CN101043445B - IO dispatch method in network storage system - Google Patents
IO dispatch method in network storage system Download PDFInfo
- Publication number
- CN101043445B CN101043445B CN2007100718360A CN200710071836A CN101043445B CN 101043445 B CN101043445 B CN 101043445B CN 2007100718360 A CN2007100718360 A CN 2007100718360A CN 200710071836 A CN200710071836 A CN 200710071836A CN 101043445 B CN101043445 B CN 101043445B
- Authority
- CN
- China
- Prior art keywords
- order
- command
- read
- iscsi
- write
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- Computer And Data Communications (AREA)
Abstract
The provided IO scheduling method in network storage system comprises: 1. when a piece of new IO entering the transmitting queue, deciding whether this command is IO read-write response command or short command for data flow control command; 2. if yes, going to step 3, or else to step 4; 3. adding this command as queue head, and returning to step 1; and 4. adding this command to queue end, and returning to step 1. This invention transmits the read-write command or short command preferentially to improve whole system performance and the resource utility.
Description
(1) technical field
The present invention relates to network storage field, be specifically related to storage area network (Storage AreaNetwork-SAN) field in the Network storage technology.
(2) background technology
Network storage technology uses the ripe network interconnection to replace the IO bus to connect main frame and memory device.In network connects, often to use switch.Use optical fiber switch for fiber optic network, and for Ethernet, promptly the iSCSI storage system is then generally used common Ethernet switch.The switch that uses in network store system is not at present generally used at the network storage and is optimized especially.For example, for the iSCSI storage system, generally use common Ethernet switch, switch is for situation the unknown of upper layer application, so can not be optimized at the network storage.In electronic letters, vol, " the short bag priority scheduling algorithm of preemptive type in the input queue " in the fourth phase in 2005 proposed the short bag priority scheduling optimized Algorithm for ordinary router, but is not the optimization method of using at the network storage.
(3) summary of the invention:
The object of the present invention is to provide the IO dispatching method in the network store system of a kind of performance that improves network store system and resource utilization.
The object of the present invention is achieved like this: it comprises the attainable step of following computer:
1) a new IO order will enter transmit queue, judge whether this order is IO read-write response command or the short command that belongs to the data flow con-trol order;
2) if this order is IO read-write response command or the short command that belongs to the data flow con-trol order, then arrive step 3), if not, then arrive step 4);
3) this order is added to the formation head, turned back to step 1);
4) add this order to the formation afterbody, turn back to step 1).
The invention is characterized in: it is a kind of a kind of optimization method of using at the network storage, and it preferentially transmits read-write response command in entire I path or the IO order only transmitted or the short command that belongs to the data flow con-trol order in switch.
The entire I path comprises main frame in the Network storage technology, storage networking and memory device.In read-write requests, memory device is waited for the read-write requests order of transmitting from main frame, if can preferentially transmit these request commands, then can reduce the stand-by period of memory device, improves the utilance of memory device.After system has finished desired operation, also to wait for a response command usually, just can be released to and finish the resource that read-write requests is applied for.If can preferentially transmit these response commands, then system can discharge resource as soon as possible, and resource is offered follow-up read-write requests, thereby has improved resource utilization.Simultaneously, because these orders all are not contain the parcel of storing data, they are preferentially transmitted to improve system throughput.
Analogue test shows: improve the dispatching priority of reading and writing the response command that finishes and can significantly improve storage system resource utilization and performance on the entire I path.The present invention proposes to use at the network storage, and read-write response command in the IO order of transmitting in entire I path or the switch or the short command that belongs to data flow con-trol are preferentially transmitted, and reaches the raising performance of storage system and improves the purpose of resource utilization.
(4) description of drawings
Fig. 1 is an iSCSI read order flow chart in the embodiment of the invention;
Fig. 2 is an iSCSI write order flow chart in the embodiment of the invention;
Fig. 3 is the sketch of an IO path in the network store system.
(5) embodiment:
The present invention is further illustrated below in conjunction with the drawings and specific embodiments:
In network store system, be example with the iSCSI storage system, its read-write process is as follows:
Read command:
When the start end user program when destination end equipment sends read request:
1. this request is converted into scsi command and be delivered to the iSCSI bottom layer driving (Low Level Driver, LLD).ISCSILLD receives orders the back to destination end transmission iSCSI " SCSI Command " PDU.
2. when destination end front end target drives (Front End Target Driver, FETD) receive this order after, its decapsulation is reduced to scsi command.
3. this scsi command is passed to the scsi target intermediate layer (SCSI Target Middle Level, STML).
4. STML is returned to start end to the read data in the buffering area, i.e. iSCSI " SCSI Data In " PDUs.After the iSCSI LLD of start end receives read data, it is kept in the buffer empty of having distributed.After the total data end of transmission, STML gives start end, i.e. iSCSI " SCSI Respond " PDU with response back.
5. after iSCSI LLD received this response, (SCSI Middle Level SML) handled to give the SCSI middle level with it.ISCSI LLD and STML are released to whole resources that this order distributes.
In conjunction with Fig. 1, the WRITE order:
When the start end user program when destination end equipment sends write request:
1. this request is converted into scsi command and is delivered to iSCSI LLD.ISCSI LLD sends iSCSI " SCSI Command " PDU to destination end.
2. after destination end FETD receives this order, its decapsulation is reduced to scsi command.
3. this scsi command is passed to STML.
4. STML distributes needed buffering area, sends iSCSI " Ready to Transfer " PDU to start end, and the notice originating end can begin to send data.
In case 5. iSCSI LLD receives this PDU, just the write data that leaves in the buffering area is sent to destination end, i.e. iSCSI " SCSI Data Out " PDUs.
6. FETD receives the write data that transmits from start end, and it is kept in the buffer empty of having distributed.Then, notice STML write data is received.
7. after the total data end of transmission, STML gives start end, i.e. iSCSI " SCSI Response " PDU the response back after handling.
8. after iSCSI LLD received this response, (SCSI Middle Level SML) handled to give the SCSI middle level with it.ISCSILLD and STML are released to whole resources that this order distributes.
In conjunction with Fig. 2, from the read-write process of iSCSI agreement as can be seen, the iSCSI PDU that transmits between start end and the destination end has two kinds: a kind of for comprising the order that reads and writes data, i.e. iSCSI " SCSI Data In " PDUs and iSCSI " SCSI Data Out " PDUs; Another kind is to wait for relevant order with resource, as request command iSCSI " SCSI Command " PDU, response command iSCSI " Ready to Transfer " PDU and iSCSI " SCSI Response " PDU.If preferentially transmit these short commands, then can reduce the stand-by period of resource, improve resource utilization.For example destination end all will send iSCSI " SCSI Response " PDU order to start end after read-write finishes, and has only start end to receive that after this order, start end and destination end can be released to all resources that this read-write operation distributes.
For each read-write operation, start end and destination end all can have reciprocal process many times, and in these message transmitting procedures, and the great majority that transmit in the IO path are finish order or belong to the short command of data flow con-trol order of request read-write and read-write.Existing document proves, can improve performance if improve the dispatching priority of short bag in router.The present invention proposes the short command in entire I path or the IO order only transmitted in switch is preferentially transmitted, to improve performance and resource utilization.
It among Fig. 3 the sketch of an IO path in the network store system, the IO path comprises the device 3 that the network of the IO order of the device 2 of handling the IO order in the network delivery device 1, switch of the IO order of starter end and object machine end sends, and wherein each IO order was all waited in line in waiting list before being handled by them.4,5,6 is respectively starter, the waiting list in switch and the object machine.
In conjunction with Fig. 3, it is all the same to handle the processing procedure of carrying out in the device 3 that the network of the IO order of the device 2 of IO order or object machine end sends in the network delivery device 1 of the IO of starter end order, switch, i.e. the taking out the IO order from team's head of transmit queue 4,5 or 6 respectively and send on the network of circulation.Its idiographic flow is as follows:
(1) in transmit queue, takes out order.
(2) this order is sent on the correct network.
(3) to step (1).
Starter, switch or the object machine handling process to transmit queue 4,5,6 respectively are as follows:
(1) new IO order will enter transmit queue.
(2) judge whether this order is IO read-write response command, as request command iSCSI " SCSI Command " PDU, response command iSCSI " Ready to Transfer " PDU and iSCSI " SCSI Response " PDU, if then arrive step (3), if not, then arrive step (4).
(3) this order is added to the formation head, to step (1).
(4) add this order to the formation afterbody, to step (1).
Claims (1)
1. the IO dispatching method in the network store system, it is characterized in that: read-write response command in entire I path or the IO that only transmits in the switch order or the short command that belongs to the data flow con-trol order are preferentially transmitted, comprise the attainable step of following computer:
1), a new IO order will enter transmit queue, judge that whether this order is that IO reads and writes response command or belongs to the short command of data flow con-trol order;
2) if this order is IO read-write response command or the short command that belongs to the data flow con-trol order, then arrive step 3), if not, then arrive step 4);
3), this order is added to the formation head, turn back to step 1);
4), add this order to the formation afterbody, turn back to step 1).
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2007100718360A CN101043445B (en) | 2007-03-06 | 2007-03-06 | IO dispatch method in network storage system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2007100718360A CN101043445B (en) | 2007-03-06 | 2007-03-06 | IO dispatch method in network storage system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101043445A CN101043445A (en) | 2007-09-26 |
CN101043445B true CN101043445B (en) | 2011-02-23 |
Family
ID=38808657
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2007100718360A Expired - Fee Related CN101043445B (en) | 2007-03-06 | 2007-03-06 | IO dispatch method in network storage system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101043445B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101694610B (en) * | 2009-10-16 | 2011-11-09 | 成都市华为赛门铁克科技有限公司 | Command processing method, device and memory device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1550993A (en) * | 2003-04-16 | 2004-12-01 | ƽ | Read priority caching system and method |
CN1773475A (en) * | 2004-11-12 | 2006-05-17 | 国际商业机器公司 | An arbitration structure and a method for handling a plurality of memory commands |
CN101000589A (en) * | 2006-12-22 | 2007-07-18 | 清华大学 | Adaptive external storage IO performance optimization method |
-
2007
- 2007-03-06 CN CN2007100718360A patent/CN101043445B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1550993A (en) * | 2003-04-16 | 2004-12-01 | ƽ | Read priority caching system and method |
CN1773475A (en) * | 2004-11-12 | 2006-05-17 | 国际商业机器公司 | An arbitration structure and a method for handling a plurality of memory commands |
CN101000589A (en) * | 2006-12-22 | 2007-07-18 | 清华大学 | Adaptive external storage IO performance optimization method |
Also Published As
Publication number | Publication date |
---|---|
CN101043445A (en) | 2007-09-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7761642B2 (en) | Serial advanced technology attachment (SATA) and serial attached small computer system interface (SCSI) (SAS) bridging | |
US7272687B2 (en) | Cache redundancy for LSI raid controllers | |
CN102065568B (en) | MAC (media Access control) software and hardware interaction method based on data descriptor and hardware implementation device thereof | |
CN100440184C (en) | DMA controller and transmit method capable of simultaneously carrying out read-write operation | |
US9858959B2 (en) | Adaptively mounting and unmounting removable storage media based on monitoring requests and states of storage drives and the storage media | |
US8924610B1 (en) | SAS/SATA store-and-forward buffering for serial-attached-SCSI (SAS) storage network | |
CN109697122A (en) | Task processing method, equipment and computer storage medium | |
CN102236543B (en) | Data decompression device and method | |
CN107153511A (en) | Memory node, mixing memory controller and the method for controlling mixing memory group | |
CN102023961A (en) | Communication method based on shared memory and apparatus thereof | |
CN101873475A (en) | Control command sending method, data transmission method, monitoring system and device | |
CN1331070C (en) | Method and equipment of data communication | |
CN101656658A (en) | Method and device for improving dequeuing efficiency in queue management | |
CN107294865B (en) | load balancing method of software switch and software switch | |
CN102843435A (en) | Access and response method and access and response system of storing medium in cluster system | |
US8078771B2 (en) | Sending large command descriptor block (CDB) structures in serial attached SCSI (SAS) controller | |
CN101043445B (en) | IO dispatch method in network storage system | |
CN101437033A (en) | Method and network appliance for supporting variable velocity | |
US20060271639A1 (en) | Multipath control device and system | |
CN104714832A (en) | Buffer management method used for airborne data network asynchronous data interaction area | |
CN108958903A (en) | Embedded multi-core central processing unit method for scheduling task and device | |
US10581748B2 (en) | Information processing apparatus, information processing method, and non-transitory computer-readable storage medium | |
CN106980474B (en) | Data recorder based on PCIE interface | |
US20140280716A1 (en) | Direct push operations and gather operations | |
US8117419B2 (en) | Storage apparatus and method for eliminating redundant data storage using storage apparatus |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110223 Termination date: 20170306 |