JP2008276320A - Virtual system control method and computer system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To determine an optimum destination in consideration of blade features and redundancy of power supply and fan when migrating a virtual system between a plurality of CPU blades. <P>SOLUTION: A computer system provided with a plurality of CPU blades and configured to migrate a virtual system running on one CPU blade to another CPU blade includes a storage space for storing a management table registering information about each CPU blade including information about a corresponding power supply and/or cooling fan, and a control means for, when a virtual system on any CPU blade must be migrated to another CPU blade, selecting a CPU blade satisfying predetermined conditions in comparison with the source CPU blade and setting it as the destination of the virtual system by reference to the management table. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a virtual system management method and system, and more particularly to a method and system for managing and controlling migration of a virtual system between a plurality of CPU blades in a system in which a plurality of computing nodes are mixed in the same chassis, such as a blade system. About.

In recent years, in a computer system such as a blade system in which a plurality of CPU blades are housed in the same casing, there is a technique for migrating a virtual system from one CPU blade to another due to a system failure or the like (for example, Patent Document 1). . In general, in a virtual computer system based on a VM (virtual machine), this migration is performed by using a virtualization software that allows a designated server to stand by as a backup, and if a certain server is overloaded or a light failure occurs frequently. The virtual system is migrated to the spare server using the system live migration function. As a result, load leveling and system failure due to failure are avoided, and smooth system operation is realized.
JP 2006-244481 A

  However, in the conventional system, since the movement destination of the virtual system is uniquely determined in advance, not only in the case of load distribution / failure avoidance using a spare server, the virtual system is moved between servers in operation. However, there is an inconvenience that an appropriate server cannot be selected as a destination at that time. In particular, CPU blades provide not only performance, but also high-availability, high-reliability, high-capacity memory, and other system-added values depending on applications and implementation environments. Although the number of variations has increased, the spare server prepared in advance does not necessarily have the same specifications as the source CPU blade, so there is a possibility that performance and safety will be reduced after moving the virtual system. It was. As described above, the conventional system cannot dynamically select a safer and more robust server, and cannot effectively utilize system resources utilizing the features of each blade.

  In view of the above problems, the present invention is a computer system having a plurality of CPU blades. When a virtual system on one blade is migrated to another blade, the characteristics of the blade, the operating environment such as the fan and power supply, and the occurrence of a failure It is an object of the present invention to provide a method and system for realizing highly available and dynamic resource allocation by selecting an optimum blade in consideration of the situation and the like.

To achieve the above object, the present invention provides a computer system configured to migrate a virtual system configured with a plurality of CPU blades and operating on one CPU blade to another CPU blade.
Refer to the management table when there is a need to migrate a storage area that stores at least a management table in which information related to each CPU blade is registered, and a virtual system on any CPU blade to another CPU blade. The main feature is that it comprises control means for selecting another CPU blade that satisfies a predetermined condition in comparison with the migration source CPU blade and setting it as the migration destination of the virtual system.

  In this system, the management table further stores information on the corresponding power supply and / or cooling fan for each CPU blade, and this information is taken into account when setting by the control means.

  The control means manages the system operating environment index for each blade obtained by performing a predetermined weighting operation on each registered item for each CPU blade by referring to the management table in advance or periodically according to a predetermined condition. And another CPU blade having an index equal to or higher than that of the source CPU blade is selected as the destination.

  In addition, the management table further includes a column for managing the CPU blades by grouping them according to an operator input, and the control unit selects a destination CPU blade from the same group as the source CPU blade. It is preferable.

  According to the computer system of the present invention, information on available CPU blades is managed by the management table, and information on each CPU blade is required when a virtual system that operates on any of the CPU blades needs to be moved. By comparing these, the most appropriate CPU blade at that time is dynamically selected as the destination, so the cost and labor of preparing a spare blade can be saved, and problems due to different performance and reliability of the destination blade can be prevented. can do.

  In addition, not only the performance of the CPU blade, but also redundancy and power saving of the power supply and fan are incorporated in the migration judgment conditions, so the virtual system can be moved appropriately while maintaining reliability with the power supply and fan as elements. Can be made.

  In addition, since the system operating environment index for each CPU blade is calculated in advance and a CPU blade having an index equal to or higher than that is determined as a destination when a failure occurs, the numerical value of each CPU blade after the failure has occurred. The time required from the occurrence of a failure to the completion of the transition is drastically shortened as compared with the configuration in which the destination is selected by acquiring the destination.

  In addition, each CPU blade is managed in groups according to operator settings, and the virtual system movement is restricted within the group, enabling virtual system control in line with the user's intentions, according to the user's intended operational purpose. It is possible to utilize resources in the right place for the right person

  The best mode for carrying out the present invention will be described in detail below with reference to the drawings.

  FIG. 1 is a schematic view showing an embodiment of the present invention. As shown in the figure, in the embodiment of the present invention, a plurality (two) of blade systems 100, 200 are each mounted with a plurality of CPU blades, a chassis management processor, a plurality of power supply units (PSU), and a cooling fan. Yes. Each CPU blade is deployed with virtualization software (VM), and a plurality of systems operate on the VM. Here, for example, the cooling fan does not need to be provided for each blade, and may be configured to cover a plurality of blades with one fan.

  Each CPU blade has a non-volatile memory (not shown) for storing BMC (Baseboard Management Controller) and FRU information (Field Replacement Unit Information). In this memory, the CPU blade type and performance ( For example, the number of CPU cores, frequency, cache size, etc.) and hardware (HW) fault tolerance set by the vendor are registered. Here, HW fault tolerance is an index that can be compared and expressed between blades according to HW reliability prediction / results and the level of supported high availability functions, etc., and is preset by the vendor at the time of blade shipment And The registered contents of the memory are read when each blade is powered on and transmitted from the BMC to the chassis management processor 10.

  The case management processor 10 receives the above-described relatively comparable information of the CPU blade from the BMC. Also, FRU information of PSU and / or cooling fan (FAN) under management of each CPU blade, redundancy of each PSU / FAN, presence / absence of power saving function, VM type information, etc. are acquired, and these information are acquired. Are registered in the management table of the storage area (not shown). An example of this management table is shown in FIG.

  As shown in FIG. 2, in the management table, for each CPU blade of each chassis, as the availability information, HW fault tolerance, VM type, cooling fan redundancy, power supply redundancy index values set by the vendor Is registered. Also, as performance information, CPU performance, memory capacity, IO, and VM type index values are registered. Further, an index value for power saving of the CPU, fan, and power source, a group number, and finally an in-group system operation index value are registered.

  In this embodiment, the HW fault tolerance index takes a value such as “1” or “2”. In the table of FIG. 4, for example, in the case 100, the CPU blades 12 and 1n have high availability such as FT blades. An index of level “2” indicating that the blade is a trusted blade is registered. In the HW failure history, for example, level 5 is set as an initial value, and is set to be lowered by 1 level when a predetermined number of failures occur in a predetermined period. In this case, information on the event of occurrence of failure and information on the number of times can be managed by each blade, and a level change notification can be notified to the management processor 10 when the above conditions are satisfied. Also, as power source / fan redundancy, PRU / FAN FRU information is read, for example, “1” when N + 1 redundancy is present, “2” when redundancy is 2N, UPS (uninterruptible power supply) The index of “3” is registered. Similarly, the performance information has a separate conversion table (not shown) in which a combination of numerical ranges such as the number of CPU cores, frequency, and cache size is associated with the performance index (not shown), and actual values are mapped to this table. Thus, the performance index can be derived. Further, it is determined from the FRU information whether the CPU, fan, power source, etc. are power saving power sources or fans. For example, when power saving is not performed, “0” is registered.

  The chassis management processor 10 also manages these CPU blades in groups by registering the group numbers of the CPU blades in the management table shown in FIG. This group number can be configured so that an operator can directly set it via a console (not shown) of the housing management processor 10. This grouping is used to limit the migration destination when migrating a virtual system to a specific group.

  The case management processor 10 may further calculate a system operating environment index from each index data of each CPU blade and register it in the management table when the above index data is prepared when the power is turned on or periodically. The system operating environment index can be derived by further individually calculating each item of the management table as necessary. In the case of the present embodiment, for example, the weighting values (coefficients) of the respective items are determined in advance so that the obtained system operation environment index becomes level 1-5, and the weighting calculation is performed according to this. The coefficient of each item can be easily determined by those skilled in the art in consideration of the implementation environment and various conditions.

  The case management processor 10 further exchanges information with the case management processor 20 of a different blade case 200 to complete and manage a management table of available CPU blades. In addition, when a new CPU blade is added during system operation, the chassis management processor 10 detects the insertion of the CPU blade, immediately obtains the CPU blade information from the BMC, and registers it in the management table. Recalculate and manage system operating environment indicators.

  The case management processor 10 determines the system operation environment index of the management table when it is necessary to migrate the virtual system of any CPU blade to another blade under a predetermined condition such as when frequent frequent failures occur. , An activation destination is selected from blades having a system operating environment index equal to or higher than that of the movement source CPU blade in the same group, and notified to the BMC of the CPU blade. In this case, the chassis management processor 10 continuously monitors the load of each CPU blade, and may select a CPU blade having a smaller load. In addition, when there is no other CPU blade whose system operating environment index is equal to or higher than that of the movement source, it may be configured to select another CPU blade having a higher index, and the index is a predetermined value. If different from the above, the system may be configured to continue operation without performing migration.

  When the VM of each CPU blade obtains the movement destination information from the chassis management processor 10, it executes the movement of the virtual system by inter-VM control. As a result, for example, if the system operation index moves to a CPU blade with the same or higher, it is possible to move to the same or improved level before the movement without lowering the system environment level in which the virtual system operates. Realizes optimal system resource allocation and robust system operation.

  An example of the migration operation of this system will be described below. The chassis management processor 10 collects detailed CPU / memory information, HW failure history information, and HW fault tolerance index information in the FRU from the BMC of the CPU blade 11-1n in the chassis when AC is on. This is reflected in each individual index in the management table of FIG. Further, the FRU information of the directly subordinate PSUs 11-13 and FAN11-13 is read, and the power supply / fan redundancy and the power saving index are registered in the management table. Further, when each of the CPU blades 11-1n is DC-on, the BMC obtains information indicating the type of the VM, and the housing management processor 10 obtains the information from the BMC and reflects it in the management table. Then, a system operation environment index is calculated from these pieces of information and registered in the management table. Further, the system operator inputs the grouping of each CPU blade from a console (not shown) as necessary, and this is registered in the management table.

  Here, the type of VM is rating information arbitrarily determined based on information such as measurement results of a plurality of VMs at a certain point in time and positioning in a vendor, for example, “high-medium-low”, “ It can be expressed by an index such as “1-5”. That is, various products of VM are released from several companies, and each product has a difference in availability, scalability, performance, etc., and even in the same product, there are differences in these indexes between editions and versions. In consideration of these differences, the vendor side arbitrarily determines the type of VM and registers it at the time of manufacture.

  On the other hand, the case management processor 20 also obtains the same type of information from the BMC of the CPU blade 21-2m in the case when AC is on, creates a similar management table, and calculates the system operating environment index. Further, communication is performed between the cases, information in these management tables is exchanged, and all information as shown in FIG. 2 is managed.

  In such a situation, when it is detected that a HW minor failure such as a correctable error such as a 1-bit error frequently occurs or the blade is overloaded in the CPU blade 12 shown in FIG. The CPU blade 12 registers failure history information in the BMC. The case management processor 10 changes the failure history index of the CPU blade 12 in the management table according to a predetermined algorithm based on the report from the BMC. This can be set, for example, by lowering the index by one rank when a predetermined number of minor faults are reported within a predetermined period or when an overload state continues for a predetermined period or longer.

  In this way, when a predetermined condition is met, for example, when the failure history of the CPU blade 12 decreases from level 5 to level 3, the case management processor 10 needs to migrate the virtual system of the CPU blade 12. judge. In this embodiment, this condition is set, for example, when the failure history index is lowered by two ranks, but this is not limited to this example, and other criteria may be applied.

  At this time, the chassis management processor 10 refers to the management table of FIG. 2 and the CPU blades are set to the same group # 1 as the CPU blade 12 (blade 1n of the chassis 100, blades 22 and 2m of the chassis 200). Among them, a CPU blade having the same system operating environment index as high level 5 and having a low load is selected as a destination CPU blade. At this time, the chassis management processor 10 communicates with the chassis management processor 20 of the chassis 200 to acquire and select information such as the load status of each blade. For example, in FIG. 3, when there are two virtual systems in the migration source CPU blade 12, among the CPU blades belonging to group # 1 with the same system operating environment index of level 5, the CPU blade of the chassis 200 with a low load An embodiment is shown in which 22, 2m is selected as the movement destination. This information is communicated between the case management processors 10 and 20, and the BMC of the CPU blade 12 is notified of the upper side of the movement destination. Receiving this, the VM of the CPU blade 12 obtains the movement destination information from the BMC for the material measurement, and executes the movement of the two virtual systems by the VM-VM control as shown in FIG.

  Through the above steps, the present invention can dynamically select the optimal destination blade of the virtual system at that point in consideration of the performance of each CPU blade and the characteristics of the power supply / fan. It is possible to allocate the optimal system resources and operate the system robustly with the conventional system configuration without adding hardware. In addition, the destination can be dynamically defined without lowering the system environment level in which the virtual system operates, and there is no need to prepare a spare standby server in the event of a failure or overload, so the user's TCO (Total Cost of Ownership) can be reduced. In addition, the virtual system can be deployed on the basis of HW characteristics such as high performance, high availability, and high reliability, and can be restricted to movement within a group that can be defined by the user. It is possible to utilize resources in the right place for the right material.

  Next, another embodiment of the present invention will be described. In the above embodiment, the CPU blades are grouped to limit the migration destination of the virtual system within the same group. However, this grouping is not an essential configuration, for example, as shown in FIG. 1 or FIG. In addition, a CPU blade with a low system load may be selected from CPU blades having a system operation environment index equal to or higher than that without considering a group. In the above embodiment, the chassis management processor selects the destination. For example, the system operating environment index of each CPU blade is notified from the BMC to the VM when the VM is started up or in operation, and the CPU blade The VM may be configured to select a destination when moving the virtual system from these pieces of information.

  Yet another embodiment will be described with reference to FIG. In the embodiment of FIG. 5, a management server 300 is provided on the network connecting the casings 100 and 200. The management server 300 manages the management table shown in FIG. 2 and controls the migration destination of the virtual system. Execute. In this embodiment, each chassis management processor obtains information on CPU blades in combination with each BMC in accordance with instructions of management software running on the management server 300, and directly obtains information on power supplies / fans. Notify

  The chassis management processors 10 and 20 of each chassis receive detailed information on CPU / memory of each CPU blade, HW failure history information, FRU from the BMC of the CPU blades 11-1n and 21-2m in each chassis when AC is on. The HW fault tolerance index information is collected and transferred to the management server 300. Also, redundancy and power saving information is acquired from the power supply / fan FRU information and transferred to the management server 300. Management software (not shown) running on the management server 300 registers these pieces of information in the management table shown in FIG. 2, and calculates the environmental index in the same manner as described above. Even when a CPU blade is newly added in any case, the case management processor detects the insertion of the CPU blade and transmits information to the management server 300. When the management software receives the information, it immediately recalculates the system operating environment index and manages the information.

  For example, when a minor failure frequently occurs or an overload occurs in the CPU blade 12, this is notified from the BMC of the CPU blade 12 to the management server 300 via the chassis management processor 10, and the management table is updated. The management software selects the CPU blades 22 and 2m having the same system operating environment index level 5 and a low load among the CPU blades valued in the same group # 1 as the migration destination of the virtual system. The 12 VMs are directly instructed to move the two virtual systems without going through the chassis management processor 10. In this way, the virtual system can be moved smoothly across a plurality of housings.

  The virtual system control method and computer system according to the present invention can be suitably used in the information processing industry including the network industry.

It is a figure which shows the structure of the Example of the computer system concerning this invention. It is a figure which shows an example of a management table. It is a figure which shows the example of a movement of the virtual system which limited the movement destination by grouping. It is a figure which shows the example of a movement of the virtual system based on a system operation environment parameter | index. It is a figure which shows the structure of another Example of the computer system concerning this invention.

Explanation of symbols

100, 200 Blade enclosure 10, 20 Enclosure management processor 11-1n, 21-2m CPU blade PSU Power supply unit FAN Cooling fan 300 Management server

Claims (8)

In a computer system configured to migrate a virtual system configured with a plurality of CPU blades and operating on one CPU blade to another CPU blade,
Refer to the management table when there is a need to migrate a storage area that stores at least a management table in which information related to each CPU blade is registered, and a virtual system on any CPU blade to another CPU blade. A computer system comprising: a control unit that selects another CPU blade that satisfies a predetermined condition in comparison with a migration source CPU blade and sets it as a migration destination of the virtual system.   2. The system according to claim 1, wherein the management table further stores information on a corresponding power supply and / or cooling fan for each CPU blade, and this information is taken into account when setting by the control means. A featured computer system.   3. The system according to claim 1, wherein the control unit is obtained by performing a predetermined weighting operation on each registered item for each CPU blade by referring to the management table in advance according to a predetermined condition or periodically. A computer system which manages a system operation environment index for each blade, and selects another CPU blade having an index equivalent to or higher than that of a source CPU blade as a destination.   4. The system according to claim 1, wherein the management table further includes a column for managing the CPU blades by grouping according to an operator input, and the control unit includes the movement source. A computer system, wherein a destination CPU blade is selected from the same group as the CPU blade.   In a virtual system control method for migrating a virtual system operating on one CPU blade to another CPU blade in a computer system comprising a plurality of CPU blades, at least CPU and memory information and fault history for each CPU blade When the information, the managed power source and / or the cooling fan information are registered in the management table, and the virtual system on any CPU blade needs to be migrated to another CPU blade, the control means can control the management table. And selecting another CPU blade that satisfies a predetermined condition related to the source CPU blade and setting it as the destination of the virtual system.   6. The method according to claim 5, wherein the management table further stores information on a corresponding power supply and / or cooling fan for each CPU blade, and this information is stored in the CPU blade of the destination by the control means. A virtual system control method which is considered at the time of selection.   7. The method according to claim 5, wherein the control means obtains a predetermined weighting operation for each registered item for each CPU blade by referring to the management table in advance or under a predetermined condition in advance. A virtual system control method that manages a system operating environment index for each blade, and selects another CPU blade having an index equal to or higher than that of a source CPU blade as a destination. The method according to any one of claims 5 to 7, wherein the management table further includes a column for grouping and managing the CPU blades according to an input from an operator, and the control means includes the source of the movement. A virtual system control method comprising selecting a destination CPU blade from the same group as a CPU blade.

Images