TW565801B - Cluster computer having distributed load balancing system - Google Patents

Cluster computer having distributed load balancing system Download PDF

Info

Publication number
TW565801B
TW565801B TW91119061A TW91119061A TW565801B TW 565801 B TW565801 B TW 565801B TW 91119061 A TW91119061 A TW 91119061A TW 91119061 A TW91119061 A TW 91119061A TW 565801 B TW565801 B TW 565801B
Authority
TW
Taiwan
Prior art keywords
server
load balancing
nodes
cluster computer
computer
Prior art date
Application number
TW91119061A
Other languages
Chinese (zh)
Inventor
Shien-Jy Lee
Yu-Cheng Chao
Chun-Ming Wu
Original Assignee
Infowrap Technologies Inc
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Infowrap Technologies Inc filed Critical Infowrap Technologies Inc
Priority to TW91119061A priority Critical patent/TW565801B/en
Application granted granted Critical
Publication of TW565801B publication Critical patent/TW565801B/en

Links

Landscapes

  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

A cluster computer is disclosed, and more particularly to a cluster computer having a distributed load balancing system. By utilizing the present invention, the calculation loading of each node is balanced and each node accesses the centralized storage system under the control of the distributed load balancing system while the cluster computer system performs calculations. Therefore, the quantity of storage device included in each node can be reduced, so that the cost of constructing the cluster computer system is decreased and the construction threshold thereof is decreased as well. Moreover, if front-terminal server of the cluster computer is broken down, the distributed load balancing system will take control of each device and node so as to keep the process of each node continuing and not affected by any influence, and the process data in the cluster computer is stored and protected safely, and the convenience of storing and backuping data are increased. Furthermore, the performance and processing speed of the cluster computer system are increased as well.

Description

經濟部智慧財產局員工消費合作社印製 565801 五、發明説明() 發明領域: 、丑本發明係有關於一種叢集電腦,特別有關於一種具有 分散式負载平衡系統之叢集電腦,讓叢集電腦在進行運算 (如二可分割運算)時,透過分散式負載平衡系統的控制可 使二母個節點(Node)共同存取同一儲存系、统,並且可妥善 保遵叢集電腦運作時的各樣資料,以及提升節點 理速度。 处 發明背景: 隨著半導體製程技術和積體電路設計之進步,個人電 '已經廣泛被採用於個人、家庭、學術研究、肖業以及工 業等方面。另一方面,隨著網際網路的蓬勃發展,大量資 訊不斷在纟罔際纟罔路内流動,特別是在電子商務和學術研究 的領域’資訊的處理和流動更是頻繁且沉重,戶斤以需要處 理月b力強大且具有高可靠度的高階飼服器等相關設備和器 材來進行穩定的支援與運作。 然而,一般具有強大處理能力的高階伺服器其價格卻 非ΐ叩貝,並不符合企業的經濟效益。由於目前一般個人 電腦用的處理器已經具有一定程度的處理能力及可靠性, 而且藉著應用超高速的網路連線(如具有Gigabit傳送速度 的網路,甚i是應A纖通道和光纖網路設備冑立的網2 連線),電腦與電腦之間的資料傳送速度得以 以可藉著平行處理和分散式計算等技術,將多台2電: 以網路連接形式建構成一個叢集電腦,提供高效能計算和 本紙張尺度適用中國國家標準(CNS)A4規格(210X 297公楚) …·:」......•裝.........訂.........# (請先閲讀背面之注意事項再填寫本頁> 565801 五 經濟部智慧財產局員工消費合作社印製 A7 發明説明() 處理予各領域。 習J參考第1圖,其所纷示為習知叢集電腦之示意圖。 為知叢集電腦1〇主要係由多個節點3〇和至少 ^成’且各設備係、透過網路連接設備(如交換器等)6〇互 ^接以構成-個區域網路20’其中主電腦7〇具有任務 为配、控制節點30的負載平衡,以及龄批金 W 从及皿控叢集電腦10内 斤有運作等功能,以於必要時加以控制。 請參考第2圖,其所繪示為根據第丨圖所示之習知叢 =電腦的節點内之設備連接示意圖。第丨圖之習知叢集電 月甸1 0中,每一節點3 〇可由一個或多個處理器⑼及記憶模 組90以共用匯流排1 〇〇的方式互相連接組成。 而且,習知叢集電腦10之每一節點30皆具有專屬的 U健存δ又備4〇或郎點储存系統等,如磁碟陣列(raid)、 網路健存設備(Network Attached Storage ; NAS)和儲存區域 網路(StorageAreaNetwork; SAN)等,藉以儲存節點30處 里的 > 料’又或作資料備份之用。一般而言,若節點3 〇係 由共用匯流排1 00連接多個處理器80構成時,節點儲存設 備40多為連接至共用匯流排1〇〇的共用儲存設備,以供多 個處理器80同時存取。 第1圖所示之叢集電腦1 〇,多係應用具有開放性原始 碼特性的作業系統,如Unix或Linux等作業系統作為其工 作平台,其長處為高穩定度、高彈性的設計及實施等,故 可根據使用者的需求與考量,進行不同的修改及調整,以 本紙張尺度適用中國國家標準(CNS)A4規格(210X297公釐) ......·裝.........訂.........# (請先閲讀背面之注意事項再填寫本頁) f輝高工作效率。 進行運算?“〇 一般係作高效能計算之用,當叢集電腦10 70依據應用°:可分割運算)時,連接網際網路50的主電腦 算住務的^夕Γ式的设定及每一節點的處理能力等,將此運 載平衡和力=子任務平均分配至各個節點3G,藉以達至負 節點3:=情況,依照應用程式的設計和安排,在多個 行動態負载平衡的處理。 主電腦;〇胳*即點30元成主電腦70交付的子任務後, 節動態負載平衡的原則,交付另-子任務予 進仃,直至此運算(如不可分割運算)完成為止。當 =時,主電腦70將根據應用程式的設計, 广的郎點储存設…取節點30進行的運 二,並將其整合統-為運算的結果。又或當各節點3二 進仃的子任務時,先將結果傳m電腦 : 待取得所有子任務的結果時,對所有子任務的結果進 統…藉以得到運算任務(如不可分割運算)的結果。 每一節點3 0在處理子J壬雜& 土入 進行資料存取,因此二:3〇都會對節點健存設備切 二 母個即點3 〇的工作效率與節點 備40不無關連。例如,節點儲存設備4〇與處理器8〇 接之間的頻寬、節點健存設備4〇的運作速度和叢集電腦 的工作平台和應用程式對節間存設備4()之優化等, 影響節點的工作效率。 會 (請先閲讀背面之注意事項再填寫本頁> 經濟部智慧財產局員工消費合作社印製Printed by the Intellectual Property Bureau of the Ministry of Economic Affairs, Consumer Cooperatives 565801 V. Description of invention () Fields of invention: The invention relates to a cluster computer, and in particular to a cluster computer with a distributed load balancing system. During the operation (such as two-division operation), through the control of the distributed load balancing system, two nodes (Nodes) can access the same storage system and system together, and can properly protect all kinds of data during the operation of the cluster computer. And increase node processing speed. Background of the Invention: With the advancement of semiconductor process technology and integrated circuit design, personal electronics have been widely used in personal, family, academic research, Xiao industry, and industry. On the other hand, with the vigorous development of the Internet, a large amount of information continues to flow within the borders, especially in the fields of e-commerce and academic research. The processing and flow of information is more frequent and heavy, Stable support and operation with related equipment and equipment that need to handle high-level feeders with high monthly power and high reliability. However, the price of high-end servers with powerful processing capabilities is not cheap, and it is not in line with the economic benefits of enterprises. Because the processors for personal computers currently have a certain degree of processing power and reliability, and through the use of ultra-high-speed network connections (such as networks with Gigabit transmission speed, it is even more suitable for A-fibre channels and optical fibers. The network equipment is connected to the network 2), and the data transmission speed between the computer and the computer can be parallelized and distributed computing technology can be used to connect multiple computers: a network connection to form a cluster Computer, providing high-performance calculations and this paper size is applicable to China National Standard (CNS) A4 specifications (210X 297 Gongchu)… ·: ”...... • Installation ......... Order ... ...... # (Please read the notes on the back before filling out this page> 565801 Five Consumers' Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs printed A7 Invention Description () to be processed in various fields. Xi J refers to Figure 1, It is shown as a schematic diagram of a conventional cluster computer. It is known that the cluster computer 10 is mainly composed of a plurality of nodes 30 and at least ^, and each device is connected to a device (such as a switch, etc.) through a network 60. ^ Connected to form a local area network 20 'where the host computer 70 has tasks The load balancing of the distribution and control nodes 30, as well as the operation of the internal control W10 and the cluster control computer 10 have functions such as operation to control them if necessary. Please refer to Figure 2, which is shown in accordance with Figure 丨The conventional cluster shown in the figure is a schematic diagram of the equipment connection in the nodes of the computer. In the conventional cluster shown in Figure 丨, each node 30 can be shared by one or more processors and memory module 90. The bus 100 is connected to each other in a way. Moreover, each node 30 of the conventional cluster computer 10 has a dedicated U storage δ and a 40 or Lang point storage system, such as a disk array (raid), Network Attached Storage (NAS) and Storage Area Network (SAN), etc., can be used to store > data in 30 nodes or for data backup. Generally, if node 3 〇 When the common bus 100 is connected to multiple processors 80, the node storage device 40 is mostly a shared storage device connected to the common bus 100 for simultaneous access by multiple processors 80. Figure 1 Cluster computer 10 shown, multiple systems Operating systems with open source features, such as Unix or Linux, are used as their working platforms. Their strengths are high stability and flexibility in design and implementation. Therefore, different operating systems can be implemented according to user needs and considerations. Revise and adjust, apply Chinese National Standard (CNS) A4 specification (210X297 mm) to this paper size ...... · Installation ............ Order ......... # (Please read the precautions on the back before filling out this page.) FHui working efficiency. "〇 Generally, it is used for high-performance computing. When the cluster computer 10 70 is based on the application °: Dividable operation), the host computer connected to the Internet 50 calculates housekeeping and the setting of each node. Processing capacity, etc., this load balance and force = sub-task is evenly distributed to each node 3G, so as to reach negative node 3: = situation, according to the design and arrangement of the application, dynamic load balancing processing in multiple rows. 〇 tick * that is 30 yuan into the sub-tasks delivered by the host computer 70, the principle of dynamic load balancing, and deliver another-sub-tasks to advance until this operation (such as indivisible operation) is completed. Based on the design of the application program, the host computer 70 will store a wide range of storage points ... Take the second operation performed by node 30 and integrate it-the result of the operation. Or when each node 3 enters the sub-task, First transfer the results to the m computer: When the results of all sub-tasks are obtained, the results of all sub-tasks are unified ... to obtain the results of computing tasks (such as indivisible operations). Each node 30 is processing sub-J 壬 Mi & amp Soil access for data access, Therefore, two: 30 will cut the two nodes of the node storage device. The working efficiency of point 30 is not related to the node backup 40. For example, the bandwidth between the node storage device 40 and the processor 80, the node The operating speed of the health storage device 40 and the optimization of the inter-node storage device 4 () by the working platform and applications of the cluster computer affect the node's work efficiency. Yes (Please read the precautions on the back before filling this page> Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs

、發明說明( 個r此外’當叢集電腦10進行運算(如不可分割運算),名 4〇及30在進仃運算任務時,需要同時存取節點儲存設传 進行動Γ來自主電腦70的通訊,例如主電腦70對節點3< 度' 1負載平衡、主電腦70詢問節點30運算任務的遠 點3〇主電知7〇對節點3〇 I求運算任務的結果等,使得節 大未能專注於處理運算任務,節點3〇的工 大幅下降。 卞口阳 專屈即點3 〇建置或維護方面,由於每個節點3 0皆需要 必點健存…°,所以當新增節點3〇時,成本勢2. Description of the invention (In addition, when the cluster computer 10 performs calculations (such as indivisible calculations), the names 40 and 30 need to simultaneously access the node storage settings to perform operations when performing computation tasks. Communication from the host computer 70 For example, the host computer 70 has a load balance of node 3 < 1 ', the host computer 70 asks the far point of the computing task of the node 30, the host computer 70, and the result of the computing task of the node 30, and so on. Focusing on processing tasks, the number of workers at node 30 has dropped significantly. 卞 Kou Yang specializes in the point of 30 construction or maintenance. Since each node 30 needs to be physically healthy ... °, when adding node 3 〇Hour, cost potential

内眾::的節點儲存設冑40而上升,而且,叢集電腦K 於Μ + βΡ點儲存^備4G係分別對應依附各個節點30,基 格为存設備的容量越大,每個儲存單位的成本越便宜之價 了'理,分散式的節點儲存設備4〇之成本偏高,同時亦降 面儲存資料的安全性和備份資料時的方便性。另一方 二為求叢集電腦10發揮高工作效率及具備高穩定性,各 ::儲存設備40的規格必須盡量統一,因而限制了採購 便:換郎點储存設備4〇的彈性’擁有叢集… 便難以降低。 ^ :......•裝.........訂.........雖 (請先閲讀背面之注意事項再場寫本頁} 經濟部智慧財產局員工消費合作社印製 ¾腦ο資 電電3的 主集點需 於叢節所 由若個算 ,’ 各運 面能,ο 方功時3 存等作點存 保衡運節儲 料平止時而 資载停同 , 在負而,控 和障作掌 配故運法 備 設 存 ¾ 節 個 每 在 分然地沒能 務突誤亦可 任70無料亦 控肖常資料 掌fa正的資 責主保生的 負 確產中 係f法所Q 700無或 1 便料 ¾) 公 97 2 X -10 2 規 4 A S) N C 標 家 國 國 中 用 適 度 尺 張 紙 本 565801 A7 B7 經 濟 部 智 慧 財 產 局 員 工 消 費 合 作 社 印 五、發明説明( 受到影響而損毀,可見習知的叢集電^ 1〇在資料保存方 面未能提供穩定的可靠度。 隨著知識經濟的發展,企業對於大量資料處理的應用 需求日趨迫切,而且資料的處理、儲存、安全和管理等需 求也呈現爆炸性成長,叢集電腦的應用將日益廣泛。然 而’傳統叢集電腦的實施和應用上卻未能同時提供低成 本同可罪度和向工作效率的運算。因此,如何建置可以 節省止業成本 '強化擴充能力、*高穩定性、高安全性的 叢集電腦’是知識經濟時代企業得以在激烈競爭中脫顆而 出的致勝關鍵。 發明目的及概述: 鑒於上述之發明背景中,隨著知識經濟的發展,企業 和學術研究等對於大量資料處理的應用需求曰料切,而 且資料的處理、儲在、容令知典挪松工, 、: 储存女王和皆理等需求也呈現爆炸性成 長’然而高階飼服器群組其價格卻非常昂貴。但隨著半導 體製程、積體電路設計和通訊網路技術的進步,一般的個 人電腦已具備可靠的處理能Λ ’網路通訊的速度亦快速提 升,因此藉由-般電腦透過網路連接建構而成的叢集電 腦,可提供接近高階飼服器群組的工作能力,但建置、維 護或擴充成本皆比高階祠服器群組便宜,於是提供了词服 器群組外的另一實施方案。纟叢集電腦中,每一節點皆具 有專屬的節點儲存設備或節點儲存網路作資料存取之用, 但當新增節點時’成本勢必因昂貴的節點儲存設備而上 本紙張尺度適用中國國家標準(CNS)A4規格(210X297公釐) ......·¥.........、可.........# (請先閲讀背面之注意事項再填寫本頁) 565801 A7 B7 經濟部智慧財產局員工消費合作社印製 五、發明説明() 升,而叢集電腦内分散的眾多節 即點儲存設備,亦導致維護 節點時成本大為增加。 此外,為求叢集電腦發捏古 r谭阿工作效率及具備高穩定 性,各個節點儲存設備的規格 乂肩盡置統一,限制了採購 或更換節點儲存設備的彈性,播士餐 擁有叢集電腦的成本便難以 進一步降低。 再者,若習知叢集電腦中堂 T旱控任務分配和負載平衡等 功能的主電腦因故障而停止運作吐 果作時,各個節點的運作和監 控’以及節點間的負載平衡# Α丨 • 衡便文到影響,儲存在每個節點 儲存設備中的資料亦可能受免| y ό 又到知毁,寶貴的資料因而流 失。 本發明的主要目的為提供了一 f種叢集電腦,特別係有 關於一種具有分散式負載平衡糸 丁何糸統之叢集電腦,讓叢集電 腦在進行運算(如不可分割運篡 ^ )寻,母一郎點共同存取同 一儲存系統,減少每一節點所包 I 3的儲存S又備,藉以降低 導入叢集電腦_。同時’亦可提升叢集電腦的處理速 度和運作可靠i ’並可增加儲存或備份資料時的安全性與 方便性。 ' 根據以上所述之目的,本發 4货听徒供了一種具有 負載平衡系統之叢集電腦,此且有八 ^ 匕具有刀散式負載平衡系統之 叢集電腦至少包括··具有複數個連接痒的網路連接 交換器广連接至網路連接裝置的储存系統;數個節點,這 些節點係用以共同執行一運算(如一不可分割運算卜且每 本紙張尺度適财_家群(CNS)A4祕⑽x297公楚 ...............、耵.........#. (請先閲讀背面之注意事項再填寫本頁) 565801 A7Inner public :: The node storage setting is increased by 40, and the cluster computer K is stored at the M + βP point. The 4G system corresponds to each node 30. The larger the capacity of the storage device, the larger the storage capacity of each storage unit. The cheaper the cost is, the higher the cost of decentralized node storage equipment 40 is. At the same time, it also reduces the security of data storage and the convenience of backing up data. On the other hand, in order for the cluster computer 10 to exhibit high working efficiency and high stability, the specifications of each :: storage device 40 must be as uniform as possible, thus restricting the procurement: the flexibility of the storage point 40 for the exchange point has a cluster ... Difficult to lower. ^: ...... • Installation ......... Order ......... Although (please read the notes on the back before writing this page) Employees of the Bureau of Intellectual Property, Ministry of Economic Affairs Printed by the consumer cooperative ¾ Brain ο The main collection point of the information electricity and electricity 3 needs to be calculated in the cluster section, 'Each transportation surface energy, ³ Fang Gongshi 3 deposits, etc., as points of storage and balance. The load is the same. In the negative, the control and the obstacles are in charge. The old law is prepared to store ¾ sections. Every time you fail to make a mistake, you can also take 70% of the charge. The main product of the main health insurance is Q 700 without or 1 note ¾) Public 97 2 X -10 2 Regulation 4 AS) NC bidding paper with a moderate rule paper 565801 A7 B7 Intellectual Property Bureau of the Ministry of Economic Affairs Employee Consumer Cooperatives Co., Ltd. 5. Description of the invention (Affected and damaged, it can be seen that the conventional cluster electricity ^ 10 has failed to provide stable reliability in data preservation. With the development of the knowledge economy, enterprises need application of large amounts of data processing Increasingly urgent, and the demand for data processing, storage, security and management is also exploding, clustering computers Applications will become increasingly widespread. However, 'the implementation and application of traditional cluster computers have failed to provide both low-cost, guilt-free, and work-efficient calculations. Therefore, how to build can save business costs'. Cluster computers with stability and high security are the key to the success of enterprises in the era of knowledge economy. In the context of the above invention, with the development of the knowledge economy, enterprises and academia The research and application requirements for large-scale data processing are cutting, and the data processing, storage, and storage of the codebooks, and the demand for storage of queens, etc. also show explosive growth. However, the high-end feeder group The price is very expensive. However, with the advancement of semiconductor manufacturing, integrated circuit design and communication network technology, ordinary personal computers have reliable processing capabilities. The speed of network communication has also increased rapidly. A cluster of computers constructed with a network connection can provide working capabilities close to high-end feeder groups, but The maintenance or expansion cost is cheaper than the high-end temple server group, so another implementation is provided outside the server group. In the cluster computer, each node has a dedicated node storage device or node storage network. For data access, but when adding nodes, the cost will inevitably be due to expensive node storage equipment. The size of the paper is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm) ... · ¥ .. ....... 、 可 ......... # (Please read the notes on the back before filling out this page) 565801 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs ), And the numerous nodes and point storage devices scattered in the clustered computer also lead to a significant increase in the cost of node maintenance. In addition, in order to achieve high working efficiency and high stability of cluster computers, the specifications of each node storage device are uniformly set, which limits the flexibility of purchasing or replacing node storage devices. It is difficult to further reduce costs. In addition, if the host computer that is familiar with the cluster computer ’s drought control task distribution and load balancing functions stops operating due to failure, the operation and monitoring of each node 'and the load balancing between nodes # Α 丨 • 衡The impact of the simple text may also exempt the data stored in the storage device of each node. It is also known that the valuable data is lost. The main purpose of the present invention is to provide a cluster computer, in particular, to a cluster computer with decentralized load balancing system, so that the cluster computer can perform operations (such as indivisible operation), Ichiro points common access to the same storage system, reducing the number of storages I and 3 contained in each node, thereby reducing the introduction of cluster computers. At the same time, it can also increase the processing speed and reliable operation of the cluster computer, and increase the security and convenience when storing or backing up data. '' According to the above-mentioned purpose, the listener of this article provided a cluster computer with a load balancing system, and there are at least eight cluster computers with a knife-type load balancing system including at least ... The network connection switch is widely connected to the storage system of the network connection device; several nodes are used to jointly perform an operation (such as an indivisible operation and each paper size is suitable for money_ 家 群 (CNS) A4 Secret x297 Gongchu ..............., 耵 ......... #. (Please read the notes on the back before filling this page) 565801 A7

請 閲 讀 背 面 - 之 注 意 事 項Please read the note on the back-Notes

訂 # A7 ^^------- 五、發明説明() 料存取,所以叢集電腦的運算資料得以妥善存放在集中式 =儲存系統中,不因外在環境的改變而受損,故此資料= B理、儲存與備份就更加安全和方便。另外,藉著利用集 中式的儲存系統,建置叢集電腦或擴充節點時的成本得以 大幅降低。 囷式簡單說明: 本發明的較佳實施例將於往後之說明文字中辅以下列 圖形做更詳細的闡述,其中: 第1圖係繪示習知叢集電腦之示意圖; 第2圖係繪示根據第i圖所示之習知叢集電腦的節點 内之設備連接示意圖; 第3圖係繪示本發明之一實施例的連接示意圖; 第4圖係繪示本發明之一較佳實施例的連接示意圖; 以及 第5A圖至第5B圖係繪示根據第4圖之中央儲存系統 的實施例之示意圖。 (請先閲讀背面之注意事項再場寫本頁> 裝 -訂· 經濟部智慧財產局員工消費合作社印製 囷號對照說明: 10 叢集電腦 20 區域網路 30 節點 40 節點儲存設備 50 網際網路 60 網路連接設備 70 主電腦 80 處理器 90 記憶模組 100 匯流排 300 叢集電腦 310 區域網路 10 本紙張尺度適用中國國家標準(CNS)A4規格(210X297公釐) 565801 經濟部智慧財產局員工消費合作社印製 五、發明説明() 320 節點 330 340 伺服器 350 360 伺服器 370 380 網際網路 390 400 控制台 410 500 儲存區域網路 510 520 儲存設備 550 560 儲存設備 發明詳細說明: A7 B7 中央健存系統 網路連接設備 前端伺服器 區域網路 連接線路 控制設備 網路儲存設備 請參考第3圖,其所繪示為本發明之一實施例的連接 示意圖。如第3圖所示,本發明所提供之叢集電腦3〇〇, 係由多個連接至區域網路310的節點32〇、一台連接網際 網路380的前端祠服器37G和至少—台具有分散式負載平 衡系統用作負載平衡等工作的伺服器34〇所組成,而且更 ^有一個連接至區域網路310的中央儲存系統33〇,其中 每一節點320可由一個或多個處理器及記憶模組以共用匯 流排的方式互相連接組成,每一節點32〇内並不包含節點 健存設備(或節點儲存系統)。 ” 當叢集電腦300運作時,前端伺服器370將 — (, %算任務 (如不可分割運算)給予用作負載平衡等工作的 3 40,然後由伺服器340將此運算任務平均分配至 ° I個節點 320 ’由眾多節點320執行此運算任務的子任務 ” , U達至負 載平衡的目的。由於本發明之叢集電腦300中,4、 、 彳糸分別利 本紙張尺度適用中國國家標準(CNS)A4規格(210X297公釐) ......•裝.........訂.........籲 (請先閲讀背面之注意事項再填寫本頁) 1 ο 8 5 6 5Order # A7 ^^ ------- 5. Description of the invention () Material access, so the computing data of the cluster computer can be properly stored in the centralized = storage system, not damaged due to changes in the external environment, Therefore, data = management, storage and backup are more secure and convenient. In addition, by using a centralized storage system, the cost of building cluster computers or expanding nodes can be greatly reduced. Simple description of the formula: The preferred embodiment of the present invention will be described in more detail in the following explanatory text with the following graphics, where: Figure 1 is a schematic diagram of a conventional cluster computer; Figure 2 is a schematic diagram FIG. 3 is a schematic diagram of device connection in a node of a conventional cluster computer according to FIG. I. FIG. 3 is a schematic diagram of connection of an embodiment of the present invention. FIG. 4 is a preferred embodiment of the present invention. 5A to 5B are schematic diagrams illustrating an embodiment of the central storage system according to FIG. 4. (Please read the precautions on the back before writing this page > Binding-Binding · Printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs, 囷 #Comparative instructions: 10 Cluster computers 20 Local area network 30 Node 40 Node storage device 50 Internet Road 60 Network connection equipment 70 Host computer 80 Processor 90 Memory module 100 Bus 300 Cluster computer 310 Local area network 10 This paper size applies to China National Standard (CNS) A4 specification (210X297 mm) 565801 Intellectual Property Bureau of the Ministry of Economic Affairs Printed by the employee consumer cooperative V. Description of the invention 320 Node 330 340 Server 350 360 Server 370 380 Internet 390 400 Control Panel 410 500 Storage Area Network 510 520 Storage Device 550 560 Storage Device Invention Details: A7 B7 The central storage system network connection equipment, front-end server, local area network connection, line control equipment, network storage equipment, please refer to FIG. 3, which is a schematic diagram of a connection according to an embodiment of the present invention. As shown in FIG. The cluster computer 300 provided by the present invention is composed of multiple nodes 32, 1, The front-end server 37G connected to the Internet 380 and at least one server 34 with a decentralized load balancing system for load balancing and other tasks, and a central storage system connected to the local network 310 33〇, of which each node 320 can be composed of one or more processors and memory modules connected to each other in a shared bus, each node 320 does not contain node storage equipment (or node storage system). " When the cluster computer 300 is operating, the front-end server 370 will give-(,% calculation tasks (such as indivisible calculation) to 3 40 for load balancing and other tasks, and then the server 340 will evenly distribute this calculation task to ° I Node 320 'The sub-tasks of this computing task are performed by many nodes 320', U achieves the purpose of load balancing. Because in the cluster computer 300 of the present invention, 4,, and 彳 糸 respectively, the paper size applies the Chinese National Standard (CNS) A4 size (210X297mm) ...... • Install ......... Order ......... (Please read the precautions on the back before filling this page) 1 ο 8 5 6 5

發明説明( 經濟部智慧財產局員工消費合作社印製 用不同㈣服器和網路設備等實施連接網際料3 =器370和用作負載平衡的飼服器34〇,而且前端飼 ^ ^ 具有防火牆的功能,所以可保護叢集電腦3〇〇内 部的運作。 5寺透過適當修改和調整叢集電腦3〇〇的運作平台 :應用程式等’例如,因為節點320及中央儲存系統330 白具有媒體存取控制位址(Media Aceess —Μ綠…; ^AC位址)’故可以此Mac位址料資料流動的參考,在 P』320進仃運算任務的子任務_,將所需資料储存至中 央儲存系統330’又或自中央儲存系統33〇讀取所需資料。 另外,由於前端飼服H 370只係負責將運算任務給予 用作負載平衡的伺服器34〇而不需負責負載平衡的工作, 所以當前端飼服器370受到不明的攻擊或故障而停止運作 時’,具有分散式負載平衡系統的飼服$ 34()仍可持 並掌管叢集電腦300内部的各設備和節點,持續對節點32〇 進仃負載平衡等工作,所以節點32〇的運作不受影響。而 且’由於本發明之叢集電腦係採用中央健存系統…而 所以經由飼服H 340和應用程式的控制,節點32〇 對中央儲存系、统330取用運算時所需之資料或存入所產生 =運算結果而不受外界影響’儲存於中央儲存系統33〇的 身料亦不會因而流失,重要的資料便得以妥善保全。。 再者,前端伺服器370和伺服器34〇可直接自中央 存系統330對節點32〇運算後所得之結果支 4,而不 12 本紙張尺度適用中國國家標準(CNS)A4規格(210x297公釐) ......•裝.........訂.........S (請先閲讀背面之注意事項再場寫本頁} 565801 五、發明説明() 4如習知叢隼雷腦 最果電知的主電腦等待節 需要連絡節點才可取得節點的,…广异、、J又或 器”〇和词…40 果。因此’前端飼服 f請先閲讀背面之注意事項再填、寫本Ϊ ]服器34〇不需要常常詢問節點32〇運 進度,亦不需對節點 便 求運异任務的結果,節點32〇 便了和用全部的資源 的工作效率得以提升,异任務,於是節點320 叢集電腦300的處理效能亦隨之大 平田’日加。 請參考第4圖,其所緣千A女欲口口 所、,θ不為本發明之一較佳實施例的 不思圖。在第4圖中,本發明之叢集電腦300,係利 用網路連接設備(如交換器等)35〇同時連接多個節點32〇、 兩台(或兩台以上)用作負載平衡等工作的飼服器跡兩台 (或兩台以上)連接網際網路38〇的前端伺服器37〇以及中 、儲存系統330,藉以構成一個區域網路39〇,其中每一節 點320可由一個或多個處王里器及記憶模組以共用匯流排的 方式互相連接組成,彳日灰 ΛΛ- ^ 仁母一即點内並不包含節點儲存設備 (或節點儲存系統)。 而且,叢集電腦3〇〇内更包括控制台4〇〇,藉由專屬網 路連線等方式,直接與前端飼服器37〇、用作負载平衡等 工作的伺服器3 60、中央儲存系統3 3 〇和眾多節點3 2〇等 經濟部智慧財產局員工消費合作社印製 進行連繫與通訊,掌握目前叢集電腦3〇〇内各設備的運作 情況,以便管理人員隨時進行監控。 由於第4圖之叢集電腦3〇〇具有連往網際網路38〇的 通道,故前端伺服器370 —般皆具有防火牆功能,而採用 13DESCRIPTION OF THE INVENTION (The Intellectual Property Bureau, the Ministry of Economic Affairs, the Employee Cooperative Cooperative, prints the implementation of the connection to the Internet with different servers and network equipment, etc. 3 = device 370 and a feeder for load balancing 34 0, and the front-end feed ^ ^ has a firewall Functions, so it can protect the internal operation of the cluster computer 300. 5 Temple through appropriate modification and adjustment of the cluster computer 300 operating platform: applications, etc. 'For example, because the node 320 and the central storage system 330 have media access Control address (Media Aceess —M green ...; ^ AC address) 'So you can use this Mac address as a reference for data flow, and enter the sub-task of the computing task at P "320 to save the required data to the central storage The system 330 'or the required information is read from the central storage system 33. In addition, since the front-end feeding server H 370 is only responsible for giving computing tasks to the server 34 for load balancing, it is not required to perform load balancing. So when the front-end feeder 370 stops operating due to an unknown attack or failure, the feeder with a distributed load balancing system of $ 34 () can still hold and manage each of the internals of the cluster computer 300 Devices and nodes continue to perform load balancing and other tasks on node 32, so the operation of node 32 is not affected. Moreover, because the cluster computer of the present invention uses a central storage system ... The control of the program, the node 32 has access to the central storage system, and the system 330 accesses the data required for the calculation or the generated = operation results are not affected by external influences. Lost, important data can be properly preserved. In addition, the front-end server 370 and the server 34 can directly calculate the results obtained from the central storage system 330 on the node 32, and the 12 paper standards are not applicable to China. National Standard (CNS) A4 Specification (210x297 mm) ...... • Installation ............ Order ......... S (Please read the precautions on the back first Write this page} 565801 V. Description of the invention (4) For example, the host computer that is most known by Cong Zhilei, who knows the most, needs to contact the nodes to obtain the nodes. … 40 results. Therefore, please read the notes on the back before writing and writing. The server 34 does not need to constantly ask the node 32 for the progress of the operation, and it does not need to ask the node for the results of different tasks. The node 32 is able to work with all the resources to improve the efficiency of the task, so The processing performance of the node 320 cluster computer 300 will also be followed by Ohira Hirata. Please refer to FIG. 4, which is related to a thousand female desires, and θ is not a schematic diagram of a preferred embodiment of the present invention. In Figure 4, the cluster computer 300 of the present invention uses a network connection device (such as a switch, etc.) 35 to connect multiple nodes 32 simultaneously, and two (or more than two) are used for load balancing and other tasks. Two (or more than two) feeders are connected to the front-end server 37 of the Internet 38 and the medium and storage systems 330 to form a local network 39, of which each node 320 may be composed of one or more The virgin ware and the memory module are connected to each other in a manner of sharing a common bus. The next day's ash ΛΛ- ^ mother does not include a node storage device (or node storage system). In addition, the cluster computer 300 also includes a control panel 400, which is directly connected to the front-end feeder 37 by means of a dedicated network connection, etc., a server 3 60 for load balancing, and a central storage system. The 3300 and many nodes 3200 and other employees of the Intellectual Property Bureau of the Ministry of Economic Affairs' consumer cooperatives print and communicate with each other to grasp the operation of various devices in the current cluster of computers 300, so that managers can monitor at any time. Because the cluster computer 300 in Figure 4 has a channel to the Internet 38, the front-end server 370 generally has a firewall function, and uses 13

565801 A7 B7 經濟部智慧財產局員工消費合作社印製 五、發明説明() 兩台(或兩台以上)的前她約 ~柯服器370之原因係用以平均分 擔工作以及胃其中—台前端伺服器370未能正常運作 時,另-台前端飼服器370可立刻接管,避免連接網際網 路380的通道完全癱瘓。同理,叢集電腦則採用兩台(或 兩台以上)的飼服器360來互相支援及備用,藉以增加叢集 電腦運作時的穩及當任-台㈣器360故障時另 一台可支持掌控各節點320持續進行運作。565801 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs. 5. Description of the invention () The reason for the appointment of two (or more than two) ~ KE server 370 is to share the work and stomach. When the server 370 fails to operate normally, another front-end feeder 370 can immediately take over to prevent the passage to the Internet 380 from being completely paralyzed. In the same way, the cluster computer uses two (or more than two) feeders 360 to support and back up each other, so as to increase the stability and performance of the cluster computer during operation-the other can support control when the Taiwanese device 360 fails. Each node 320 continues to operate.

另外’請同時參考第4圖、第5A圖和第5B圖,第5A 圖和第5B圖分別繪示為根據第4圖之中央儲存系統的實施 例之示意圖。第5A圖所示之中央儲存系統330係以健存區 域網路(SAN)·實施之,料請㈣接祕川自網路 連接設備350傳送至儲存區域網路5〇〇的控制設備川, 然後控制設備51〇將储存資料儲存至適當的健存設備 520 〇 第5B圖所示之中央儲存系統330係以網路儲存設備 (NAS)550實施之,儲存資料可經由多條連接線路41〇自網 路連接設備350傳送至網路儲存設備55〇中適當的儲存設 備560 。 。 又 第4圖之中央儲存系統33〇,可由如磁碟陣列、網路儲 存設備和儲存區域網路等實施’但並未限制於如上所述之 設備或系統。 應用本發明之叢集電腦300,除了節點32〇的工作效率 得以提升外,由於叢集電腦300的節點32〇内不包含節點 14 本紙張尺度適用中國國家標準(CNS)A4規格(210X 297公釐) ......•裝.........訂.........· (請先閲讀背面之注意事項再填寫本頁) 565801 A7 B7 五、發明説明() =存設備(或節點储存系統),所以當擴充節點32〇時成本 知乂減v仁部不會降低節點32〇的工作效率,而且因叢 集電腦300採用集中式的中央儲存系統33〇,而非習知叢 集電腦:呈分散式的節點儲存設備,基於儲存設備的容量 越大,母個館存單位的成本越便宜之價格原理,可得知應 用中央儲存系統3 3 〇的儲存成本較為便宜,並可改善節點 32〇使用專屬的節點儲存設備(或節點储存網财役 備的使用率不佳之問題。 诚仔°又 再者,由於應肖了集中式的中央儲存系、统330,叢隼電 腦300的架構猓以銪儿 ^ ^ 取亲电 9 ,/、應用程式和作業平台的架構也 于β ,便可提高執行的速度。而資料數據得 理和储存,大大增加健存管理人員於維護和備份時的 性,同時亦提升了資料的整合統一和安全性。 叢集明:優點為提供一種具有分散式負載平衡系統之 和對外遠结》本發明之叢集電腦係由負責運算任務委派 和=連線的前端飼服器、負責運算任務分配和負載平衡 的負載平衡词服考、# 、 ' ^ ^ ^ ° P .、和中央儲存系統組成,所以當對 外連線的刖端伺服器 衡㈣器、節點和“ 叢集電腦中負載平 電腦執行的運算可㈣^存系統等設備的運作,所以叢集 央儲存系統亦不合因i續進行而不受影響,同時储存於中 各節點皆存取飼服器故障而流失。而且,由於 運算,所以^點的一卢健存系統,節點便可集中資源處理 ”的處理能力得以提高,運算速度亦可加快, 本紙張尺賴财 ......·裝: (請先閲讀背面之注意事項再填寫本頁) 訂· 15 565801 A7 B7 五、發明説明( 同時使得資料的管理與備份更加可靠方便,並 集電細或擴充郎點時的成本。故此,本發明之 供了低成本、高可靠度和高工作效率的運算。 此外’由於本發明之具有分散式負載平衡 電服具有建置及客製化(Customize)容易、低導 擴充彈性但低擴充成本、高可靠性及穩定度等 別適合企業、學術研究和工程應用等需要高效 域之用。 如熟悉此技術之人員所暸解的,以上所述 之較佳實施例而已,並非用以限定本發明之 圍,凡其它未脫離本發明所揭示之精神下所完 變或修飾,均應包含在下述之申請專利範圍内 降低建置叢 叢集電腦提 系統之叢集 入成本、高 特點,故特 能計算的領 僅為本發明 申請專利範 成之等效改 (請先閲讀背面之注意事項再填寫本頁) 經濟部智慧財產局員工消費合作社印製 本紙張尺度適用中國國家標準(CNS)A4規格(210X297公釐)In addition, please refer to FIG. 4, FIG. 5A, and FIG. 5B at the same time, and FIG. 5A and FIG. 5B are schematic diagrams of an embodiment of the central storage system according to FIG. 4, respectively. The central storage system 330 shown in FIG. 5A is implemented by a health storage area network (SAN). It is expected that it will be transmitted from the network connection device 350 to the storage area network 500 control device. Then the control device 51 stores the stored data to a suitable health storage device 520. The central storage system 330 shown in FIG. 5B is implemented by a network storage device (NAS) 550. The stored data can be connected via multiple connection lines 41. From the network connection device 350 to the appropriate storage device 560 in the network storage device 55. . Also, the central storage system 33 in FIG. 4 can be implemented by, for example, a disk array, a network storage device, and a storage area network ', but is not limited to the devices or systems described above. With the application of the cluster computer 300 of the present invention, in addition to the improved working efficiency of the node 32, the node 32 of the cluster computer 300 does not include the node 14. The paper size is applicable to the Chinese National Standard (CNS) A4 specification (210X 297 mm) ...... • Installation ......... Order ......... (Please read the precautions on the back before filling this page) 565801 A7 B7 V. Description of the invention () = Storage equipment (or node storage system), so the cost is reduced when the node 32 is extended. The kernel will not reduce the working efficiency of the node 32, and because the cluster computer 300 uses a centralized central storage system 33, and Non-knowledge cluster computers: Based on the principle of decentralized node storage devices, the larger the storage device's capacity, the cheaper the cost of the parent storage unit. It can be known that the storage cost of the central storage system 3 3 0 is cheaper. And can improve the use of dedicated node storage equipment for node 32 (or the problem of poor utilization rate of node storage network financial reserve.) Cheng Zi ° Furthermore, due to the centralized central storage system, system 330, cluster猓 Architecture of computer 300 猓 猓 ^ ^ Pro-electricity 9 /, the structure of the application and operating platform is also in β, which can improve the speed of execution. The data and data processing and storage greatly increase the performance of maintenance personnel during maintenance and backup, and also improve Data integration, uniformity, and security. Clustering: The advantage is to provide a decentralized load-balancing system with a remote connection. "The cluster computer of the present invention is responsible for computing task delegation and = connected front-end feeders, responsible for computing The task of load distribution and load balancing is based on the load balancing test, #, '^ ^ ^ ° P., and the central storage system, so when externally connected server balancers, nodes, and load balancers in the cluster computer The calculations performed by the computer can save the operation of the storage system and other equipment, so the cluster central storage system is also unaffected by the continuous operation, and at the same time, all the nodes stored in the storage access server are lost and lost. Moreover, due to the calculation Therefore, the processing power of the “one-point health storage system, nodes can be centralized resource processing” has been improved, and the calculation speed can be accelerated. : (Please read the notes on the back before filling out this page) Order · 15 565801 A7 B7 V. Description of the invention (Meanwhile, the management and backup of the data are more reliable and convenient, and the cost of collecting electricity or expanding the land price is increased. Therefore, The invention provides low-cost, high-reliability, and high-efficiency operations. In addition, 'as the invention has a decentralized load balancing electrical service, it is easy to build and customize (Customize), low-conductivity expansion flexibility, but low expansion Cost, high reliability, and stability are not suitable for enterprises, academic research, and engineering applications that require high-efficiency fields. As those skilled in the art understand, the above-mentioned preferred embodiments are only used to limit the present invention. The scope of the invention, all other changes or modifications that do not depart from the spirit disclosed by the present invention, should be included in the scope of the following patent applications to reduce the clustering cost and high characteristics of the cluster computer system, so it can The calculation is only the equivalent change of the patent application for this invention (please read the precautions on the back before filling this page). Cooperatives printed in this paper scale applicable Chinese National Standard (CNS) A4 size (210X297 mm)

Claims (1)

565801565801 申請專利範圍 ABCD 經濟部智慧財產局員工消費合作社印製 1· 一種具有分散式負載平衡系統之叢集電腦,至少包 括: 至少一網路連接設備,該至少一網路連接設備具有複 數個連接埠; 一儲存系統,該儲存系統係對應連接至該些連接埠之 至少一連接埠,其中該儲存系統具有可擴充性,以及該儲 存系統係選自於一儲存區域網路和一網路儲存設備中之一 者; 複數個節點,該些節點係用以執行一運算任務之複數 個子任務’且每一該些節點係連接至該些連接埠之對應一 第連接槔’藉以在執行該些子任務時對該儲存系統進行 一存取步驟;以及 至少一刚端伺服器’該至少一前端伺服器之一端係連 接至一網際網路; 至少一電腦伺服器,該至少一電腦伺服器之一端係連 接至該至少一前端伺服器之另一端,該至少一電腦伺服器 之另一端係連接至該些連接埠之對應一第二連接埠,且該 至少一前端伺服器係用以將該運算任務委派至該至少一電 腦伺服器,而該至少一電腦伺服器係用以對該些節點進行 一運算任務分配和一負載平衡控制。 2.如申請專利範圍第1項所述之具有分散式負載平衡 I 裝 訂------------J. (請先閱讀背面之注意事項再填寫本頁) 17Patent application scope ABCD Printed by the Consumer Cooperative of Intellectual Property Bureau of the Ministry of Economics 1. A cluster computer with a distributed load balancing system, at least including: at least one network connection device, the at least one network connection device having a plurality of ports; A storage system corresponding to at least one port connected to the ports, wherein the storage system has expandability, and the storage system is selected from a storage area network and a network storage device One of a plurality of nodes, the nodes are used to perform a plurality of sub-tasks of a computing task, and each of the nodes is connected to a corresponding first connection of the ports, so as to execute the sub-tasks An access step is performed on the storage system at the time; and at least one rigid server is connected to an Internet; one of the at least one front-end server is connected to an Internet; at least one computer server is one of the at least one computer server. Connected to the other end of the at least one front-end server, and the other end of the at least one computer server is connected to the The second port corresponds to a second port, and the at least one front-end server is used to delegate the computing task to the at least one computer server, and the at least one computer server is used to perform a calculation on the nodes. Task distribution and a load balancing control. 2. Binding with decentralized load balancing as described in item 1 of the scope of patent application Binding ------------ J. (Please read the precautions on the back before filling this page) 17 565801565801 ^、申請專利範園 ABCD 系統之雪隹I⑽ . 取果电細,其中上述之至少一網路連接設備為至 一交換器。 3 ·如申請專利範圍第1項所述之具有分散式負載平衡 系統之叢集電腦,其中上述之運算任務分配係將該運算任 務之該些子任務平均分配至該些節點。 4 ·如申請專利範圍第1項所述之具有分散式負载平衡 系統之叢集電腦,其中每一該些節點係由複數個處理器和 至少—記憶模組以共用匯流排的連接方式構成。 5·如申請專利範圍第丨項所述之具有分散式負载平衡 系統之叢集電腦,其中上述之叢集電腦更包括至少一控制 台’用以監控該至少一網路連接設備、該儲存系統、該些 節點及該至少一電腦伺服器的運作。 C請先閲讀背面之注意事項再填寫本頁} 裝. 、一叮. 圍 範 利 專 請 申 如 其 甾 月 Mr 之 統。 系算 衡運 平割 載分 負可 式不散一 分 有 具 之 述 所 頃 述 11 上 第中 為 務 任 算 il 之 經濟部智慧財產局員工消費合作社印製 括 包 少 至 腦 1^¾ 集 叢 之 統 系 衡 平 載 負 式 散 分 有 具 t 種 埠 接 4βΐϋ 個 數 複 有 具 器 換 交 - 少 至 該 器 換 交 1 少 至 8 本紙張尺度適用中國國家標準(CNS)A4規格(210X297公釐) 申請專利範圍 至少一前端伺服器,該 接一網際網路,該至少一寸t夕一前端伺服器之一端係連 至該些連接埠之至,|、 〗端伺服器之另一端係對應連接 (請先閲讀背面之注意事項再填寫本頁) 乂 一連接埠; 一儲存系統,該儲存系 該至少-連接埠,纟應連接至該些連接琿之 储存系統係選自卜=存系統具有可擴綠’以及該 —者; A網路和一網路儲存設備中之 複數個節點,該此節 個子任藤日—一即點係用以執行一運算任務之複數 于任務’且母一續此於 第Ί 冑係連接至料連接彳之對應一 第一連接埠,藉以在執杆纺 .^ ^ 執仃該些子任務時對該儲存系統進行 一存取步驟;以及 一至少一負載平衡词服器,該至少一負載平衡伺服器之 一=係連接至該些連料之對應—第二連料,且該至少 =前端伺服器係用以將該運算任務委派至該至少—負載平 衡祠服器u至少-負載平衡飼服器係用以對該些節點 進打一運算任務分配,藉以將該運算任務之該些子任務分 配至該些節點,以及對該些節點進行—負載平衡控制。 經濟部智慧財產局員工消費合作社印製 8·如申請專利範圍第7項所述之具有分散式負載平衡 系統之叢集電腦,其中每一該些節點係由複數個處理器和 至少一記憶模組以共用匯流排的連接方式構成。 9·如申請專利範圍第7項所述之具有分散式負載平衡 19 本紙張尺度適用中國國家標準(CNS)A4規格(210X 297公釐) ABCD 565801 々、申請專利範圍 系統之叢集電腦,其中上述之叢集電腦更包括至少一控制 台,用以監控該至少一交換器、該至少一前端伺服器、該 儲存系統、該些節點及該至少一負載平衡伺服器的運作。 10.如申請專利範圍第7項所述之具有分散式負載平衡 系統之叢集電腦,其中上述之運算任務為一不可分割運 算。 1 1.如申請專利範圍第7項所述之具有分散式負載平衡 系統之叢集電腦,其中上述之至少一前端伺服器更具有該 叢集電腦之一防火牆功能。 (請先閲讀背面之注意事項再填寫本頁) 經濟部智慧財產局員工消費合作社印製 20 本紙張尺度適用中國國家標準(CNS)A4規格(210X 297公釐)^, Apply for patent Fanyuan ABCD system's snow 隹 I. Take the fruit of electricity, where at least one of the above network connection equipment is to a switch. 3. The cluster computer with a decentralized load balancing system as described in item 1 of the scope of the patent application, wherein the above-mentioned computing task allocation is an equal distribution of the sub-tasks of the computing task to the nodes. 4 · A cluster computer with a distributed load balancing system as described in item 1 of the scope of the patent application, wherein each of these nodes is composed of a plurality of processors and at least a memory module connected in a shared bus connection. 5. The cluster computer with a decentralized load balancing system as described in item 丨 of the patent application scope, wherein the cluster computer further includes at least a control console for monitoring the at least one network connection device, the storage system, the The nodes and the at least one computer server. CPlease read the precautions on the back before filling out this page} Installation, Yiding, and Fan Li, please apply for the same system as Mr. Wang. The calculation of Heng Yunping ’s load-carrying load can be described in detail. 11 The above is the Ministry of Economic Affairs Intellectual Property Bureau employee consumer cooperative printed by the Ministry of Economic Affairs. The system of the cluster is the flat load negative sub-distribution with t types of port 4βΐϋ. The number of devices is exchanged-as little as the device is exchanged 1 as little as 8 This paper size applies to China National Standard (CNS) A4 specifications (210X297 The scope of the patent application is at least one front-end server, which is connected to the Internet. One end of the at least one inch front-end server is connected to the ports, and the other end of the server is Corresponding connection (please read the precautions on the back before filling this page) 乂 A port; a storage system, the storage is the at least -port, and the storage system that should be connected to these connections is selected from the following: The system has expandable green 'and the one; A network and a plurality of nodes in a network storage device, this section of the child Ren Tengri-a point is used to perform a plurality of computing tasks in a task 'And the mother continues to connect to the first connection port corresponding to the material connection at the first (i.e., the first connection port), so as to perform an access step on the storage system when performing the sub-tasks; and A at least one load balancing server, one of the at least one load balancing server = connected to the corresponding ones of the connected materials-the second connected material, and the at least = front-end server is used to delegate the computing task to The at least-load-balanced server server at least-load-balanced server is used to assign an operation task assignment to the nodes, thereby assigning the sub-tasks of the operation task to the nodes, and to the nodes. Nodes perform-load balancing control. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs8. A cluster computer with a distributed load balancing system as described in item 7 of the scope of patent application, wherein each of these nodes is composed of multiple processors and at least one memory module It is constituted by a common bus connection. 9 · With decentralized load balancing as described in item 7 of the scope of patent application 19 This paper size applies Chinese National Standard (CNS) A4 specification (210X 297 mm) ABCD 565801 々 Cluster computer for patent application system, where the above The cluster computer further includes at least one console for monitoring operations of the at least one switch, the at least one front-end server, the storage system, the nodes, and the at least one load balancing server. 10. The cluster computer with a distributed load balancing system as described in item 7 of the scope of the patent application, wherein the above-mentioned calculation task is an indivisible operation. 1 1. The cluster computer with a distributed load balancing system as described in item 7 of the scope of the patent application, wherein the at least one front-end server further has a firewall function of the cluster computer. (Please read the precautions on the back before filling out this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 20 This paper size applies to China National Standard (CNS) A4 (210X 297 mm)
TW91119061A 2002-08-22 2002-08-22 Cluster computer having distributed load balancing system TW565801B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
TW91119061A TW565801B (en) 2002-08-22 2002-08-22 Cluster computer having distributed load balancing system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
TW91119061A TW565801B (en) 2002-08-22 2002-08-22 Cluster computer having distributed load balancing system

Publications (1)

Publication Number Publication Date
TW565801B true TW565801B (en) 2003-12-11

Family

ID=32502620

Family Applications (1)

Application Number Title Priority Date Filing Date
TW91119061A TW565801B (en) 2002-08-22 2002-08-22 Cluster computer having distributed load balancing system

Country Status (1)

Country Link
TW (1) TW565801B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8332483B2 (en) 2003-12-15 2012-12-11 International Business Machines Corporation Apparatus, system, and method for autonomic control of grid system resources
TWI400655B (en) * 2008-12-25 2013-07-01 Ind Tech Res Inst Apparatus and method for dynamic score-based load balancing
TWI448975B (en) * 2011-07-15 2014-08-11 Chunghwa Telecom Co Ltd Dispersing-type algorithm system applicable to image monitoring platform
CN113553171A (en) * 2021-06-07 2021-10-26 用友汽车信息科技(上海)股份有限公司 Load balancing control method, device and computer readable storage medium

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8332483B2 (en) 2003-12-15 2012-12-11 International Business Machines Corporation Apparatus, system, and method for autonomic control of grid system resources
TWI400655B (en) * 2008-12-25 2013-07-01 Ind Tech Res Inst Apparatus and method for dynamic score-based load balancing
TWI448975B (en) * 2011-07-15 2014-08-11 Chunghwa Telecom Co Ltd Dispersing-type algorithm system applicable to image monitoring platform
CN113553171A (en) * 2021-06-07 2021-10-26 用友汽车信息科技(上海)股份有限公司 Load balancing control method, device and computer readable storage medium

Similar Documents

Publication Publication Date Title
US7478177B2 (en) System and method for automatic reassignment of shared storage on blade replacement
US7509506B2 (en) Hierarchical system and method for managing power usage among server data processing systems
CN102473106B (en) Resource allocation in virtualized environments
US9329792B2 (en) Storage thin provisioning and space reclamation
US20100223618A1 (en) Scheduling jobs in a cluster
TWI285304B (en) Extendable storage apparatus for blade server system
US11809252B2 (en) Priority-based battery allocation for resources during power outage
CN103873559A (en) Database all-in-one machine capable of realizing high-speed storage
KR20170097132A (en) System for efficient processing of transaction requests related to an account in a database
US11397653B2 (en) Technologies for fast recovery of distributed storage systems on disaggregated storage
CN110727950A (en) Distributed cooperative computing system and cooperative processing method
US20200244533A1 (en) Single management connection automatic device stack configuration system
EP4002139A2 (en) Memory expander, host device using memory expander, and operation method of server system including memory expander
TW565801B (en) Cluster computer having distributed load balancing system
CN116074179B (en) High expansion node system based on CPU-NPU cooperation and training method
US20200125494A1 (en) Cache sharing in virtual clusters
US20170249279A1 (en) Riser matrix
US10860334B2 (en) System and method for centralized boot storage in an access switch shared by multiple servers
CN105005547A (en) NUMA-based multi-path server complete physical partitioning method
Deng et al. The performance of a supercomputer built with commodity components
US20200142618A1 (en) Cabinet server system and server
US9639427B1 (en) Backing up data stored in a distributed database system
US20040024942A1 (en) Super scalable multiprocessor computer system
US20230195528A1 (en) Method and apparatus to perform workload management in a disaggregated computing system
Bai et al. Cloud Computing (3)

Legal Events

Date Code Title Description
GD4A Issue of patent certificate for granted invention patent
MM4A Annulment or lapse of patent due to non-payment of fees