200912728 九、發明說明： 【發明所屬之技術領域】 本發明係^於具有複數儲存單位的大量儲存解 二具體而言，本發明具體例係關於在儲存裝置^資‘ 資料控制資訊之分裂方法和奘w 什展置上貝枓和 【先前技術】 Μ 財之高速雜記錄，會涉及 谱妹才4率。使用習知磁碟機，像㈣心 =徑上有限度延遲。已知系統:足此Pi;;身 控制路徑功能在較高位階分開 =疋將貝科和 理、維護和/或使用者界面二2統把資料處 理很多控制資訊。惟已知策略仍需徑盡量處 j完ί在資料路徑内實施’便是“制路;ί分;址:f签： 制處理器設立來往於儲存址f間内之資料’以控 I/O操作。 师裝置，和來/往於通訊網路之所有 一種2_2219號旨在揭示 統包含三統連資料路徑。儲存系 在資料路徑和控制路’資料服務平台位 和控制路徑。資料服矛;平二控：器位於資料路徑 為界面，含控制路和β,機界面，與儲存器應用主機 路和段，人右次4cL Γ又連口於服務處理器。平台包含資料 路=束，只有利用資料以二 段，含有資料路徑功有控制界面。控制器包含資料路徑 200912728 機建築H國=5,8G2,366號旨在揭示—種槽案伺候 與局部理等t位加从本身之處理器， 在槽案控dt4理’ 層。虛賴齡系統是 部窖藏器檑冗其本身局 (可記憶體 副系統，以及資=^!5’5=39()遽旨在揭示—種資料儲存 接出控制器’連接至-或以上的直 發出資料轉移命令，發動=在^^理系統對副系統 體内轉移之資料’以及資料要送入的主機記憶 特定“閱二i!!儲存副系統之裝^^制器’能夠重； 誤，眘二d4轉移。此外，如在閱讀操作中發生錯 的主的裝置控，可特定更換資料應送入 a 士it:6知從資料路徑分離控制路徑之方法，需尊的尤 =有，建築（必然、為複雜的資料路 虑 在因為”構化不谷許有良好程度的預估能力，由於 此ί^；ίΐΐ相關課題時可能發生估計最壞情況之延遲。、 此舉導致硬體資源之無效利用。 、逖 故巫需有改進之祕和方法，枝在高速轉移系統中從 200912728 控制路徑分離資料路徑。 【發明内容】 料館存裝置列於申請專利範11第1項。資料健 = 料路徑接受的資料加 陣列控制器接受的輸出控制她 裳置另外包括控制路徑，適於對流路驅 本發明，資到儲存裝置陣列上之位址。按照 此適於只為簇集轉移;門=動， 法包法規定於申請專利範圍第6項。本方 至;器内緩衝之資料輸送 一部份加以關聯。 制貝訊與從陣列控制器接受的資料 陣列it器ί:: 括i相當於控制資訊之處 ，把從 資訊可“ 上2m;之陣列。控制 裝置陣列上之位址。從《列t更普遍關係到儲存 車列，做為—序歹,=集接4到的資料，可轉移到 例說；圖佳具體例只是舉 發能修錦。本 圖為以彔糸統之方塊圖，具有系統控制器模组，按 200912728 照本發明具體例，在儲存單位陣列上進行儲存功能。第j圖 所示記錄系統概以100指示。記錄系統丨⑻包含使用者界面 102，讓使用者控制記錄系統1〇〇之整體操作，並觀看有關系 統狀態等等之資訊。在本發明一具體例中，使用者界面包 LCD觸墊顯示器。 记錄系統100包含系統控制器模組⑴4。系統控制 組H)4適於轉移資料至HDD陣列114，並由此接受資料^ 陣列包括複數個別HDD。資料簇集轉移來往於HDD陣列 ^14之磁碟^是利用設定或銘記適當數值於暫存器内，表示 簇集大小、簇集開始位址和相關命令，像「閱讀」或「書 寫」。 ^ 系統控制器模組104包含内嵌式軟體處理器系統，第i =斤為 PPC 106。PPC 是 P〇wer pc 的縮寫。ppc 1〇6 經由 Μ控制路徑116與外部模組通訊。由ppc 1〇6進行的 2 ’τϋ含f行槽案系統維護演算，與外部委託機通^， ΐί 觸或其他使用者界面或υι，控制使用者出入。 例如儲存器的邏輯定址和實體定址間之界 ί理憶體部署、垃圾收集、解斷篇、版本 i例之ίίίΐ 如下述所規定，最好進行本發明具 體例^貝祕理，不用檔⑽、統/定址任務之常務。 模衡徑118，構成纽控制器 麄哭10S 在第圖所不具體例中，PPC 106構成空 I" ： ； ;t: r^t:n:5 流路驅動If 112提餘制資1 雜繼徑118，對 在第1祕目2 吏控制資訊與資料有關聯。 路徑120接受的;。。，告藏-1〇8適於轉移經由資料 資料，在轉移到RAID #制’^由^路^接受的即時 衝。RATD扯座丨控制0之前，於窖藏器1〇8内缓 二，再把資料輸送至流路驅動器模組112，提 200912728 供資料流送至HDD陣列114内所附裝置。凡專家均 統1〇〇可另外適應流送來自HDD陣列114之資料。°己錄系 RAID控制器11〇把進入的資料流多工化。把進入 料k加以多工化’是把―體進人㈣料流分成複數並遠冷 ^過程’各旨在書寫於儲存|置陣列内之不同儲存。= 家均知RAID控制器n〇亦可適於把出去的資料 資料S要增加冗贅’從二以下故障腦回收資料，而無 流路驅動賴組112把資料流逐字分裂成n塊 ^相對應DMA弓丨擎，服務所附標準ΑΤΑ-HDD。專家均知， ，具體例中可採用按照任何規約作# κ是可採用規約之二例。定址ΐΐ (ί 儲存裝置·"之麵他）亦可麵路驅動器 伽明具體例提供有效率綠和裝置，把高速資料流分 裝置上。「晚裝紮」指就在資料要書寫於 扇：J/1Λ賢料與控制資訊，諸如儲存裝置上所欲磁 ‘二位址資訊，加以關聯。此外，内部控制路徑 蝱哭ϋ接提蚊址資訊給流路轉11额112，而不給窖 次控制110。構想是在很低位階，把進入/ 情、、兄：亩、ΐί裝置上之指定簇集關聯，亦即在最後控制之 單—儲料位或儲存裝置通訊。以此方式， 而士，控制常務中，達成高速資料記錄績效。具體 士/^if制資訊之晚關聯’簡化記錄系》統應之設 前，不ίϊ=在相關資料剛要書寫於_陣列…之 士，明具體例進一步開發DMA之已知原理。具體而 ° MA-般容許處理!I在段資料轉移中，卸除每單一記憶 200912728 行特殊I性戶;ί;:課題。此係藉提供專用硬體，進 發明具體例提址範圍之情形。此外，本 如此DMA曰格2更體，甚至進行設置課題，監督並影響 題。 a格系列。此進-步卸除PPC 106的蠻力例行課 題，乎完讀除㈣職12G進行控制課 t 預估能力。結果是更佳利用硬體資 理器。 斷心或放棄冗贅組件，能夠使用低功率控制處 以參態\用來說明本發明具體例之操作。狀態圖 訊之触本發明賤例，触#料和控制資 系統操!^之前’第2圖内以p°wer pc或ppc標註之 = I*在空轉狀態202。PPC發動轉移，經由暫存辱 移的簇集大錢值，要用的磁扇開始位址2 Ϊ 是為了 _操作或書寫操作。在本發明一具體例 = 範圍可達每裝置16 。此等數值 j 規約（像12C)提供。暫存器在第2圖内^ 读s备Iff」。暫存器數值在圖示狀態204閱讀。把間斷輸 ' 控1111’發出轉移情涉及儲存裝置發動之訊號。 熊斷訊號職之前或先前完成㈣料傳輸狀 二、/Τί動次一轉移。「小」或「大」資料簇集均可定址， …。易言之，導:出連群=g ΑτΓ拉罢#^族集大小。流路驅動器模組112可含有邏輯為 褒置计异特殊磁扇開始位址和/或其他位址資訊。在本 200912728 =。具體例中，對全部DMA引擎可一起或同時發動6袖 群轉顯構’其侧轉移是以連 度戈簇集大小遞減連群長度。 測Ϊ箱集大小為零，如狀態208所示，表示DMA異 疋成。，此，系統控制器回到空轉狀態2〇2。 、 則表ίΪίϋ6ί測得簽集大小大於零’如狀態210所示’ :整：丄狀=新=移在之第後：所附健存裝置 216 ° 蝴靖 ϋ目關儲？裝置備妥時’狀態機迭代環路？以再度計算 剩下要轉移的簇集大小和位址，如狀態2G6所示。又° 址發？频财，流送資料描m慎理器的位 址=間内。反而’ PPC 1〇6 (如第1圖所示）為被出入㈣， 只輸运隨後、持續或接續的簇集位址。對大型、 田二’ 硬體模組需要較長時槽，以流送資料填 、^奶、二 箱集和適當資料率，會有較長間距才要的 動器模組m至PPC 1G6之_· 很低的時計頻率作業。 谷夺處理益在 在本發明-具體例中，狀態位址是在 一次之前，由PPC 106提供給流路驅動器^轉^序 移之開始位址是由流路驅動器模組112 、定 始位址之數值。此外，對剩餘之連群轉移 200912728 提供額外定址資料經内部控制路徑118至流路驅動器模組 112。在接到要求資料簇集轉移後，流路驅動器模組112適於 通知處理器，先前完成資料簇集轉移之完成狀態。在本發明 具體例中，設置第一次DMA出入後，流路驅動器模組112 適於向PPC 106要求次一開始位址，用於現時簇集轉移後 行的次-簇集轉移。 逛 、第3圖是製法流程圖，表示本發明具體例之方法。此方 法以參照數字鮮。專家均知方法需以資料儲存系 統100進行。本發明方法具體例另外可按照狀態圖 iiS η 只 在步驟302 办 負料在窖藏器（諸如1〇8)内緩衝。在步驟 ίο器内ί衝過之資料輸送至陣列控制器，諸如歸0 步驟306輸送至流路驅動 器接ίϊϊ二驅動器模組把控制資訊與從陣列控制 部份，由流路驅最後’從陣列控制器接到的資料 存裝置_ 4組在相#於控織訊之處，書寫於儲 要。專家均知，把轉虹述任何加咖合，亦屬需 【圖式簡單說明】 ί 資料記錄系統之方塊圖； 1 〇〇 §己錄系統 104系統控制器模組 108窖藏器 112流路驅動器模組 【主要元=^=明具體例方法之製法流程圖。 102使用者界面 106 PPC 110 RAID 控制 114 HDD陣列 12 200912728 118内部控制路徑 2〇〇狀態圖 204閱讀暫存數值器狀態 116外部控制路徑 120資料路徑 202空轉狀態 208測得簇集大小為零狀態210測得蔟集大小大於零狀態 206測定箱集大小和位址狀態 212啟動新的DMA連群轉移狀態 214狀態機DMA_control狀態 216 HDD群集備妥狀態 302把資料在窖藏器加以緩衝 304把窖藏器内緩衝過的資料輸送至陣列控 306把資料從陣列控制器輸送至流路驅動=200912728 IX. Description of the invention: [Technical field of the invention] The present invention relates to a large number of storage solutions having a plurality of storage units. Specifically, the present invention relates to a method for splitting data control information in a storage device.奘w 什 展 上 上 上 上 上 上 上 上 上 上 上 上 上 上 上 上 上 上 上 上 上 上 上 上 上 上 上 上Use a conventional disk drive, like (four) heart = limited delay on the path. Known systems: this Pi;; the body control path function is separated at a higher level = 疋 The Beca and the management, maintenance and / or user interface system handle a lot of control information. However, the known strategy still needs to be done as far as possible. In the data path, it is implemented as 'the road; 分; address: f: the processor sets up the data between the storage address f to control I/ O operation. The division device, and all kinds of communication networks to and from the communication network 2_2219 are intended to reveal the system including the three unified data paths. The storage system is in the data path and the control road 'data service platform position and control path. Ping Erkong: The device is located in the data path interface, including the control road and β, the machine interface, and the storage application host road and segment. The right 4cL port is connected to the service processor. The platform contains data path = bundle, only The use of data in the second paragraph, including the data path has a control interface. The controller contains the data path 200912728 machine building H country = 5,8G2, 366 is intended to reveal the kind of trough case service and local rationality t-bit plus itself In the slot case control dt4 rational 'layer. The virtual lag system is a part of the 窖 檑 檑 其 本身 本身 本身 本身 ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( Storage and export controller 'connected to - or above direct data transfer Order, launch = in the ^ ^ system to the sub-system transfer data 'and the data to be sent to the host memory specific "read two i!! storage subsystem installed ^ ^ system can be heavy; mistake, careful two D4 transfer. In addition, if the master device control that has made a mistake in the reading operation, the specific replacement data should be sent to the ashiit:6 method of separating the control path from the data path, and it is necessary to respect the building. For the complicated data path, because "the structure does not have a good degree of predictability, due to this ί^; ΐΐ ΐΐ related issues may occur in the worst case delay. This action leads to invalid hardware resources The use of the 巫 逖 需 需 需 需 需 需 需 需 需 需 需 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在The data received by the path plus the output control of the array controller controls her skirt to additionally include a control path, which is suitable for the flow path to drive the invention to the address on the storage device array. According to this, it is suitable for only cluster transfer; move, The legal package law stipulates that the scope of patent application is item 6. The party to the data transmission part of the buffer is connected. The system and the data array received from the array controller ί:: includes i is equivalent to control Where information is available, the information can be "on the array of 2m; the address on the array of control devices. From "column t is more common to the storage train, as - serial, = collection 4 to the data, can Transfer to the example; Tujia specific example is just a hair repair. This figure is a block diagram of the system, with a system controller module, according to the specific example of the invention according to 200912728, on the storage unit array storage function. The recording system shown in Figure j is generally indicated at 100. The recording system (8) includes a user interface 102 that allows the user to control the overall operation of the recording system 1 and view information about system status and the like. In one embodiment of the invention, the user interface includes an LCD touch pad display. The recording system 100 includes a system controller module (1) 4. The system control group H) 4 is adapted to transfer data to the HDD array 114 and thereby accept the data array comprising a plurality of individual HDDs. Data cluster transfer to and from the HDD array ^14 disk ^ is set or remember the appropriate value in the scratchpad, indicating the cluster size, cluster start address and related commands, such as "read" or "book". ^ The system controller module 104 includes an embedded software processor system, and the i-th is a PPC 106. PPC is an abbreviation for P〇wer pc. Ppc 1〇6 communicates with the external module via the Μ control path 116. The 2'τϋ f-slot system maintenance calculation performed by ppc 1〇6 controls the user's access with the external client, ΐί touch or other user interface or υι. For example, the logical addressing of the storage and the boundary between the physical addressing. 理 部署 部署 部署 垃圾 垃圾 垃圾 垃圾 垃圾 垃圾 垃圾 垃圾 垃圾 垃圾 垃圾 垃圾 垃圾 、 ί ί ί ί ί ί ί ί ί ί ί ί The commonality of the system/location task. Mode balance diameter 118, constitutes the new controller, crying 10S. In the non-specific example of the figure, PPC 106 constitutes an empty I":;;t: r^t:n:5 flow path drive If 112 surplus capital 1 miscellaneous Following the path 118, there is a correlation between the control information and the data in the first secret. Path 120 accepted; . , 告 〇 -1 〇 8 is suitable for transfer through the data, in the transfer to the RAID # system '^ ^ ^ ^ accepted instant flush. Before the RATD is controlled, the data is sent to the flow driver module 112, and the data is sent to the device attached to the HDD array 114. Additional information from the experts can be adapted to stream data from the HDD array 114. ° Recorded system RAID controller 11 multiplexed the incoming data stream. The multiplexing of the incoming material k is to divide the body into a plurality of streams and to cool them. The processes are intended to be written in different storages in the storage array. = Everyone knows that the RAID controller n〇 can also be used to increase the redundant data data S to recover the data from the second or lower fault brain, while the no-flow drive Lai group 112 splits the data stream into n blocks by word ^ Corresponding to the DMA bow engine, the service is accompanied by the standard ΑΤΑ-HDD. Experts know that, in the specific example, it can be used according to any statute. # κ is the second example of the statute. Addressing ΐΐ (ί 储存 储存 ) ) ) ) 亦可 亦可 亦可 亦可 亦可 亦可 亦可 亦可 亦可 亦可 亦可 亦可 亦可 亦可 亦可 亦可 亦可 伽 伽 伽 伽 伽 伽 伽 伽 伽 伽 伽 伽 伽 伽 伽 伽"Night dressing" means that the information is to be written in the fan: J/1 Λ 料 与 与 与 与 与 与 与 与 与 与 与 与 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 In addition, the internal control path screams and picks up the mosquito information to the flow path 11 for 112, without giving the control 110. The idea is to associate the specified clusters on the entry/event, brother: acre, ΐί device at a very low level, that is, at the last control-storage level or storage device communication. In this way, in the control, the high-speed data record performance is achieved. The specific relationship between the special information of the shi/^if system and the 'simplified record system' is set before the sequel, and the relevant information is just written in the _ array... Specifically, MA-likely allows processing! I in the segment data transfer, remove each single memory 200912728 line special I-type household; ί;: subject. This is done by providing dedicated hardware and in the scope of the invention. In addition, this DMA 曰 2 is more complex, and even set the subject, supervise and influence the problem. a grid series. This step-by-step removal of the brute force routine of PPC 106 is almost complete (4) for 12G to control the course. The result is better use of hardware agents. Breaking the core or discarding the redundant components, the low power control can be used to describe the operation of the specific example of the present invention. The state diagram touches the example of the invention, and the material and control system are operated! ^Before the figure 2 is marked with p°wer pc or ppc = I* is in the idling state 202. The PPC initiates the transfer, and the magnetic fan start address 2 要 is used for the _ operation or writing operation via the clustered big money value of the temporary storage. In a specific example of the invention = range up to 16 per device. These values are provided in the j specification (like 12C). The scratchpad is in the second picture ^ read s ready Iff". The register value is read in the illustrated state 204. The transfer of the intermittent transmission 'control 1111' involves a signal from the storage device. The bear was disconnected before or before the completion of the job (four) material transmission second, / Τ 动 move one transfer. "Small" or "Large" data clusters can be addressed, .... Easy to say, guide: out of the group = g ΑτΓ pull stop # ^ family size. The stream driver module 112 can include logic to determine a particular magnetic fan start address and/or other address information. In this 200912728 =. In a specific example, all of the DMA engines can be launched together with or at the same time. The side transition is decremented by the size of the cluster. The test box set size is zero, as indicated by state 208, indicating that the DMA is different. Here, the system controller returns to the idle state 2〇2. , 表 Ϊ ϋ ί ί ί 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签 签When the device is ready, the state machine iterates over the loop? To recalculate the cluster size and address to be transferred, as shown in state 2G6. Also ° address? Frequently, the data is sent to the address of the data processor. Instead, 'PCC 1〇6 (as shown in Figure 1) is the access (4), which only transports subsequent, persistent or consecutive cluster addresses. For large and Tian's hardware modules, it takes a long time slot to fill the data, milk, two boxes and the appropriate data rate. There will be a long distance between the actuator modules m to PPC 1G6. _· Very low timepiece frequency operation. In the present invention - the specific example, the status address is provided by the PPC 106 to the stream driver before the primary address is rotated by the stream driver module 112, the starting bit The value of the address. In addition, additional addressing data is provided to the remaining connected group 200912728 via internal control path 118 to stream driver module 112. After receiving the request data cluster transfer, the stream driver module 112 is adapted to notify the processor to complete the completion status of the data cluster transfer. In the specific example of the present invention, after the first DMA access is set, the stream driver module 112 is adapted to request the PPC 106 for the next-first address for the sub-cluster transfer after the current cluster transfer. FIG. 3 is a flowchart of the manufacturing method, and shows a method of a specific example of the present invention. This method is fresh with reference numbers. Experts are aware that the method needs to be performed with the data storage system 100. In particular, the method of the present invention can be buffered in a trap (such as 1〇8) in accordance with the state diagram iiS η only in step 302. In step ίο, the data rushed to the array controller is sent to the array controller, such as returning to step 306 to the stream driver, and the second driver module is used to control the information from the array control portion, and the flow path is driven to the last 'from the array. The data storage device received by the controller _ 4 groups is written in the storage area at the point where the control is woven. Experts know that it is necessary to transfer any coffee to the company. It is also required to be a simple diagram of the data recording system. 1 〇〇§ Recording system 104 system controller module 108 hidden device 112 stream driver Module [main element = ^ = specific method example method flow chart. 102 User Interface 106 PPC 110 RAID Control 114 HDD Array 12 200912728 118 Internal Control Path 2〇〇 State Diagram 204 Read Temporary Valuer State 116 External Control Path 120 Data Path 202 Idle State 208 Measured Cluster Size Zero State 210 Measured 大小 set size is greater than zero state 206 Measured box set size and address state 212 starts new DMA contiguous group transfer state 214 state machine DMA_control state 216 HDD cluster ready state 302 buffers data in the hidden device 304 in the hidden device The buffered data is sent to the array control 306 to transfer the data from the array controller to the flow path drive =
書寫到儲存裝置之陣列 13Writing to an array of storage devices 13