability coefficient.
[0006]
Also, the present invention calculates the difference between the CPU capacity of the physical computer and the total CPU capacity by summing up the CPU capacity required for each virtual machine in a virtual machine system in which a plurality of virtual machines operate on the physical machine. Then, a virtual machine having the CPU capacity of the difference is generated.
[0007]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0008]
FIG. 1 shows a configuration of a physical computer system to which the present invention is applied. 10, 11,..., 1 n indicate a physical processor IP 0, a physical processor IP 1,. Reference numeral 20 denotes a main storage device. Reference numerals 30, 31,..., 3n denote an I / O processor IOP0, an I / O processor IOP1,. Reference numeral 40 denotes a service processor (SVP), 50 denotes a console display device (CD), and 60 denotes a hard disk device (HD). Reference numeral 70 denotes a control program (hypervisor) for the entire virtual machine system.
[0009]
FIG. 2 shows a configuration of a virtual machine system according to the present invention. Reference numeral 70 denotes a hypervisor, and 71, 72, and 7X denote virtual computers (hereinafter referred to as LPAR) LPAR1, LPAR2, and LPARX, respectively. LPAR1 and LPAR2 are user-defined virtual machines, and LPARX is a maintenance virtual machine newly provided in the present invention. 81n represent n logical processors (hereinafter referred to as LIP) LIP1,..., LIPn under LPAR1 control. Similarly, 821,..., 82n represent n logical processors LIP1,..., LIPn under LPAR2 control. 8Xn represents n logical processors LIP1,..., LIPn under LPARX control. The hypervisor 70 exists in the main storage device 20. In this embodiment, the number of virtual machines defined by the user is two, but the number is not limited to this.
[0010]
FIG. 3 shows a configuration of the hypervisor 70. The hypervisor 70 includes a scheduler 90 that schedules each LPAR, an initial processing unit 91 that performs initial processing and activation of each LPAR, a resource management unit 92 that manages physical resources supplied from the SVP and logical resources in each LPAR, It comprises a logical processor control unit 93 that controls the execution of the logical processor of each LPAR, a frame control unit 94 that controls screen display of resource allocation of each LPAR, the operating status of each LPAR, and the like.
[0011]
FIG. 4 shows data storage states of the main storage device 20 and the hard disk HD 60. The hard disk HD 60 is provided with a hypervisor storage unit 61, an I / O configuration information unit 62, and a monitor program storage unit 63. When the hypervisor is activated, the hypervisor stored in the hypervisor storage unit 61 is loaded into the corresponding part (23) of the extended hardware system area (EHSA) 22 of the main storage device 20. The LPARX monitor program stored in the monitor program storage unit 63 is also loaded into the corresponding part (24) of the EHSA 22 when the hypervisor is activated. A hardware system area (HSA) 21 stores hardware control information when the computer is started up. Reference numeral 24 denotes an LPAR information table in which configuration information of each LPAR is stored.
[0012]
FIG. 5 shows details of the LPAR information table 100. Reference numeral 101 denotes an LPAR name column. Reference numeral 102 denotes a main storage capacity setting field, which indicates the main storage capacity given to each LPAR. Reference numeral 103 denotes a logical IP definition column, which indicates a logical IP generated by each LPAR. Reference numeral 104 denotes a user-defined CPU service rate setting field. Reference numeral 105 denotes a CPU service rate (actual service rate) that is actually assigned to each LPAR.
[0013]
FIG. 6 shows an SVP CPU capability control frame newly provided in the present invention. The CPU capability control frame specifies the user's required CPU capability by a coefficient when the CPU capability of the computer system introduced by the user is 100. For example, the CPU capacity of this computer system is 100 MIPS, the CPU capacity of LPAR1 and LPAR2 of the virtual machine used by the user is A1 MIPS and A2 MIPS, respectively, and the sum of A1 and A2 is 80 MIPS. In this case, a value of 80 is designated as the CPU capability coefficient. The CPU capability coefficient specified in the CPU capability control frame is stored in the CPU capability specification table (FIG. 7) and managed by the resource management unit 92 of the hypervisor 70.
[0014]
Next, a CPU capacity control method in this embodiment will be described.
[0015]
First, the system is started by designating the physical computer in the virtual computer mode. When the system is started up, the hypervisor is loaded from the hard disk drive (HD) to the extended hardware system area (EHSA) and activated.
[0016]
In the initial process, the hypervisor prepares a maintenance LPAR (LPARX) separately from the LPAR group used by the user, and automatically sets the main memory and logical IP for LPARX. When the CPU capability coefficient is designated, a capability other than the designated CPU capability (20 in the above example) is assigned to LPARX and activated. Thereafter, LPAR1 and LPAR2 are activated, and 3LPAR is activated as a whole system. LPAR1 and LPAR2 are assigned main memory and CPU capacity designated by the user, respectively. The CPU capacities of LPAR1 and LPAR2 are specified by a ratio value (actual service rate) of the entire system in accordance with the required capacity values (A1, A2). Note that the user designates only a ratio value (user-defined service rate) according to the values (A1, A2) designated as the necessary CPU capacity.
[0017]
Next, the allocation (service rate) of the CPU capacity of the entire system to each LPAR will be specifically described. Conventionally (when there is no CPU capability designation function), for example, when A1: A2 = 60: 40 for the required CPU capabilities A1 and A2 of each LPAR, the hypervisor moves to LPAR1 and LPAR2 according to the settings in the LPAR frame Allocate 60% and 40% of the CPU capacity of the entire system. In contrast, in the present invention, 80% of the CPU capacity of the entire system is assigned to LPAR1 and LPAR2, and the remaining 20% is assigned to LPARX. Specifically, 48% (60% × 0.8) of the CPU capacity of the entire system is allocated to LPAR1, and 32% (40% × 0.8) of the CPU capacity of the entire system is allocated to LPAR2. .
[0018]
In order to prevent the CPU service rate of the user-defined LPAR from fluctuating, LPARX needs to keep its CPU capacity constant (20% in this embodiment). For this purpose, a resource cap function (RC function) that guarantees the upper limit of the CPU service rate is set for LPARX. The RC function is a function that inhibits a CPU service exceeding a specified upper limit value of the CPU service rate during a certain interval. Furthermore, a wait completion function (WC function) that guarantees the lower limit of the CPU service rate is set. The WC function is a function that keeps the control suspended by the H / W even if the program WAITs, and can use the CPU up to the specified value of the CPU service rate even when the frequency of WAITing is high. The RC function and the WC function are known functions, and are disclosed in Japanese Patent Laid-Open No. 9-81401. As described above, by setting the RC function and the WC function, LPARX can always maintain a constant CPU service rate.
[0019]
Finally, a schematic processing flow of the hypervisor will be described with reference to FIG. FIG. 8 shows from the hypervisor initial process to the user-defined LPAR registration process to the user-defined LPAR activation process. In the user-defined LPAR registration processing step 401, the specified LPAR name is extracted from the I / O configuration information 62 of the hard disk device HD60 and registered in the LPAR information table 100 managed by the hypervisor resource management unit. In the LPARX registration processing step 402, the LPAR newly provided in the present invention is registered under the name LPARX. Also, the definition information of the main storage area and logical IP assigned to LPARX is registered in the LPAR information table 100. At this time, as many logical IPs as the physical IPs are registered. This logical IP is set to a shared mode (a mode in which a physical IP is shared with another LPAR). In an all LPAR initial processing step 403, initial processing of the LPAR registered in the LPAR information table 100 is performed. Specifically, a logical IP corresponding to each LPAR is generated and an internal table of the logical IP is initialized. In a CPU capability designation determination step 404, the capability designation value table 300 managed by the resource management unit 92 is read, and if there is a CPU capability designation (when the CPU capability coefficient is not 100), the LPARX activation processing step 405 is executed. To do. When there is no designation (when the CPU capacity coefficient is 100), LPARX activation processing is not performed. In the LPARX activation processing step 405, LPARX activation processing is performed. Specifically, it is a process of bringing the main memory, logical IP, etc. allocated to LPARX online, and online processing is performed using the configuration information set in the LPAR information table 100. Thereafter, the monitor program is automatically IPLed. Thus, the hypervisor initial process is completed. Thereafter, the user-defined LPAR configuration information setting process 406 can be performed in the LPAR frame. After the configuration is set, a user-defined LPAR activation process 407 is performed, and a CPU service rate based on the CPU capability coefficient described above is assigned. Thereafter, after IPLing the OS for each user-defined LPAR, the system is put into actual operation. Note that a monitor program operating in LPARX can also see the operating state of a user-defined LPAR in the LPAR frame.
[0020]
When dynamically changing the CPU capacity coefficient, the hypervisor receives the state change notification from the SVP, reads the information, and changes the capacity specification value table 300 of the resource management unit 92. Thereafter, the hypervisor determines an allocation value for the entire system in the same way as the CPU allocation method described above for all LPARs, and determines the service rate of each LPAR. In order to dynamically change the CPU capability coefficient, an interface (external interrupt) for notifying the hypervisor from the SVP is used. That is, by adding a function of reading CPU capacity coefficient data to the HVA instruction that controls the exchange between the SVP and the hypervisor, the CPU capacity coefficient is notified to the hypervisor at the time of interrupt. The process of dynamically changing the CPU allocation value of each LPAR can be realized by using an LPAR frame.
[0021]
FIG. 9 compares the CPU usage rate (service rate) of the prior art and the CPU usage rate (service rate) according to the present invention. For LPAR1 and LPAR2, the CPU usage rates α1 and α2 corresponding to the required CPU capacities A1 and A2 do not change according to the present invention. That is, conventionally, the event search uses a CPU capability (100-α1-α2) other than the necessary CPU capabilities A1 and A2, but this CPU capability (100-α1-α2) is assigned to LPARX according to the present invention. .
[0022]
【The invention's effect】
According to the present invention, in a virtual machine system, a user can use only necessary CPU capacity, and system operation management can be performed by allocating CPU capacity unnecessary for the user to a maintenance virtual machine. As a result, CPU resources can be effectively utilized.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of a physical computer system to which the present invention is applied.
FIG. 2 is a configuration diagram of a virtual machine system according to the present invention.
FIG. 3 is a diagram showing a configuration of a hypervisor in FIG. 2;
4 is a diagram illustrating a storage state of data in the main storage device and the hard disk HD in FIG. 1. FIG.
FIG. 5 is a diagram showing details of an LPAR information table in FIG. 4;
FIG. 6 is a diagram showing a CPU capability control frame according to the present invention.
FIG. 7 is a diagram showing a CPU capability designation table according to the present invention.
FIG. 8 is a diagram showing a schematic processing flow of the hypervisor in FIG. 2;
FIG. 9 is a table showing CPU usage rates (service rates) in the prior art and the present invention.
[Explanation of symbols]
10, 11, ..., 1n Physical processor (IP)
20 Main storage devices 30, 31, ..., 3n I / O processor (IOP)
40 Service Processor (SVP)
50 Console display device (CD)
60 Hard disk device (HD)
70 Hypervisor (CP)
71, 72, 7n Virtual machine (LPAR)
811,..., 81n Logical processor (LIP)
90 scheduler 91 initial processing unit 92 resource management unit 93 logical processor control unit 94 frame control unit 200 CPU capability control frame

Claims (3)

A virtual computer used by a user on a physical computer (hereinafter referred to as a user-defined virtual computer) and a system maintenance virtual computer (hereinafter referred to as a maintenance virtual computer) operate under the control of the system control means . In the virtual machine system, the virtual machine system stores configuration information storage means for storing the configuration information of each virtual machine and a ratio value (hereinafter referred to as a CPU capacity coefficient) with respect to the CPU capacity of the entire system allocated to the user-defined virtual machine. CPU capacity storage means for storing, the user-defined virtual machine configuration information includes a user-defined service rate, the system control means registers the user-defined virtual computer in the configuration information storage means, A maintenance virtual machine is registered in the configuration information storage means, and the user-defined virtual machine and the maintenance virtual machine are registered. Perform initial processing of the machine, on the basis of the CPU capacity of the entire system to the rest of the percentage value obtained by subtracting the CPU power coefficients stored in the CPU power storage unit activates the virtual machine the maintenance of the user-defined virtual machines A CPU capacity adjustment method, comprising: activating the user-defined virtual machine based on a service ratio obtained by multiplying the user-defined service ratio by the CPU capacity coefficient. The physical computer has a service processor, the system control unit CPU power adjustment method according to claim 1, wherein changing the CPU power coefficient according to an instruction from the service processor. In a virtual machine system in which a virtual machine used by a user on a physical machine (hereinafter referred to as a user-defined virtual machine) and a virtual machine for system maintenance (hereinafter referred to as a maintenance virtual machine) operate, Configuration information storage means for storing computer configuration information and CPU capacity storage means for storing a ratio value (hereinafter referred to as a CPU capacity coefficient) to the CPU capacity of the entire system allocated to the user-defined virtual machine, The configuration information of the user-defined virtual machine includes a user-defined service rate and operates based on the remaining ratio value obtained by subtracting the CPU capability coefficient stored in the CPU capability storage means from the CPU capability of the entire system . Virtual machine and service obtained by multiplying the user-defined service rate of the user-defined virtual machine by the CPU capability coefficient Virtual computer system characterized by comprising a user-defined virtual machine that operates based on.

Images