外文翻譯--和關(guān)于windows模式中高速存儲(chǔ)技術(shù)的研究

1、<p><b>　　附錄Ⅱ</b></p><p>　　----外文翻譯部分</p><p><b>　　原文：</b></p><p>　　Research of High-speed Data Storage Technique in Windows Mode</p><p>　　Ab

2、stract: In the application of modern industry measure and control fields,the transient state single should be incepted to computer for storage and processing real time. With increasing design target,it becomes more and m

3、ore important to the problems in the data storage processing and the ways to solve these problems in the data storage processing and the ways to solve these problems in Windows mode.</p><p>　　Keywords: Data

4、storage; transfers speed; WinDriver; ASPI; virtual disk.</p><p>　　1 Foreword</p><p>　　With the measure instrument precision and practice demand target increasing, the data transfer speed between

5、 instrument and computer improves, and most of the data should be processing real time. Common CPU operation speed has already exceeded 1 GHz, so it is easy for the operation ability of computer to satisfy the command of

6、 industry measure and control. However, the key point is the transfer bottleneck between the instrument and host computer. If the data cannot be sent to computer in time, t</p><p>　　2Memory Management </p

7、><p>　　One of the key concepts of memory managements is the correlation between the address space (logic address) and storage and storage space (physical address). When program is running, memory management uni

8、t (MMU) executes the mapping from logic address to physical address. The logic address is added by the value in relocation register before sent to memory, and the value in relocation register is transparent to user. It i

9、s obvious that in Windows mode, memory management is used the virtual address</p><p>　　Considering of the mapping affection on data transfer speed, operation physical address directly is the best choice. The

10、 operation step is as follow : first , send the data to the buffer area (physical memory) that allotted by the management soft; second, send the data to the buffer area (logic memory) that allotted by the management soft

11、; third, send the data from logic memory to storage disks. Both physical address and logic address should be continuous. The reasons are listed as follow: first</p><p>　　In the fist step, data acquisition ca

12、rd should obtain the process of transfer data ,and the physical address must be continuous. The buffer area’s size lies on the data’s size transferred once. And in the third step, since suing ASPI interface technique (it

13、 will be explained in fourth part), the logic address should be continuous, too. If the buffer areas in first and third step can be united into one buffer area, the processing is only concluding two steps: fist, send the

14、 data to continuous buf</p><p>　　To avoid the storage operation effect on the data read, the ping-pong mode is accepted in the storage processing . The allotted buffer area is divided into two same parts. Wh

15、en the first buffer area is full of data , the system sends an interrupt signal to management soft , and transfers the data to hard disk.. The following data will data will be sent to the next buffer area. When this buff

16、er area is full of data, the system sends an inteeeupt signal to management soft, too. The precondition of</p><p>　　3. Method on Allotting</p><p>　　Continuous Buffre Area </p><p>　　

17、There are several methods on allotting continuous area in Windows mode, and two of these are compared here .</p><p>　　One is using VxD (Virtual Device Driver) to do this job. It can allot one continuous buff

18、er area and provide the head address of this buffer area to management program and mapping form this buffer area to virtual address space. In the pro gram, first hold one buffer area in the virtual address space, then tr

19、ansfer VxD with sending the head address of the virtual address space to VxD, so that VxD allot one buffer area in the physical memory and obtain the hcad address. At last, refer the physical</p><p>　　Anothe

20、r is using the function of card’s DMA operation in WinDriver: WD_DMALock. By using this function, the continuous buffer area can be obtained for DMA operation area. WinDriver is only used to drive cards, so it requires a

21、t least one card in the acquisition system. But it is common for most acquisition system than the first way, and its dependability is higher, too. Moreover, it is convenience to upgrade. So it is a good choice for indust

22、ry measure fields.</p><p>　　The buffer area is allotted by this way: setting the reasonable parameter of the function WD_DMALock to obtain the continuous physical page , and the corresponding logic address i

23、s continuous, too. The head address of first physical page is the physical memory address .The return value of puseraddr , one of the parameter of WD_DMALock, is the physical memory. In addition ,winDriver collates every

24、 acquisition card by logic mumber (from l to the maximum ,the logic number is not corresponding to ph</p><p>　　One detail should be paid much attention : one of parameter of this function must be set to “DMA

25、_KERNEL_BUFFER_ALLOC”, so that WinDriver can allot a continuous physical memory space. Moreover, the process of allotting buffer area must be placed in the initialization phase in the program, and other program should no

26、t run before the memory management soft runs. Otherwise, the process of allotting memory may be failure for the system has not enough continuous physical memory space. When the process</p><p>　　Data storage

27、buffer area 2</p><p><b>　　(1MB)</b></p><p>　　Data storage buffer area 1</p><p><b>　　(1MB)</b></p><p>　　Fig.1 Allotted Physical Memory Model</

28、p><p>　　4. Storage in Hard Disk </p><p>　　Hard disk operation has two levels: Windows and logic sector level. Accessing disk in Windows level , the operation system will do most of the job. The ope

29、ration is so simple that the Windows98/2000 operation system can call API function directly to access the disk. However, the speed of this access is too low to satisfy the data storage’s requirement. Accessing disk in lo

30、gic sector level , the soft developed by user self can operate disk directly in logic sector. This method gains higher speed </p><p>　　To improve the data storage speed , system adopts SCSI hard disk instead

31、 of IDE hard disk . In the forepart of SCSI, users develop the SCSI device’s driver. The job is very complex and often makes mistakes, and the driver’s portability is not well. ASPI (Advanced SCSI Program Interface) tech

32、nique solves this problem. Problem. ASPI is developed by Adaptec Company, and defined in most operation system, including Windows system, The process of accessing SCSI hard disk by ASPI includes four steps co</p>

33、<p>　　The concrete method about ASPI will not be explained here, which can be found in many books about SCSI. In the process of data storage, the time of searching track and sector is reduced after the disk rotate s

34、peed keeps normal. But a detail should be paid much attention. If the magnetic head is far form the sector where to write data when the first time to write data, the magnetic head will spend more time to write to seek tr

35、ick and sector. Moreover, if the disk is at dormancy state (disk rotate</p><p>　　5. Virtual Disk </p><p>　　Generally, the disk operation speed is lower than the memory operation speed of 4 to 6

36、order. So it can improve the storage speed to use memory as disk, this way is also called RAM disk. RAM disk’s device driver allows the normal disk operation, though the operation is executed in memory instead of disk. S

37、o it is transparent for the user. This way is not a good choice for the high dependability system, because it has many disadvantages such as the memory capability should be enough, the data in </p><p>　　6. P

38、ostscript</p><p>　　All the above items are the key points in high-speed data storage technique. In addition, there are some other points also as RAID (Redundant Arrays of Inexpensive Disks), disk dispatch al

39、gorithm, etc .The affection range and the cost of the system are all different. So the choices are different. too. “Four Channels PXI Bus Preprocessing Dynamic Data Acquisition System” is design for the aero engine’s tra

40、nsient and dynamic state, wind tunnel examination. This task belongs to the subject: the res</p><p><b>　　Reference</b></p><p>　　[1]Chen xiangqun XiangYong, WangLei MaHongbing, Zhen

41、gKougen Dave Probert. Windows Operation System Principle </p><p>　　[2]CaoZiyuan. Physical Memory’s Allocation and Direct Access in Win32 Mode,Observe and Control Technique, Vol.20,No.1,Jan2001.</p>&l

42、t;p>　　[3]Brian Sawert. The Programmer’s Guide To SCSI, China Electric Power Press,2001</p><p>　　[4]Walter Oney. Programming the Microsofy Windows Driver Model, Microsoft Press,1999</p><p>　　[

43、5]ZhangHao. The Archive of PHD2000 Data Acquisition System’s Sofr, Sep, 2001</p><p>　　Author Biographies</p><p>　　Zhang Lei: work in the First Aeronaut Institute of Air Foree, instructor, gradua

44、te student of Beijing University of Aeronautics&Astronautics, engaged in data acquisition system development:</p><p>　　Xu Xing: work in Beijing University of Aeronautics&Astronautics ,School of Instr

45、umentation Science & Opto-electronics Engineering, professor, tutor of graduate student, engaged in computer measure and control technique research.</p><p><b>　　翻譯：</b></p><p>　　

46、和關(guān)于Windows模式中高速存儲(chǔ)技術(shù)的研究</p><p>　　摘要：在現(xiàn)代工業(yè)測(cè)量和技術(shù)領(lǐng)域的應(yīng)用中，單一的短暫狀態(tài)輸入計(jì)算機(jī)存儲(chǔ)，并卻實(shí)時(shí)處理。隨著設(shè)計(jì)目標(biāo)的增加，重要數(shù)據(jù)的存儲(chǔ)速度變的越來(lái)越重要。本文主要講的是數(shù)據(jù)存儲(chǔ)過(guò)程中的問(wèn)題和在Windows模式中解決這些問(wèn)題的方式。</p><p>　　關(guān)鍵詞：數(shù)據(jù)存儲(chǔ) 轉(zhuǎn)換速度 Windows ASPI 虛擬磁場(chǎng)</p>

47、<p><b>　　1 .展望</b></p><p>　　隨著測(cè)量工具的精確和實(shí)際需求目標(biāo)的曾加，數(shù)據(jù)轉(zhuǎn)換速度在測(cè)量?jī)x器和計(jì)算之間改進(jìn)，并且大部分?jǐn)?shù)據(jù)都可以實(shí)時(shí)處理。普通CPU的運(yùn)行速度已經(jīng)超過(guò)1G赫茲。對(duì)于計(jì)算機(jī)的運(yùn)行能力來(lái)講，需滿足工業(yè)測(cè)量?jī)x器和控制的要求是很容易的：然而主要的問(wèn)題在于測(cè)量?jī)x器與主機(jī)同的轉(zhuǎn)換障礙。如果數(shù)據(jù)沒(méi)有及時(shí)地傳送到計(jì)算機(jī)，那么計(jì)算機(jī)的操作能力就不能有效地

48、被運(yùn)用，這個(gè)問(wèn)題將按如下步驟進(jìn)行探討：第一步應(yīng)先確定操作系統(tǒng)的類別最流行的操作系統(tǒng)是Windows系統(tǒng)。因此，最重要的數(shù)據(jù)轉(zhuǎn)換技術(shù)將在Windows模式中討論。</p><p><b>　　2.存儲(chǔ)器管理</b></p><p>　　存儲(chǔ)器管理的一個(gè)主要的概念是在“位置空間”（邏輯地址）和“存儲(chǔ)空間”（物理地址）間的相互關(guān)系。當(dāng)程序運(yùn)行時(shí)，存儲(chǔ)器管理單元執(zhí)行從邏輯地址

49、到物理地址間的制作工作。</p><p>　　在傳到存儲(chǔ)器前，邏輯地址就是通過(guò)重新注冊(cè)的價(jià)值得到補(bǔ)充，并且重新登記的的價(jià)值對(duì)用在虛擬地址中使用，而非物理地址。所以這個(gè)程序存到的是虛擬地址而非物理地址。</p><p>　　考慮到數(shù)據(jù)轉(zhuǎn)換速度中制圖因素的影響，操作物理地址顯然是最好的選擇，操作步驟如下：第一：將數(shù)據(jù)傳到由管理軟件分配的緩沖區(qū)域（物理區(qū)域）；第二，將數(shù)據(jù)傳到由管理軟件分配的緩沖

50、區(qū)域（邏輯區(qū)域）；第三，將數(shù)據(jù)從邏輯存儲(chǔ)器傳送到存儲(chǔ)磁盤上，但是物理地址和邏輯地址是無(wú)法連續(xù)的，原因如下：第一，必須保證向上述那樣分配存儲(chǔ)器，并且下一步可以顯示出來(lái)；第二，相關(guān)的分配幾個(gè)緩沖區(qū)域，分配一個(gè)緩沖區(qū)域并不能將零碎的片段添加的記憶中去。第三，分配到這種緩沖區(qū)域可以預(yù)先處理此程序。</p><p>　　在第一步中，數(shù)據(jù)獲得下可以獲得數(shù)據(jù)轉(zhuǎn)換過(guò)程中，數(shù)據(jù)存儲(chǔ)分配區(qū)域的物理地址，并且物理地址必須是連續(xù)的。緩沖

51、區(qū)域的大小在于一次性轉(zhuǎn)換數(shù)據(jù)的長(zhǎng)度。并在有第三步中，因?yàn)槭褂昧薃SPI界面技術(shù)（將在第四部分中介紹），邏輯地址也應(yīng)是連續(xù)的，如果第一階段和第三階段的緩沖區(qū)域可以統(tǒng)一在一個(gè)緩沖區(qū)域，那么這個(gè)過(guò)程只包括兩步：第一，將數(shù)據(jù)傳送到連續(xù)的緩沖區(qū)域；第二，將數(shù)據(jù)傳送到數(shù)據(jù)磁盤，通過(guò)這種方式，數(shù)據(jù)轉(zhuǎn)移到存儲(chǔ)器系統(tǒng)的時(shí)間將縮短，且數(shù)據(jù)存儲(chǔ)速度將增加。</p><p>　　為了避免存儲(chǔ)操作技術(shù)影響數(shù)據(jù)只讀功能，PING-PONG模

52、式有存儲(chǔ)過(guò)程中被接受，分配緩沖區(qū)域被劃分為兩個(gè)相同的部分。第一個(gè)緩沖區(qū)域充滿數(shù)據(jù)，系統(tǒng)傳送一個(gè)中斷信號(hào)到管理軟件，并且將數(shù)據(jù)轉(zhuǎn)到硬盤上，剩下的數(shù)據(jù)被傳到下個(gè)緩沖區(qū)域，當(dāng)這個(gè)緩沖區(qū)域充滿數(shù)據(jù)，系統(tǒng)也會(huì)向管理軟件傳送中斷信號(hào)，這個(gè)模式的前提是：數(shù)據(jù)的存儲(chǔ)速度應(yīng)高于獲得速度。WIDEULTRA2 SCSI磁盤的轉(zhuǎn)換速度可以達(dá)到80MB/S,且數(shù)據(jù)的寬度是16位,可以滿足這個(gè)前提.</p><p>　　3.分配連續(xù)緩沖區(qū)

53、域的方式</p><p>　　在WINDOWS模式下,有多種分配連續(xù)緩沖區(qū)域的方式,這里將比較其中的兩種.</p><p>　　一種是利用VXD(虛擬裝置驅(qū)動(dòng)器)來(lái)做此工作,它可以分散一些連續(xù)的緩沖區(qū)域,切給管理程序提供這個(gè)緩沖區(qū)域的首地址,為從這個(gè)緩沖區(qū)域到虛擬地址提供制圖,在此程序中,首先在虛擬地址中選擇一個(gè)緩沖區(qū)域,然后將附有首地址的虛擬地址的VXD傳送到VXD,只有這樣,VXD分散

54、一個(gè)在物理存儲(chǔ)器與已確定的虛擬地址聯(lián)系起來(lái),這種方法可以在WIN32中使用,它可以直接收到一個(gè)連續(xù)的物理緩沖區(qū)域和邏輯緩沖區(qū)域.但是,這個(gè)模式的執(zhí)行比較復(fù)雜.</p><p>　　另一種是是用WinDreir中卡的DMA運(yùn)行的功能:WD-DMALOCK.通過(guò)使用其功能,對(duì)DMA來(lái)說(shuō),這個(gè)連續(xù)的緩沖區(qū)域能夠獲得,這個(gè)緩沖區(qū)域可以作為數(shù)據(jù)存儲(chǔ)緩沖區(qū)域使用. WinDreir只是被用來(lái)驅(qū)動(dòng)卡,因此,在獲得系統(tǒng)中要求最少

55、一張卡.但是對(duì)大多時(shí)獲得系統(tǒng)來(lái)講,使用數(shù)據(jù)獲得卡是很普通的一件事,這種方法較第一種來(lái)講是比較簡(jiǎn)單的,且它的依賴性也更高,更多的是,對(duì)上層來(lái)講是簡(jiǎn)便的,所以對(duì)工業(yè)測(cè)量是個(gè)很好的選擇.</p><p>　　緩沖區(qū)域是按此方式分散的,為了獲得一個(gè)連續(xù)的物理頁(yè)碼,設(shè)定一個(gè)合理的WD-DMALOCK功能的參數(shù),切相應(yīng)的邏輯地址也是連續(xù)的,首個(gè)物理頁(yè)碼的首地址是一個(gè)物理存儲(chǔ)器地址.PuserAddr的回服價(jià)值, WD-DMA

56、LOCK的一個(gè)參數(shù),是物理存儲(chǔ)器的地址.另外WinDreir通過(guò)邏輯數(shù)字(從1到無(wú)窮大,邏輯數(shù)字與物理順序不是對(duì)應(yīng)的)分散獲得卡,在設(shè)有運(yùn)行獲得卡或這些卡的任何影響的情況下,這項(xiàng)功能就可運(yùn)行.因此邏輯數(shù)字在此項(xiàng)功能中被使用設(shè)為1,且對(duì)所有的獲得卡承擔(dān)責(zé)任.</p><p>　　有一個(gè)細(xì)節(jié),我們應(yīng)給予更多的關(guān)注:次功能的一個(gè)參數(shù)一定要被轉(zhuǎn)送到DMA-KERNEL-BUFFER-AUOC中.這樣, WinDreir可

57、以分散一個(gè)連續(xù)的物理記憶空間,而且分散緩沖區(qū)域的過(guò)程應(yīng)放在程序初始化階段.且其它的程序不應(yīng)在存儲(chǔ)器管理軟件運(yùn)行前運(yùn)行.否則,由于系統(tǒng)沒(méi)有足夠的連續(xù)的物理存儲(chǔ)空間分散存儲(chǔ)器過(guò)程可能會(huì)失敗.當(dāng)完成此過(guò)程時(shí),通過(guò)WD-DMALOCK的功能,分散記憶一定要釋放,這個(gè)表格展示了通過(guò)PING-PONG模式(每個(gè)緩沖區(qū)域的大小是1MB)分散物理存儲(chǔ)器的模型.</p><p>　　Data storage buffer area

58、 2</p><p><b>　　(1MB)</b></p><p>　　Data storage buffer area 1</p><p><b>　　(1MB)</b></p><p>　　Fig.1 Allotted Physical Memory Model</p><

59、p><b>　　4.硬盤存儲(chǔ)</b></p><p>　　硬盤操作有兩個(gè)層次:WINSOWS和邏輯階段水平.在WINDOWS模式的存儲(chǔ)磁盤,運(yùn)行系統(tǒng)會(huì)做大部分工作.如此簡(jiǎn)單的操作,比較于WINDOWS98/2000操作系統(tǒng)可以稱為可以直接進(jìn)入磁盤的API功能,然而,進(jìn)入的速度太低以至于不能滿足數(shù)據(jù)存儲(chǔ)的需要,在邏輯階段的存儲(chǔ)磁盤.在沒(méi)有使用操作系統(tǒng)中的文件系統(tǒng)的情況下,這種方式存儲(chǔ)磁盤

60、的速度需比WINDOWS模式更高.如果數(shù)據(jù)被存儲(chǔ)在連續(xù)的邏輯階段.存儲(chǔ)數(shù)據(jù)的速度會(huì)更高.因此,相比在WINDOWS模式下,在邏輯階段且連續(xù)的情況下,存儲(chǔ)磁盤會(huì)更好.</p><p>　　為了提高數(shù)據(jù)存儲(chǔ)速度,系統(tǒng)采用SCSI磁盤取代IDE硬盤.在SCSI前期,用戶自己改進(jìn)了SCSI驅(qū)動(dòng)器.此工作很復(fù)雜且容易出錯(cuò),驅(qū)動(dòng)器也不便捷.ASPI(先進(jìn)的SCSI程序界面)技術(shù)解決了此問(wèn)題.ASPI是由AOLOPTEC公司改

61、進(jìn)的,并且愛(ài)大多數(shù)的操作系統(tǒng)中給予解釋.包括WINDOWS系統(tǒng).通過(guò)ASPI存儲(chǔ)SCSI磁盤的過(guò)程一般分為四步,構(gòu)建存儲(chǔ)磁盤要求;將要求轉(zhuǎn)送給ASPI管理者;等待執(zhí)行要求;解釋由ASPI管理者反饋回來(lái)的錯(cuò)誤信號(hào).</p><p>　　關(guān)于ASPI的具體方法這里將不做介紹,它可以在很多有關(guān)SCSI的書中找到.在數(shù)據(jù)存儲(chǔ)過(guò)程中,在磁盤速度轉(zhuǎn)換正常后,搜索階段的時(shí)間被減少,但是一個(gè)細(xì)節(jié)我們應(yīng)更加注意.如果磁頭處于何時(shí)讀

62、寫數(shù)據(jù)何處讀寫階段,那么磁頭就需要花費(fèi)更多的時(shí)間去尋找技巧和狀態(tài)階段.此外,如果磁盤處于非激活狀態(tài)(磁盤旋轉(zhuǎn)速度為0),那么磁盤從非激活狀態(tài)到運(yùn)行狀態(tài)將消耗大量時(shí)間.而且,數(shù)據(jù)從獲得下到分散緩沖區(qū)域的轉(zhuǎn)化速度需保持一致.這些情況可能發(fā)生在一些數(shù)據(jù)丟失的情況下,因?yàn)閿?shù)據(jù)的存儲(chǔ)速度要比數(shù)據(jù)從獲得卡到存儲(chǔ)器中的的轉(zhuǎn)換速度一定要保持運(yùn)行狀態(tài).當(dāng)數(shù)據(jù)存儲(chǔ)數(shù)據(jù)過(guò)程開始時(shí),每個(gè)磁頭都應(yīng)轉(zhuǎn)移到cylinoler,數(shù)據(jù)將被傳送那,通過(guò)軟件操作磁盤磁頭比較

63、困難,并且這種方法將傷害磁盤.所以應(yīng)該尋找一個(gè)更好的方法,而不是直接的轉(zhuǎn)換磁頭,一個(gè)有效的方法就是在數(shù)據(jù)轉(zhuǎn)移到存儲(chǔ)器的這個(gè)階段讀取數(shù)據(jù),然后在同寫到階段狀態(tài)中.通過(guò)這種方式磁頭可以轉(zhuǎn)移到cylinoler中,并保持運(yùn)行狀態(tài).</p><p><b>　　5.虛擬吸盤</b></p><p>　　一般地磁盤的操作速度要比存儲(chǔ)器操作速度低一4-6個(gè)指令,因此我們可以通過(guò)改

64、變存儲(chǔ)速度來(lái)像使用磁盤一樣使用存儲(chǔ)器.這種方式也被稱做RAM磁盤.RAM磁盤的裝置的驅(qū)動(dòng)器允許普通磁盤的操作,盡管操作技術(shù)是在存儲(chǔ)器中執(zhí)行而非磁盤中,所以,對(duì)用戶來(lái)說(shuō)它是透明的.這種方式對(duì)高依賴性的系統(tǒng)來(lái)說(shuō)并不是一個(gè)恰當(dāng)?shù)倪x擇,因?yàn)樗性S多諸如存儲(chǔ)器容量應(yīng)該足夠大的缺點(diǎn),如果系統(tǒng)或電源遇到了一些意外,虛擬磁盤中的數(shù)據(jù)就會(huì)丟失.</p><p><b>　　6.附筆</b></p>

65、<p>　　所有的以上陳述在高速數(shù)據(jù)存儲(chǔ)技術(shù)中都是關(guān)鍵,此外,有很多可以影響數(shù)據(jù)存儲(chǔ)速度的其他因素.如RAID(廉價(jià)的磁盤的多余信息),磁盤信息算法等.所有因素的影響力都不同,因?yàn)檫\(yùn)用程度和系統(tǒng)的成本都不同,所以選擇也不同.“四種渠道PXI汽車平行能量數(shù)據(jù)獲得系統(tǒng)”是因?yàn)轱w行引擎的短暫停留.能量狀態(tài)和風(fēng)向檢測(cè)而設(shè)計(jì)的,這個(gè)任務(wù)屬于此項(xiàng)目:飛行自動(dòng)測(cè)量和檢測(cè)搜索技術(shù),在系統(tǒng)軟件的發(fā)展過(guò)程中,數(shù)據(jù)存儲(chǔ)技術(shù)已經(jīng)提供了強(qiáng)有力的支持