支持IP-SAN远程映射与引导的固件协议栈模型的设计与实现
|
摘要
IP存储区域网(IP-SAN)是存储领域的一项重要技术,它同时具备IP网络的优势以及SAN架构的强大功能和扩展性。iSCSI协议是IP-SAN中实现存储设备访问的重要协议标准。以iSCSI协议为基础的远程启动技术,通过利用PCI ROM(如网卡)从远程的IP-SAN中启动,简化了基础架构,使得管理任务变得集中和简单。IP-SAN远程引导技术支持客户主机从远程SAN的映射磁盘上引导启动系统,就像访问本地的SCSI设备一样简单。那么如何将分配给某个登录用户的存储卷映射为该远程用户主机的本地硬盘?该映射硬盘上的操作系统如何引导?
本论文针对这些问题,提出了一种基于块的、面向可靠连接的iSCSI/TCP固件协议栈模型及设计实现方法,利用iSCSI协议对存储卷进行本地映射,实现了基于IP存储区域网络的远程引导。该技术的核心是运行在客户主机平台上网卡扩展ROM中的固件协议栈。研究的具体内容如下:
(1)提出了一种可移植的非路由传输层协议实现原型;简化设计了适用于嵌入式系统的传输层协议栈和TCP协议状态机;同时提出了几种简化算法来保证协议栈的性能与可靠性,包括乐观预测、固定回避重传算法和避免协议栈重入的输出请求挂起策略。
(2)在仔细分析iSCSI启动器体系结构的基础上,建立了一种便于操作系统内核iSCSI启动器平滑接管的固件iSCSI启动器有限状态模型;设计和实现了一个简洁、高效、稳定,且对存储资源要求很少的iSCSI启动器;最后设计了磁盘访问重定向服务和引导连接向量,给出了基于IP-SAN的存储卷远程映射引导的技术实现。
(3)对实现的iSCSI/TCP协议栈拟定了完整的功能、性能和可靠性测试方案;给出了基于该固件协议栈的IP-SAN远程启动方案与其他相关系统方案的比较;同时提出了进一步优化和改进系统性能的可行方案。通过将协议栈实现代码嵌入到X86裸机平台的64K PCI网卡扩展ROM中,实现了基于IP-SAN的网络引导。测试结果表明该固件协议栈有较好的性能和可靠性,并且使部署、移动和重新分配主机与存储卷之间的映射关系变得更加便利;真正实现网络启动和网络存储。
IP Storage Area Network (IP-SAN) technology is very important in the area of storage. It not only has the advantages of simple configuration and management, but also provides high capability and expansibility, which are the features of SAN. iSCSI is an critical protocol in the realization of accessing storage devices. Remote boot technology based on iSCSI protocal can boot from IP-SAN with the help of PCI ROM (such as NIC), it simplifies the infrastructure, and centralizes the the tasks of management. IP-SAN rebmote boot technology gives clients the ability to boot from mapping disks, which seems like every client has a local SCSI device, but how to allocate a mapping storage volume to a client, and how can they lauch the operating system on the mapping disk?
In order to solve these problems, a block-based, reliable iSCSI/TCP protocol stack model is designed and implemented in this paper. A new remote boot technology based on IP storage area network was designed, the core technology of it is the firmware protocol stacks running under bare machine platform. The details are as follows:
(1) A non-routing TCP protocol prototype was proposed and implemented based on firmware. The TCP protocol state machine is redesigned to make it simple. At the same time, several optimization algorithms were introduced to enhance the performance and reliability of remote boot, including optimistic forecast, fixed avoiding retransmit algorithm and output request suspend strategy that solved the problem of reentry.
(2) After carefully analysed the iSCSI initiator architecture, we designed a firmware iSCSI initiator Finite State Machine model which can be easily taken over by kernel mode initiator. On top of this, we designed and implemented a simplified iSCSI initator, which is efficient, stable and does not require many resources; finally we designed the service of Disk Access Redirection and Boot Connection Vector, shows the details of how to boot from IP-SAN.
(3) We fully tested the function, performance and reliability of the iSCSI/TCP protocol stack; then compared the IP-SAN remote boot system based on iSCSI/TCP firmware protocol stack with other relevant systems; After that, we suggested some further optimization and improvement options. The firmware protocol stack was integrated into IP-SAN extension BIOS, and applied to client hosts with no local storage devices, achieves network boot based on IP-SAN. Test results show it has a very good performance. Deployment, movement and re-distribution of relations between hosts and storages becomes more easily. Truely realized network boot and storage.
本论文针对这些问题,提出了一种基于块的、面向可靠连接的iSCSI/TCP固件协议栈模型及设计实现方法,利用iSCSI协议对存储卷进行本地映射,实现了基于IP存储区域网络的远程引导。该技术的核心是运行在客户主机平台上网卡扩展ROM中的固件协议栈。研究的具体内容如下:
(1)提出了一种可移植的非路由传输层协议实现原型;简化设计了适用于嵌入式系统的传输层协议栈和TCP协议状态机;同时提出了几种简化算法来保证协议栈的性能与可靠性,包括乐观预测、固定回避重传算法和避免协议栈重入的输出请求挂起策略。
(2)在仔细分析iSCSI启动器体系结构的基础上,建立了一种便于操作系统内核iSCSI启动器平滑接管的固件iSCSI启动器有限状态模型;设计和实现了一个简洁、高效、稳定,且对存储资源要求很少的iSCSI启动器;最后设计了磁盘访问重定向服务和引导连接向量,给出了基于IP-SAN的存储卷远程映射引导的技术实现。
(3)对实现的iSCSI/TCP协议栈拟定了完整的功能、性能和可靠性测试方案;给出了基于该固件协议栈的IP-SAN远程启动方案与其他相关系统方案的比较;同时提出了进一步优化和改进系统性能的可行方案。通过将协议栈实现代码嵌入到X86裸机平台的64K PCI网卡扩展ROM中,实现了基于IP-SAN的网络引导。测试结果表明该固件协议栈有较好的性能和可靠性,并且使部署、移动和重新分配主机与存储卷之间的映射关系变得更加便利;真正实现网络启动和网络存储。
IP Storage Area Network (IP-SAN) technology is very important in the area of storage. It not only has the advantages of simple configuration and management, but also provides high capability and expansibility, which are the features of SAN. iSCSI is an critical protocol in the realization of accessing storage devices. Remote boot technology based on iSCSI protocal can boot from IP-SAN with the help of PCI ROM (such as NIC), it simplifies the infrastructure, and centralizes the the tasks of management. IP-SAN rebmote boot technology gives clients the ability to boot from mapping disks, which seems like every client has a local SCSI device, but how to allocate a mapping storage volume to a client, and how can they lauch the operating system on the mapping disk?
In order to solve these problems, a block-based, reliable iSCSI/TCP protocol stack model is designed and implemented in this paper. A new remote boot technology based on IP storage area network was designed, the core technology of it is the firmware protocol stacks running under bare machine platform. The details are as follows:
(1) A non-routing TCP protocol prototype was proposed and implemented based on firmware. The TCP protocol state machine is redesigned to make it simple. At the same time, several optimization algorithms were introduced to enhance the performance and reliability of remote boot, including optimistic forecast, fixed avoiding retransmit algorithm and output request suspend strategy that solved the problem of reentry.
(2) After carefully analysed the iSCSI initiator architecture, we designed a firmware iSCSI initiator Finite State Machine model which can be easily taken over by kernel mode initiator. On top of this, we designed and implemented a simplified iSCSI initator, which is efficient, stable and does not require many resources; finally we designed the service of Disk Access Redirection and Boot Connection Vector, shows the details of how to boot from IP-SAN.
(3) We fully tested the function, performance and reliability of the iSCSI/TCP protocol stack; then compared the IP-SAN remote boot system based on iSCSI/TCP firmware protocol stack with other relevant systems; After that, we suggested some further optimization and improvement options. The firmware protocol stack was integrated into IP-SAN extension BIOS, and applied to client hosts with no local storage devices, achieves network boot based on IP-SAN. Test results show it has a very good performance. Deployment, movement and re-distribution of relations between hosts and storages becomes more easily. Truely realized network boot and storage.
引文
[1] Garth A, Gibson, Rodney Meter. Network Attached Storage Architecture. Communications of the ACM,2000,43(11):37-45
[2] Marc Farley,孙功星,蒋文保等译.SAN存储区域网络.北京:机械工业出版社, 2002,10-12
[3] Tom Clark.A Guide to iSCSI, iFCP, and FCIP Protocols for Storage Area Networks. Addsion-Wesley,2001,9-15
[4] SNIA.iSCSI Technical White Paper. http://www.snia.org/forums/ipsf/programs/ about/isci/iSCSI_Technical_whitepaper.PDF, 2001
[5] FCP, SCSI-3 Fibre Channel Protocol. ANSI X3,1996,269
[6] Benner A. Fibre Channel: Gigabit Communications and I/O for Computer Networks.McGraw Hill,New York,1996
[7] Prasenjit Sarkar.draft-ietf-ips-iscsi-boot-05.IBM,2002
[8] J. Satran,K.Meth.Internet Small Computer Systems Interface. http://www.ietf. org/rfc/ rfc3720.txt,2003
[9] White Paper. More Functionality, Less Infrastructure: Managed IT with netBoot/I. http://www.doubletake.com/,2008
[10]俞建新,杨小虎.网络存储新技术评析.计算机工程,2006,32(20):120-122
[11] Novell Inc. NetWare 4 feature guide. http://www.novell.com/documentation/lg/ nw42/pdfdoc/newftenu.pdf,1998
[12] Preboot Execution Environment (PXE) Specification. Version2.2, http://developer. intel. com/ial/wfm/wfmspecs.htm,1998
[13] RFC 1350.The TFTP Protocol (Revision 2). http://www.Faqs.org/rfcs/rfc 1350. html,1995
[14] Alexander S,Droms R.DHCP options and BOOTP vendor extensions.RFC2132, 1997
[15]杨华杰,张尧学.一种基于透明计算的远程启动协议MRBP2.小型微型计算机系统,2006,27(9):1658-1660
[16]周悦芝,张尧学.王勇.一种用于网络计算的可定制启动协议.软件学报,2003, 14(3):538-546
[17] Adam Dunkels.ulP-A Free Small TCP/IP Stack. http://www.sics.se/`adam/uip. 2002-0l-15
[18] Adam Dunkels.lwIP-A Ligntweight TCP/IP Stack. http://www.sics.se/`adam/lwip, 2001-04-18
[19] Cohen Amit,Cohen Reuven.A dynamic approach for efficient TCP buffer alloca- tion.In:Proc of the 7th International Conference on Computer Communications and Networks.MSST,1998,233-238
[20] Dave Simpson.Intel ships iSCSI host adapter. http://www.iscsistorage.com/ intelhba.htm,2002-03
[21] E.Yeh, H. Chao, V. Mannem, et al. Introduction to TCP/IP Offload Engine (TOE). 10Gigabit Ethernet Alliance,2002
[22] Droms R. Dynamic host configuration protocol. RFC 2131, 1997
[23] Todd R Weiss.Logical Volume Manager .Computer world,2003,37(11):228-232
[24] Compaq. Phoenix.Intel.BIOS Boot Specification,1999
[25] SCSI Architecture Model-2 (SAM-2), 2002, 9 (24)
[26] RFC3721.Internet Small Computer Systems Interface(iSCSI)Naming and Disco- very,2003,3-22
[27] B.Aboba, Joshua Tseng.IP Storage Security. http://www.ietf.org/intemet-drafts/ draft-ietf-ips-security-19.txt,2003-01-14
[28] S. Kent, R. Atkinson.Security Architecture for the Internet Protocol (IPSec). RFC2401, 1998,10
[29] RFC2945, The SRP Authentication and Key Exchange System, 2000,9
[30] RFC1994, PPP Challenge Handshake Authentication Protocol(CHAP),1996,8
[31] RFC1321, The MD5 Message-Digest Algorithm, 1992,4
[32] RFC3174, US Secure Hash Algorithm 1 (SHA1), 2000,9
[33] T Herbert.Embedding TCP/IP.http://www.embedded.com/2000/0001/0001ia1. htm, 2003-09
[34] RFC 1122, Requirements for Internet Hosts--Communication Layers. http://www. IETF.org, 1989
[35] RFC 793,Transmission Control Protocol,1981
[36] W.Richard Stevens. TCP/IP Illustrated Volume 1 The Protocols. Addison-Wesley, 1994,105-120
[37] Gauger C M,Kohn M,Gunreben S.Modeling and performance evaluation of iSCSI storage area networks over TCP/IP-based MAN and WAN networks. In: BroadNets. 2nd International Conference.Oakland,2005,850-858
[38]汤宏萍,王竹平.嵌入式TCP/IP协议栈的设计与实现.微电子学与计算机,2008, 25(6):168-170
[39]王力生,梅岩,曹南洋.轻量级嵌入式TCP/IP协议栈的设计.计算机工程, 2007, 33(2):246-248
[40] J. Chase. High Performance TCP/IP Networking.Prentice-Hall, 2003,221-255
[41] Girish Motwani, Gopinath K. Evaluation of advanced TCP stacks in the iSCSI environment using simulation model. In: Proc of the 13th IEEE NASA Goddard Conference on Mass Storage Systems and Technologies. MSST,2005, 210–217
[42] Garyr Wrightw, Stevens Richard.TCP/IP illustrated: The Implementation M. Boston: Addison Wesley, 1995,2,145-152
[43] Kancherla Bhargava K, Narayan Ganesh M, Gopinath K. Performance Evaluation of Multiple TCP connections in iSCSI. In:Proc of the 24th IEEE Conference on MassStorage Systems and Technologies. MSST, 2007,239-244
[44]游海峰,何泾沙,张伟.基于TCP重传计时器的算法分析.网络通讯与安全, 2007, 1(2):83-84
[45]葛志辉,陈志刚.一种新的基于RTT的丢包率估计算法.计算机工程与应用,2005, 41(19):26-27
[46] Abhijeet Joglekar,Michael Kounavis,Frank Berry. A Scalable and High Performance Software iSCSI Implementation. In:Proc of the 4th USENIX Conference on File and Storage Technologies.Oakland,2005,267-280
[47]郝克刚,丁剑洁.层次结构的Petri网.计算机科学与探索,2008,2(2):123-130
[48]陈大恒,闰立味,陈南等.基于Windows内核模式的iSCSI启动器设计.计算机工程与设计,2007,28(9):2161-2165
[49] Vishwakarma, Smita, Bagaria. iSCSI Simulation Study of Storage System. In: Tenth International Conference on Computer Modeling and Simulation,2008,703- 707
[50]李蕾,苏金树,张银福.iSCSI协议中错误恢复机制的设计与实现.计算机工程, 2006,32(2):119-121
[51] Intel.PnP BIOS Specification, 1999
[52] Lu Y, Du, D. H. C. Performance study of iSCSI-based storage subsystems. Communications Magazine, IEEE, 2003, 41(8):76-82
[53] Karn Phil. KA9Q TCP/IP Source Code. http://www.ka9q.net/code/ka9qnos, 2005-03-28
[54] Palekar A, Russell D. Design and Implementation of a SCSI Target for Storage Area Network.In:Proceedings of the 5th Annual Linux Showcase&Conference. Oakland,USA,2001,93-96
[55]许先斌,赵睿.一种改进的iSCSI存储系统中的缓冲策略.计算机工程与设计,2005,26(8):2207-2209
[56]余胜生,李志峰,周敬利.一种改善iSCSI存储系统性能的缓冲策略.计算机工程, 2004,30(12):73-74
[2] Marc Farley,孙功星,蒋文保等译.SAN存储区域网络.北京:机械工业出版社, 2002,10-12
[3] Tom Clark.A Guide to iSCSI, iFCP, and FCIP Protocols for Storage Area Networks. Addsion-Wesley,2001,9-15
[4] SNIA.iSCSI Technical White Paper. http://www.snia.org/forums/ipsf/programs/ about/isci/iSCSI_Technical_whitepaper.PDF, 2001
[5] FCP, SCSI-3 Fibre Channel Protocol. ANSI X3,1996,269
[6] Benner A. Fibre Channel: Gigabit Communications and I/O for Computer Networks.McGraw Hill,New York,1996
[7] Prasenjit Sarkar.draft-ietf-ips-iscsi-boot-05.IBM,2002
[8] J. Satran,K.Meth.Internet Small Computer Systems Interface. http://www.ietf. org/rfc/ rfc3720.txt,2003
[9] White Paper. More Functionality, Less Infrastructure: Managed IT with netBoot/I. http://www.doubletake.com/,2008
[10]俞建新,杨小虎.网络存储新技术评析.计算机工程,2006,32(20):120-122
[11] Novell Inc. NetWare 4 feature guide. http://www.novell.com/documentation/lg/ nw42/pdfdoc/newftenu.pdf,1998
[12] Preboot Execution Environment (PXE) Specification. Version2.2, http://developer. intel. com/ial/wfm/wfmspecs.htm,1998
[13] RFC 1350.The TFTP Protocol (Revision 2). http://www.Faqs.org/rfcs/rfc 1350. html,1995
[14] Alexander S,Droms R.DHCP options and BOOTP vendor extensions.RFC2132, 1997
[15]杨华杰,张尧学.一种基于透明计算的远程启动协议MRBP2.小型微型计算机系统,2006,27(9):1658-1660
[16]周悦芝,张尧学.王勇.一种用于网络计算的可定制启动协议.软件学报,2003, 14(3):538-546
[17] Adam Dunkels.ulP-A Free Small TCP/IP Stack. http://www.sics.se/`adam/uip. 2002-0l-15
[18] Adam Dunkels.lwIP-A Ligntweight TCP/IP Stack. http://www.sics.se/`adam/lwip, 2001-04-18
[19] Cohen Amit,Cohen Reuven.A dynamic approach for efficient TCP buffer alloca- tion.In:Proc of the 7th International Conference on Computer Communications and Networks.MSST,1998,233-238
[20] Dave Simpson.Intel ships iSCSI host adapter. http://www.iscsistorage.com/ intelhba.htm,2002-03
[21] E.Yeh, H. Chao, V. Mannem, et al. Introduction to TCP/IP Offload Engine (TOE). 10Gigabit Ethernet Alliance,2002
[22] Droms R. Dynamic host configuration protocol. RFC 2131, 1997
[23] Todd R Weiss.Logical Volume Manager .Computer world,2003,37(11):228-232
[24] Compaq. Phoenix.Intel.BIOS Boot Specification,1999
[25] SCSI Architecture Model-2 (SAM-2), 2002, 9 (24)
[26] RFC3721.Internet Small Computer Systems Interface(iSCSI)Naming and Disco- very,2003,3-22
[27] B.Aboba, Joshua Tseng.IP Storage Security. http://www.ietf.org/intemet-drafts/ draft-ietf-ips-security-19.txt,2003-01-14
[28] S. Kent, R. Atkinson.Security Architecture for the Internet Protocol (IPSec). RFC2401, 1998,10
[29] RFC2945, The SRP Authentication and Key Exchange System, 2000,9
[30] RFC1994, PPP Challenge Handshake Authentication Protocol(CHAP),1996,8
[31] RFC1321, The MD5 Message-Digest Algorithm, 1992,4
[32] RFC3174, US Secure Hash Algorithm 1 (SHA1), 2000,9
[33] T Herbert.Embedding TCP/IP.http://www.embedded.com/2000/0001/0001ia1. htm, 2003-09
[34] RFC 1122, Requirements for Internet Hosts--Communication Layers. http://www. IETF.org, 1989
[35] RFC 793,Transmission Control Protocol,1981
[36] W.Richard Stevens. TCP/IP Illustrated Volume 1 The Protocols. Addison-Wesley, 1994,105-120
[37] Gauger C M,Kohn M,Gunreben S.Modeling and performance evaluation of iSCSI storage area networks over TCP/IP-based MAN and WAN networks. In: BroadNets. 2nd International Conference.Oakland,2005,850-858
[38]汤宏萍,王竹平.嵌入式TCP/IP协议栈的设计与实现.微电子学与计算机,2008, 25(6):168-170
[39]王力生,梅岩,曹南洋.轻量级嵌入式TCP/IP协议栈的设计.计算机工程, 2007, 33(2):246-248
[40] J. Chase. High Performance TCP/IP Networking.Prentice-Hall, 2003,221-255
[41] Girish Motwani, Gopinath K. Evaluation of advanced TCP stacks in the iSCSI environment using simulation model. In: Proc of the 13th IEEE NASA Goddard Conference on Mass Storage Systems and Technologies. MSST,2005, 210–217
[42] Garyr Wrightw, Stevens Richard.TCP/IP illustrated: The Implementation M. Boston: Addison Wesley, 1995,2,145-152
[43] Kancherla Bhargava K, Narayan Ganesh M, Gopinath K. Performance Evaluation of Multiple TCP connections in iSCSI. In:Proc of the 24th IEEE Conference on MassStorage Systems and Technologies. MSST, 2007,239-244
[44]游海峰,何泾沙,张伟.基于TCP重传计时器的算法分析.网络通讯与安全, 2007, 1(2):83-84
[45]葛志辉,陈志刚.一种新的基于RTT的丢包率估计算法.计算机工程与应用,2005, 41(19):26-27
[46] Abhijeet Joglekar,Michael Kounavis,Frank Berry. A Scalable and High Performance Software iSCSI Implementation. In:Proc of the 4th USENIX Conference on File and Storage Technologies.Oakland,2005,267-280
[47]郝克刚,丁剑洁.层次结构的Petri网.计算机科学与探索,2008,2(2):123-130
[48]陈大恒,闰立味,陈南等.基于Windows内核模式的iSCSI启动器设计.计算机工程与设计,2007,28(9):2161-2165
[49] Vishwakarma, Smita, Bagaria. iSCSI Simulation Study of Storage System. In: Tenth International Conference on Computer Modeling and Simulation,2008,703- 707
[50]李蕾,苏金树,张银福.iSCSI协议中错误恢复机制的设计与实现.计算机工程, 2006,32(2):119-121
[51] Intel.PnP BIOS Specification, 1999
[52] Lu Y, Du, D. H. C. Performance study of iSCSI-based storage subsystems. Communications Magazine, IEEE, 2003, 41(8):76-82
[53] Karn Phil. KA9Q TCP/IP Source Code. http://www.ka9q.net/code/ka9qnos, 2005-03-28
[54] Palekar A, Russell D. Design and Implementation of a SCSI Target for Storage Area Network.In:Proceedings of the 5th Annual Linux Showcase&Conference. Oakland,USA,2001,93-96
[55]许先斌,赵睿.一种改进的iSCSI存储系统中的缓冲策略.计算机工程与设计,2005,26(8):2207-2209
[56]余胜生,李志峰,周敬利.一种改善iSCSI存储系统性能的缓冲策略.计算机工程, 2004,30(12):73-74