JP2003208345A - Network type storage device - Google Patents

Abstract

(57) [Problem] To configure a network-type storage device with a plurality of storage devices, it is difficult to adopt a configuration that balances availability at the time of failure and cost. So far, when a network type storage device is configured by a plurality of storage devices, control information (metadata) for managing the storage device and actual data are arranged without particular distinction. In the present invention, a distinction is introduced between these two arrangement controls, and the redundancy of control information alone is increased to improve fault tolerance. By adopting the above configuration,
A network-type storage device that achieves high availability without causing a large increase in cost can be realized.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a configuration of a network type storage device, and more particularly to a technique for realizing a network type storage device having a low cost configuration which can operate without loss of availability against a failure.

[0002]

2. Description of the Related Art A disk array device is a device that uses RAID technology to improve data transfer performance and fault tolerance by accessing a plurality of disk devices, that is, a plurality of storage devices in parallel. For this RAID, refer to "A Case for Redundant Arrays of Inexpensiv
e Disks (RAID) "; In Proc. ACM SIGMOD, June 1988
Detailed information (published by the University of California, Berkeley). A disk array device using this RAID technology is currently widely used as a large-capacity storage device.

A disk array device adopting a cluster configuration is known as a development of the disk array device. As a method of constructing a high-performance and highly-reliable system at a lower cost than in the past, in the field of servers such as databases, a technology has been known from a comparatively early stage in which multiple independent servers are combined in a high-speed connection network to form a cluster configuration. Was there. As a large-scale system configuration, rather than constructing a single system with extremely high performance, a configuration in which a large number of systems with slightly inferior performance are combined in a high-speed connection network has a certain level of performance or more. This is because it has been proved that the performance of the system can be improved with respect to the cost.

Up to now, a disk array device has generally been constructed using a single controller, but recently, a similar configuration has begun to be studied even if it is placed in a disk array. For example, Japanese Patent Laid-Open Publication No. 2001-27972 discloses the configuration of such a disk device.

Another conventional technique which is a premise of the present invention is a technique relating to a network-connected disk array device. In current disk array systems, storage array networks (hereinafter abbreviated as SAN) and network storage systems (hereinafter abbreviated as NAS) are widely used as network technologies.
First, SAN is a network technology that uses Fiber Channel, which is the interface to the disks currently used as standard, and a disk array is treated as a normal disk. As a feature, there is an advantage that it is easy to pursue high speed because of the protocol with little overhead on Fiber Channel. However, since it is treated as a disk, it lacks the functions necessary for client connection such as disk write exclusion, and Fiber Channel is less popular than ordinary IP networks and the system becomes expensive. Since the protocol overhead of the IP network is large and the overhead for realizing the file system is also relatively large, the disk array tends to be expensive compared to its performance.

On the other hand, the NAS is a method for realizing a general-purpose file system on a disk array and connecting it to an IP network to provide it as a file server. The advantage is that it is easy to add various functions as a file system. However, the protocol overhead of the IP network is large, and the overhead for realizing the file system is also relatively large, so that the disk array itself tends to be expensive compared to the performance.

Even in such a network connection type system, various studies have been made on the above cluster type system in order to improve data integrity particularly by distributed arrangement. As a typical example thereof, US Pat. No. 5,832,222
In JP-A-No. 1994-2003, there is disclosed a technique for configuring a RAID system among network-type disk array devices distributed geographically at a plurality of locations for disaster response. US Patent Publication 5,
No. 742,792 discloses a copying technique for configuring RAID 1 with a remote site. The cluster type disk array device configuration based on the distributed arrangement and the like as described in these publications is a system assuming a dedicated line and a SAN, and is not intended for NAS technology. Of course, it can be applied as it is as a technology for simply connecting to a network, but since it is necessary to maintain consistency as a file system in the NAS device, the control data and data of the target file system are updated synchronously. However, there is a problem that the overhead becomes larger. Besides, various additional functions such as time synchronization for surely reflecting the final update in the file system are required, which makes it difficult to realize.

Therefore, in order to take a cluster configuration with the NAS system, a storage system as a cluster is simply configured to realize a large-capacity file system and improve performance by load balancing, or a SAN or the like. Place one or more servers as the front end of the cluster type disk array system,
Usually, the server implements the NAS function. Although it has not been implemented as a product yet, for example, in the white paper issued by BlueArc, multiple disk systems are placed in the back end of the latter, integrated management of disk systems and NAS by multiple servers. An overview of a system that provides functionality is disclosed.

[0009]

As described in the section of the prior art, in the NAS type disk array system, a plurality of systems are simply combined, or a configuration in which a server is placed as a front end of a SAN type system is used. It has been announced.

[0010] However, since the former system file system integrity control data is lost in the failure of any subsystem for placement divided into multiple disk subsystems, resiliency as a file system total Is significantly reduced. Of course, if the entire system is duplicated, the fault tolerance will be greatly improved, but the cost will more than double, and the communication cost of the management information between the duplicated systems for maintaining consistency is also large. In addition, the system configuration in which a server is placed has a large software overhead, and it is extremely difficult to construct a high-performance system in which the load on the network is large due to the communication between the servers for maintaining the file system consistency.

First, in a cluster type disk array device, one disk controller receives a data request from a server serving as a host, transfers it to the disk array device which actually has the data, and transfers the actual data. By reading and writing, a single image disk array that is not just a set of disk array devices is realized. However, due to cost and electrical restrictions, the data path for exchanging control information and data between disk array devices has to be slower than the internal path of the disk array, and as a result, a cluster type disk is used. In the array device, a difference in performance occurs depending on the position of the interface connected to the host. Especially when access to specific data (hotspots) is concentrated from multiple hosts, both the disk array device that forms part of the cluster and the coupled network between the clusters become a bottleneck. The original high performance cannot be obtained. Conventionally, in such a case, software is used to cope with the load by placing a copy of data or the like, but it is difficult to deal with such a usage in an Internet server where hot spots change rapidly. .

Another problem is that when a cluster is composed of a plurality of disk array devices, it is necessary to consider that one of the disk array devices in the configuration is out of order due to a failure or maintenance stop. It is difficult for the cluster type disk array device disclosed in the prior art to cope with such a case. Specifically, since all the data on the disks connected to the disk array devices that make up the cluster become inaccessible, it is difficult for the disk array device as a cluster to reduce the availability.

The present invention solves the above problems by arranging data in consideration of the cluster configuration, and in particular, provides a network-type storage device having a low-cost configuration that can operate without loss of availability in the event of a failure. The purpose is to provide.

[0014]

According to the present invention, in order to achieve the above object, a network and a plurality of storage devices connected to the network and using magnetic disks,
A client system connected to the network, file data provided to the client system and management data for controlling the same are stored in the storage device, and a file access function is provided to the client system using the management data. In the network-type storage device having means for providing, the management data and the file data are distributed and stored in the plurality of storage devices with different redundancy.

[0015]

BEST MODE FOR CARRYING OUT THE INVENTION According to the present invention, in a cluster type disk device which is a network type storage device, communication of management data for configuring a NAS by multiplexing only control data or adopting a redundant configuration by RAID. We have constructed a cluster-type NAS system with practical reliability while keeping the minimum.

By adopting such a configuration, although some of the failed data cannot be accessed, other data can be read and written normally, and the consistency of the file system is not lost.

Preferred embodiments of the present invention will be described below in more detail with reference to the drawings. FIG. 1 shows the configuration of a disk array device which is a network type storage device of the present invention, which is a premise of this embodiment. That is, the storage device of the present invention is
A network, a plurality of storage devices (composed of a disk array control device and a collective disk device described below) using magnetic disks connected to the network, a client system connected to the network, and the client system File data to be provided to the client and management data for controlling the file data are stored in the storage device, and the management data is used to provide a function of accessing the file to the client system. A network-type storage device in which the management data is distributed and stored in the plurality of storage devices with different redundancy.

In this embodiment, each disk array controller 11 corresponding to four network file systems,
An example of a configuration in which 12, 13, 14 are connected by a local area network by the IP network 5 is shown. The entire disk array control device has a control device (processor, buffer, etc.) 11 for processing a file system therein.
1, 121, 131, 141, and provides a function of controlling access to a file system configured by the entire disk (or a part thereof) connected to each disk array control device.

This is basically the same processing as the network file system, and is controlled so that the entire system is equivalently viewed as one file system. Each disk array control device has an aggregate disk device (that is, a plurality of storage devices) 11 including a plurality of disks.
6, 126, 136, 146 are combined, and the collective disk device 116 is connected to the disk array controller 11.
The aggregate disk device 126 is the disk array controller 12
, Which are combined in the following order. Further, there is a management server 30 coupled to all the disk array control devices 11, 12, 13, and 14, and settings and processing at the time of failure are carried out through this management server 30. A client system 40 using this disk array device group is connected to this disk array via a network, in this case an IP network. Since a plurality of clients essentially perform the same operation in this embodiment, one of them is
The operation when the disk array system of this embodiment is used from this client system will be described below.

Next, the structure of the file system used in this embodiment will be described. The internal data structure of the file system in this embodiment is shown in FIG. In this embodiment, a file system using a hierarchical structure is used. Even in such a file system, the file itself is just block data, and the file placed in the file system is managed by the management information separated from the file data which is the control data. The management information in this embodiment is stored in the inode data structure. 6 in this FIG.
1 is the inode data structure, and the actual data 6
The 2nd and so on are accessed by a direct pointer (for speeding up small data) and an indirect pointer (used for normal data). The other fields are management information and need to be updated consistently.
Further, two disk array device numbers 64 and 65 are added as characteristic information in this embodiment.

In this embodiment, this control data is controlled so that it is always managed and placed so as to have the same contents on the two disk array devices 64 and 65. In this embodiment, only this control data is duplicated and no redundant structure is provided for the data. The two disk array devices are indicated by the disk array device numbers 64 and 65 in FIG. In this embodiment, it is assumed that the IP address used for communication is stored in this field at the time of implementation.

The two disk array devices included in the inode information are set when the file system is created. In this embodiment, specifically, four disk array devices are used, and the inode number 0 is stored in the disk array devices 11 and 116 and 12 and 126.
The node number 1 is the disk array devices 12 and 126,
13 and 136, and the inode number 2 is the disk array devices 13 and 136 and 14 and 146.
Stored in the disk array devices 14 and 146 and 11 and 116, the inode number 3 is repeatedly set.

This is to measure the load distribution to some extent at the time of initial setting. Although actual data is accessed from this inode, this data is simply distributed and arranged on this disk array device. It may be arranged in any specific manner. In actual implementation, it is conceivable to specify it by maintenance personnel in consideration of load balancing or dynamically relocate. Further, in this embodiment, the disk array device numbers 64 and 65 are set at the time of this initialization. It is necessary to change this field at the same time when the process of dynamically changing the arrangement of control information is introduced as described above.

The operation of the disk array device having the above structure will be described below. First, when operating a file on the disk array device from the client 40, first, the management server 30 is inquired about the disk array control device to be inquired. In the configuration of this embodiment, the file system function is individually possessed by each disk array control device, and the result can be obtained by inquiring any individual disk array control device. The inquiry to the management server 30 is made because the management server 30 centrally manages the failure information of the disk array control device. Using the result, the client makes an inquiry to the disk array controller and obtains the result.

The disk array device number obtained here is cached by the client in order to reduce the number of inquiries to the management server 30 and improve the speed. When the client accesses the same inode again, the cached data directly accesses the disk array device having the inode from this data without inquiring the management server. Such cache management is general and can be easily implemented by those skilled in the art.

Regarding the multiplex management for embodying the redundancy of the control data, which is a feature of the present invention, in the present embodiment, the above i
The node information is controlled by the disk array control devices and updated synchronously. Specifically, with respect to the update from the client, two corresponding inodes are locked and exclusively updated from the disk array device receiving the update. The problem is the processing when one of the inodes becomes unreadable due to the occurrence of a failure. In this embodiment, the following processing is performed.

(1) When an error is found in the internal processing, each disk array control device informs the management server that it will be closed and closes it.

(2) Each client caches a box number having an inode when it tries to read data, but there is no response when it tries to read data from the respective disk array controller. For a predetermined time, the contents of the cache are purged, and the disk array controller that should read the inode in question is queried.

(3) In response to the above inquiry, the management server has the necessary inode and returns the operating disk array controller name. To obtain the disk array control device name, the disk array device number fields 64 and 65 in the inode are referenced.

(4) (1) to (3) are repeated until the response to the file request inquired from the client is completed.

On the contrary, when reconnecting the disk array control device which has been blocked due to a failure, the management server gradually spreads the load to the reconnected disk array device server without performing any processing. It is possible to recover by going on. Of course, it does not matter if the processing for reconnection is actively performed. However, as described above, the position of the predetermined inode is only cached, and if the replacement of the cache contents progresses with time, it approaches the original state. In addition, if such blockage and reconnection are repeated, load imbalance may occur in some cases. In that case, the management server intervenes and purges the cached data on the client. It is easy to add a modification such as the process of rebalancing the load to the above procedure.

In this embodiment, the speed is emphasized, and the configuration example in which the client actively detects the time-out and re-inquires to the management server 30 has been described. It is also possible to configure the management server 30 to reply with an appropriate disk array device for each transfer inquiry, and since the management server 30 is changed only to the management server in step (2) of the above procedure. It would be easy for a trader to change to this configuration. In this case, a high-speed management server 30 is required, but there is an advantage that the operating system on the client side need not be aware of the disk array configuration.

Further, in this embodiment, at the time of starting communication and at the time of failure, control is performed by intervention of the management server on the system (that is, when the client system is connected to the network by one or a plurality of management servers). As illustrated, a management server is not placed in a high-performance disk array, and a configuration is adopted in which failure processing is autonomously performed by communication between disk arrays (that is, when the client system is connected to the network by cooperative operation). You can also In this case, for example, when a disk array self-detects a failure, the processing is taken over to the paired disk array that has control data instead of the failed disk array, and the disk array that performs the alternative processing fails. It is possible to take over the processing transparently from the client by controlling the alternate response to the IP network address of the disk array that caused the error. In such a configuration, the client accesses an arbitrary disk array device or a disk array device designated in advance for each client, whereby communication is performed between the disk array devices and predetermined data is returned. Even in such a case, a highly available disk array device can be realized by adopting a configuration in which only the control data of the present invention is made redundant. The basic control is almost the same as that shown in this embodiment except the above description, and a person skilled in the art can easily realize such a disk array device.

In this embodiment, all the disk array control devices are connected to one local area network in FIG. 1, but the network is divided and the disk array control devices are separated by distance. It is possible to easily modify / expand the network configuration such as distributively arranging them in the network, and it is not necessary to make any changes in the control aspect. Of course, you can make changes to include network switches and routers. Further, although the present embodiment has been described in the context of a network using a widely used IP network, it does not utilize the characteristics of the IP network alone, and may be used when other network technologies are used. Can be similarly applied.

In the present embodiment, as already described, by duplicating only the control data, a network type disk array apparatus having a low cost and a high availability is realized as compared with the duplication of all. It is easy to change this to a configuration with higher redundancy, such as triple control data and double data, to increase the availability of data. By maintaining the redundancy higher than the redundancy, it is possible to realize a network type storage device, that is, a disk array device, whose availability can be greatly improved at a low cost.

[0036]

According to the present invention, it is possible to realize a network type storage device, that is, a disk array device, which provides a network type file system in which reliability and cost are optimized.

[Brief description of drawings]

FIG. 1 is a network type storage device which provides a network type file system according to a preferred embodiment of the present invention;
That is, an outline of a network-compatible cluster type disk array device is shown.

FIG. 2 shows a configuration of an inode which is control data in a preferred embodiment of the present invention.

[Explanation of symbols]

5 ... Network (IP network), 11, 12, 1
3, 14 ... Disk array control device, 30 ... Management server, 40 ... Client system, 61 ... Inode data structure, 62 ... Data, 64, 65 ... Disk array device, 111, 121, 131, 141 ... File control device, 112, 122, 132, 142 ... Control storage device, 113, 123, 133, 143 ... Disk control device, 114, 124, 134, 144 ... Cache, 1
16, 126, 136, 146 ... Collective disk device.

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI theme code (reference) G06F 12/16 320 G06F 12/16 320L

Claims (3)

[Claims] 1. A network, a plurality of storage devices connected to the network and using magnetic disks, a client system connected to the network, file data provided to the client system, and controlling the file data. Means for storing management data in the storage device and providing a file access function to the client system using the management data, the management data and the file data having different redundancy. A network type storage device characterized by being distributedly stored in. 2. The network type storage device according to claim 1, wherein the client system is a client system connected to a network by cooperative operation. 3. The network type storage device according to claim 1, wherein the client system is a client system connected to a network according to an instruction from one or a plurality of management servers.