KR20030044464A - A distributed controlling apparatus of a RAID system - Google Patents

Abstract

PURPOSE: A distributed control device of a RAID(Redundant Array Independent Disks) system is provided to improve a performance in proportion to the number of controllers by supporting a fault allowance of at least two RAID controllers, processing a controller in parallel, and selecting a dual PCI bus structure in the controller. CONSTITUTION: A plurality of disc groups(290,291,292,293) store data using a fiber channel as a matching unit for sharing the sate. A plurality of distributed sharing RAID control units(240,250,260,270) share the disc groups(290,291,292,293) and control a RAID. A plurality of fiber channel switches(280,281) connect the disc groups(290,291,292,293) to the RAID control units(240,250,260,270). At least one host computer(200) reads and writes data finally using the distributed sharing RAID control units(240,250,260,270) as the same storing medium. A network switch(230) is a communication network for connecting the host computer(200) and the distributed sharing RAID control units(240,250,260,270). A distributed locking unit is connected to the RAID control units(240,250,260,270) for supplying the RAID control units(240,250,260,270) as the same data medium. A diagnosis client(231) diagnoses and sets the RAID control units(240,250,260,270) through an Ethernet.

Description

A distributed controlling apparatus of a RAID system

The present invention relates to a structure of a multi-host computer and a redundant array of independent disks (RAID), each controller through a network for high speed disk sharing and control period sharing control through a fiber channel switch The present invention relates to a distributed control apparatus of a RAID system for operating with the same RAID medium.

RAID system improves I / O characteristics by bringing together multiple independent disk storage devices into one to distribute data to each disk and to allow access to multiple disks at the same time. By making them look like a disk, they enable the implementation of mass storage devices. In addition, by simultaneously accepting auxiliary data such as disk duplication, Error Checking and Correction (ECC) codes, or parity data, you can automatically recover data in the event of a disk failure in any of the RAID systems. It can increase the availability of the system.

RAID systems are divided into various levels according to their configuration. According to a paper published by Patterson et al., RAID systems are divided into five types, that is, RAID levels 1 to 5.

First, RAID level 1 stores N disk data, stores data on another N disks, and copies and stores the same data on another mirror disk. When writing data, data should always be stored on two disks, and during reading, data can be read by selecting the faster disk among the two disks. If one disk fails, service can continue with the other disk. RAID level 2 uses Hamming code to protect data, which costs less disk than RAID level 1 mirroring.

RAID level 3 consists of adding one parity disk to a group of N data disks. When writing data, data is distributed in bits or bytes, and each data is stored on the disk, and the parity disk stores parity calculated with an exclusive logical sum (E-OR) for the data stored on the data disk. When reading, N disks must be accessed at the same time. If one disk fails, information can be recovered using the parity data stored in the parity disk.

Like RAID level 3, RAID level 4 consists of N + 1 disks. N disks store data, and the other disk stores parity. The difference from RAID level 3 is that data is distributed in blocks. Therefore, when writing data, access of one data disk and parity disk is required, and only one disk needs to be accessed when reading. In the event of one disk failure, information can be recovered using the parity stored in the parity disk.

Similar to RAID level 4, RAID level 5 stores data in blocks, but does not store and store parity on one disk.

Reading and writing data and recovering data in the event of a single disk failure are the same as for RAID level 4.

In addition, there is RAID level 6, which uses RAID level 0 and Reed-Solomon code that does not use auxiliary data, but simply distributes the data to accommodate P + Q error recovery.

RAID level 6 is more available to the system than using parity because information can be recovered even if two disks fail at the same time. Most RAID systems currently support RAID levels 0, 1, 3, 5, and RAID levels 0/1, which are a combination of RAID levels 0 and 1, and use a type suitable for the user's application environment.

If an array controller fails in such a RAID system, a fatal event may occur that causes not only the RAID system but also the entire system to stop. Therefore, in a system where availability of the system is important, a controller of a RAID system is dually configured and two controllers share a disk, so that even if one controller fails, the other controller can replace the failed controller. .

Hereinafter, a brief description will be made of a RAID system according to the prior art.

FIG. 1 is a diagram illustrating a configuration example of a general connection method between a RAID having two general controllers and host computers.

The first and second controllers 130 and 140 connected to the host computer 110, two fiber channels 150 and 151 and disk groups 160 for connecting disks.

Each RAID controller 130 or 140 is a two disk matching module 132, 133 or 142, 143, which is connected to a shared Just Bunch of Disks (JBOD) 160. The host bus adapters 101 and 111 of the host computer 110 are connected to the host matching modules 131 and 141 of the controllers 130 and 140 via the two fiber channels 120 and 121, respectively, to store data to the host. To provide.

The disks 160 to which the fiber channel is connected are divided into two logical groups by initial RAID system configuration, and each group belongs to one controller 130 or 140.

Therefore, when both array controllers 130 and 140 operate normally, each controller controls only the corresponding disk group. If one controller 130 or 140 fails, the other controller 130 or 140 ) Takes over control of the disk group belonging to the failed array controller 140 or 130.

However, in the existing RAID system, two controllers are used only for terminal error recovery and do not improve the performance of one storage medium. That is, there is a problem that the storage speed is limited to the performance of a single controller and the performance of a single fiber channel 120 or 121 connected to a host. In particular, there is a problem in that the controller performance is much restricted.

Accordingly, the present invention has been made to solve the above-described problems according to the prior art, an object of the present invention, while supporting the fault tolerance of two or more RAID controllers to improve performance through parallel processing of the controller, It is to provide a distributed control device of RAID system that improves performance in proportion to the number of controllers by adopting a dual PCI bus structure inside.

1 is a diagram showing a configuration example of a general connection method between a host computer and a RAID having two general controllers.

2 is a diagram illustrating an embodiment of a distributed sharing structure between a RAID controller and a host computer serving as the same storage medium through a shared disk group and a control period communication unit according to the present invention.

3 is a diagram illustrating a detailed configuration of a distributed shared RAID controller according to the present invention.

* Description of the symbols for the main parts of the drawings *

200: host computer 201: host bus adapter

230: network switch 240: RAID switch

241: host matching module 242: Ethernet matching module

243, 244: first and second disk matching module 245: distributed lock control module

280: Fiber Channel switch 290: Disk group (JBOD)

320: central processing unit 330: PCI bus

331: I / O chipset

341 Parity Engine and Mass Cache Controller

361: expanded memory

According to an aspect of a distributed control apparatus of a RAID system according to the present invention for achieving the above object, in the connection device of a distributed shared RAID system, a plurality of data storing the data using the Fiber Channel (matching) as a matching means Disk group; A plurality of distributed shared RAID control means for sharing the plurality of disk groups as storage means to perform RAID control; A plurality of fiber channel switch devices which are communication networks connecting the plurality of disk groups and the plurality of RAID control means; At least one host computer which is a subject that performs final reading and writing of data using the plurality of distributed shared RAID control means as the same storage medium; A network switch which is a communication network connecting the host computer and the plurality of distributed shared RAID control means; Distributed lock means connected with said plurality of RAID control means for control for providing said plurality of RAID control means as a same data medium; It may include a diagnostic management client responsible for diagnosing and setting the RAID control means via Ethernet.

The distributed shared RAID control means may include a central processing unit in charge of RAID control and a host matching means for matching with a host computer; First and second disk matching means connected to two fiber channels of the disk group; Distributed lock control means for sharing the disk among the RAID control means; An Ethernet matching module and a first PCI bus bridge connected with a client for diagnosis and configuration management of the RAID control means; A parity engine and a large-capacity cache control means for improving the input / output speed of the RAID control means, increasing the memory capacity, and calculating the RAID parity; An input / output chipset connecting all of the above means; A high speed first PCI bus connecting the input / output chipset, the first bus bridge, the distributed lock control means, and the second disk matching module; A second PCI bus bridge connecting the first disk matching module, the host matching module, the parity engine, the large-capacity cache control means, and the input / output chipset; A main memory of the central processing unit connected to the expansion memory and the input / output chipset of the parity engine and the large-capacity cache control means; A battery power supply that preserves the contents of the extended memory and the main memory even when the power is cut off; A flash memory for storing a RAID control program; A real time clock and nonvolatile memory for providing time information and storing nonvolatile data; RS-232C serial communication means for serial communication with the RAID control means.

The network switch includes at least one of an infiniband switch, a fiber channel switch, and an ethernet switch, which are high-speed communication networks, and the disk group includes a Fiber Channel Arbitrated Loop (FC-AL) that supports a fiber channel disk supporting the fiber channel. It can be composed of).

In addition, the distributed lock control unit, at least one of the SCI, Fiber Channel, Infiniband, Ethernet as a matching means, to form a dedicated connection network between the RAID control unit to maintain the consistency of data, the parity engine and large-capacity cache control unit Has a memory matching unit connected to the PCI bus and the expansion memory connected to the PCI bus, and may perform an exclusive OR of the data of the expansion memory, and performs data transmission and reception between the PCI bus and the expansion memory.

The host matching unit uses at least one of InfiniBand, Fiber Channel, and Ethernet as a matching means, and performs data transmission and reception with the host computer.

Hereinafter, with reference to the accompanying drawings a preferred embodiment of the distributed control device of the raid system according to the present invention will be described in detail.

FIG. 2 is a diagram illustrating an embodiment of a distributed sharing structure between a RAID controller and a host computer serving as the same storage medium through the shared disk group and the control period communication unit according to the present invention.

As shown in FIG. 2, a host computer 200 having one host bus adapter (HBA), a host computer 210 having two bus adapters 211, 212 or three or more bus adapters. Two or more RAID controllers 240 that provide the same storage area through a network switch 230, which may be an InfiniBand or Fiber Channel or Ethernet, with a host computer 220 (not shown). It is connected to host matching modules 241, 251, and 261 of 250, 260, and 270.

The RAID controllers 240, 250, 260, and 270 all provide the host computers 200, 210, and 220 as the same disk area. In other words, the data provided by the first RAID controller 240 provides the same data in the second or third RAID controllers 250 and 260. The opposite is also true.

One or more disk families 290, 291, 292 and each RAID controller 240 through dual Fiber Channel switches 280, 281 so that all RAID controllers 240, 250, 260, 270 provide the same RAID storage space. The two disk matching modules 243, 244, 253, 254, 263, and 264 of, 250, 260, and 270 are connected to two fiber channel switches 280 and 281, respectively. The first disk matching module 243, 253, and 263 and the second disk matching module 244, 254, and 264 correspond to the first disk matching module 342 and the second disk matching module 336 of FIG. 3, respectively. .

The first disk matching modules 243, 253, and 263 and the second disk matching modules 244, 254, and 264 are connected to the two fiber channel switches 280 and 281, respectively, to operate even when a failure occurs in any one channel. It is designed to improve performance during normal operation. The first fiber channel switch 280 may include the first disk matching modules 243, 253, and 263 of the RAID controllers 240, 250, 260, and 270 and the first disk channels of the disk groups 290, 291, and 292. 282, 284, and 286, and the second Fiber Channel switch 281 connects the second disk matching modules 244, 254, and 264 of the RAID controllers 240, 250, and 260 to the disk groups 290, 291, and the like. The second disk channels 283, 285, and 287 of 292 are connected. Here, in the disk group 290, 291, and 292, a fiber channel disk supporting a fiber channel may be configured with a Fiber Channel Arbitrated Loop (FC-AL).

Since the RAID controllers 240, 250, 260, and 270 store some data in the internal memory for improving performance, that is, the data is written to the memory and not yet written to the disk because of the cache (dirty data: Dirty Data). ) Becomes inconsistent with data between the RAID controllers 240, 250, 260, and 270, so that the dedicated channels 245, 255, and 265 between the RAID controllers 240, 250, 260, and 270 are connected to each other to process dirty data. Used for locking These communication channels 245, 255, 265 correspond to the distributed lock control means 355 of FIG. 3.

Each of the RAID controllers 240, 250, 260, and 270 includes Ethernet matching modules 242, 252, and 262, and uses the RAID controllers 240, 250, 260, and 270 as a diagnostic client 231 through the Internet. Diagnostic management is provided.

The RAID controller will be described in detail with reference to FIG. 3.

3 is a diagram illustrating a detailed configuration of a distributed shared RAID controller according to the present invention.

Internal PCI bus 340 and secondary PCI bus 330 to increase the data transfer speed, the primary PCI bus 340 has a host matching module 343 and the first disk mainly used for general RAID services Of the input / output chipset 331 connected to the matching module 342 and the parity engine and the large-capacity cache controller 341 to support the RAID parity operation and the large-capacity cache memory, the central processing unit 320, and the secondary PCI bus 330. The primary PCI matcher 344 is connected.

The host matching module 343 is connected to the host computers 200, 210, and 220 illustrated in FIG. 2, used as an Infiniband or Fiber Channel, or an Ethernet, and reads data from the host computers 200, 210, and 220. Receive a write command.

The first disk matching module 342 is connected to the disk groups 290, 291, 292, and 293 through the first fiber channel switch 280, and receives read and write commands received from the host computers 200, 210, and 220. It processes and reads and writes data to disk.

The parity engine and the large-capacity cache controller 341 have a synchronous dynamic random access memory (SDRAM) as an expansion memory 361, and transmit and receive data between the main PCI bus 340 and the expansion memory 361. In addition to the main memory 362 connected to the input / output chipset 331, the expansion memory 361 does not lose data even when the power is cut off through the battery power source 360.

The secondary PCI bus 330 includes a second PCI matching module 336 and a second PCI matching module 336 and the first PCI bus bridge 334 and the I / O chipset 331 connected to the distributed lock control unit 335 and the Ethernet matching module 333. It is connected to the unit 336.

The Ethernet matching module 333 is used by the diagnostic client 231 to diagnose and manage the RAID controller via the Internet.

The distributed lock control means 335 is used for a control channel dedicated communication channel for maintaining the consistency of the dirty data, and may be a fiber channel, an infini band, a scalable coherent interface (SCI), or an ethernet. The second disk matching module 336 is connected to the second fiber channel switch 281 to perform the same function as the first disk matching module 342.

If a failure occurs in the first and second disk matching modules 336 and 342, the normal module is transferred to the function of the failed module and performs this.

The primary PCI bus 340 and the secondary PCI bus 330 are connected through a first PCI bus bridge 334.

The present invention uses the two buses as described above and has up to twice the input / output performance. The above PCI bus refers to 66 MHz 64-bit PCI or 133 MHz 64-bit PCI-X.

The input / output chipset 331 performs serial communication with the main memory 362 and the flash memory 302 for storing the program in the flash memory bus 310, the real time clock and the nonvolatile memory 301, and the computer required for executing the remaining programs. The serial communication RS-232C 300 is connected.

The distributed control apparatus of the RAID system according to the present invention described above is not limited to the above-described embodiments and drawings, and various substitutions, modifications, and changes are possible within the scope without departing from the technical idea of the present invention. Those of ordinary skill in the art will appreciate.

In the distributed control apparatus of the RAID system according to the present invention as described above, multiple shared distributed controllers sharing multiple disk groups each provide the same storage area to the host computer to overcome the limitations of the performance of the single controller and the bandwidth of the channel. Therefore, it has the effect of linearly increasing the data storage performance according to the number of controllers and improving the fault tolerance of the control period.

Claims (8)

In a distributed shared RAID system, A plurality of disk groups for storing data for sharing data using a fiber channel as a matching means; A plurality of distributed shared RAID controllers which share the plurality of disk groups and perform RAID control; The communication network connecting the plurality of disk groups and the plurality of RAID controllers includes a plurality of fiber channel switching units; One or more host computers that are the subjects of the final reading and writing of data using the plurality of distributed shared RAID controllers as the same storage medium; A network switch which is a communication network connecting the host computer and the plurality of distributed shared RAID controllers; Distributed locking means connected to the plurality of RAID control units for controlling the plurality of RAID control units to be provided as the same data medium; Distributed control device for a RAID system comprising a diagnostic management client responsible for the diagnosis and setting of the RAID control means via Ethernet. The method of claim 1, The RAID control unit, A central processing unit in charge of RAID control; A host matching unit matching with the host computer; First and second disk matching portions connected to two fiber channels of the disk group; A distributed lock controller for sharing the disk between the RAID controllers; An Ethernet matching module and a first PCI bus bridge connected to the diagnostic management client for diagnosis and configuration management of a RAID controller; A parity engine and a large-capacity cache controller for improving the input / output speed of the RAID controller, increasing memory capacity, and calculating RAID parity; An input / output chipset for connecting all the devices; A high speed first PCI bus connecting the input / output chipset, the first bus bridge, distributed lock control means, and a second disk matching module; A distributed control apparatus of a RAID system including a first disk matching module, a host matching module, a parity engine, a second cache bus control means, and a second PCI bus bridge connecting the input / output chipset. The method according to claim 1 or 2, The RAID control unit, A main memory of the central processing unit connected to the expansion memory and the input / output chipset of the parity engine and the large-capacity cache control means; A battery power supply that preserves the contents of the extended memory and the main memory even when the power is cut off; A flash memory for storing a RAID control program; A real time clock and nonvolatile memory for providing time information and storing nonvolatile data; Distributed control device for a RAID system further comprising an RS-232C serial communication unit for serial communication with the RAID control means. The method of claim 1, The network switch, Distributed control device of a RAID system including at least one of the high-speed communication network Infiniband switch, Fiber Channel switch, Ethernet switch. The method of claim 1, The disk group, The distributed control device of a RAID system in which the Fiber Channel disk supporting the Fiber Channel is configured by Fiber Channel Arbitrated Loop (FC-AL). The method of claim 2, The distributed lock control unit, At least one of the SCI, Fiber Channel, Infiniband, Ethernet as a matching means, Distributed control device of a RAID system to maintain the consistency of data by forming a dedicated connection network between the RAID control unit. The method of claim 2, The parity engine and large capacity cache control unit, A PCI matching circuit connected to the PCI bus and a memory matching circuit connected to the expansion memory, and an exclusive logical matching of the data of the expansion memory, and the distribution of the RAID system for performing data transmission and reception between the PCI bus and the expansion memory. controller. The method of claim 2, The host matching unit, A distributed control apparatus of a RAID system using at least one of InfiniBand, Fiber Channel, and Ethernet as a matching means, and performing data transmission and reception with the host computer.