1258282 九、發明說明： 【發明所屬之技術領域】 本發明係關於網路交換奘罟，士、贫3 士„ 吩又侠衣置，尤其疋有關於網路交換裝 置搜尋並決定封包轉發路徑的電路架構。 【先前技術】 不μ有線的或是無線的區域網路’因騎線的長度與 • 電波的涵蓋範圍有理論與實務的限制，區域網路交換器 (switch)㈣著聯繫各個獨立的有線或無線網段、成為^ 涵蓋完整的區域網路的樞紐的角色。 隨著技術的進步，交㈣·的魏也日漸提升，除了支援 更夕阜口外’其他m整合光纖、無線的傳輸界面與媒介， 乃至於提供更高層次的封包處理（例如layer 3 s所提供 的負載平衡等）等功能的交換器，在市場上也愈來愈普遍了。 # <不哪疋那種父換器，它的最基本功能還是在於所謂網路封 包的轉發（f〇rwardjng )。 第1圖係一般區域網路之架構示意圖。如第丨圖所示， 又換為10的埠口 0、2、5分別銜接三個網段20、30、40。 $著三個網段上的電腦開始通訊與交換資料，交換器i。的硬 或内建的初體會根據它所收到的、由這些電腦所發出的封 匕裡的來源與目的地位址（即網路第二層的所謂Μ。位 、v建立起個查找表（lookup table )。這個查找表（未 ")裡。己錄了區域網路裡每一台電腦的MAC位址、是位於 6 1258282 哪一個埠口所連接的網段 _有電腦的網路第三層地址 (例如ίρ地址）、以及其他相關的資料。 假設電腦21要傳送資料給電㈣ ^ 又換π w從網段20 _ 21發出、以電腦31為目的地的封包，藉由查 表，交換U得知電腦31是位於連接於心2的網段曰 ^千這些封包轉發到網段30(而不是網段2〇和4〇)。同樣的， 〇 是要發給電腦22,交換器H)從網段2〇 上收到由電腦21發出、以電腦U為目的地的資料封包，藉 由查表’讀器H)得知電腦22是位於和_源的電腦h 目_網段2G’_1G知道電㈣也會制_封包， 於疋捨棄這些封包、不將這些封包轉發出去。藉由這樣的方 ^交換器1〇確使網段2〇、3〇、4〇裡不會有不必要的封包 h來影響網路的效能。需要注意的—點是，交換器10的上 述封包轉發動作是有時間限制的，交換器H)必須在-個埠口 收到下-個封包的第—個位元前，完成這個埠口目前已智收 ，封包㈣發’㈣可能有_情形會發生，—種情形是 :為來不及對這個目前已經收到的封包做出處置，而造成這 :封包的吾失’另—種情形是為了完成這個封包的處置，而 :失了下-個封包前面若干位元的收取，也會造成下—個封 ^的丢失。這樣的封包H在目前習知通訊協定下，會使 發送封包的電腦在-段時間後重送同-封包，而這些重送的 1258282 封包-樣會消耗網路頻寬，而傳輸時間的延長也使得資料傳 輸的速度降低’可以想見，當這種前後封包「衝撞」的情形 很多時，區域網路的整體效能會顯著的降低。 ’ 隨著交換器網路埠σ、與所連接電腦數目的增加、以及 網路的速度由HK)Mblts/see提升到刪⑽乃至於 1’交換器的基本的封包轉發功能也面臨相當大的 挑戰。因為連線的電腦數目的增加，查找表所記錄的内容增 加，搜尋查找表的時間必然會延長，而隨著網路傳輸速度的 增快’交換器所能進行轉發等處置的時間更形縮短，因此如 何有效的搜尋查找表成為㈣||研發的重要課題。 習知的交換器的封包路徑搜尋方法主要有線性搜尋 (nnear search )、雜凑（hashing )、以及内容定址記憶體 (content addressable memory，CAM )三種。利用線性搜尋的 查找表的記錄是依照某種順序排列，而從名字可以想見，線 性搜尋是㈣這個順序比對查找表㈣各個電腦的記錄，直 到找到和封包目的地的MAC位址符合的腦記錄為 止。這種方法因為非常簡單，用於存放查找表的記憶空間也 可以被完全湘。線性搜尋通f是以款體的方式實施，因此 為了加快搜尋速度以避免封包衝撞，f需要快速的處理器， 使得實施的成本提高。 雜凑的方法則是利用-個事料定好的雜凑函數，這個 8 1258282 雜凑函數以像是封包的M A C位址為輸入，換算出—個杳找♦ 的記錄地址。將一個電腦的資料記錄在查照表裡時，是利用 這個雜凑函數決定出存放在查找表的哪_筆記錄。而在收到 -個封包要決定它的轉發路徑時，也是利用這個雜凑函數計 算出目的地電腦的記錄是位於查找表什麼地方。因為只需要 -次運算，而且雜凑函數常以硬體的方式實施，所以雜凑的 搜尋方式速度非常快速。但是因為雜凑函數可能會對不同的 mac位址也計算得到相同的記錄地址，也就是所謂雜凑衝撞 “ashing coUision)的情形，這種幾乎是無法避免的雜凑衝 撞使得查找表的記憶”無法充分彻（有些地址可能永遠 也用不到），而雜凑衝撞因為會使得多個電腦的記錄都對應到 同-個位址’所⑽了前述的—次運算外，還要再增加對這 些不定數目的電腦記_比對，造成搜尋時_不確定。CAM 的做法和雜湊類似是利用封包的内容來計算出—個查找表的 地址’而CAM採用複雜的硬體定址方式以達料算的結果是 個唯一的地址，以避免衝撞的情形，所以cam同時具有搜 尋陕速：14充77利用5己憶空間的優點，但CAM的缺點就是成本 過於高昂。 【發明内容】 基於智知的區域網路交換裝置的查找表的搜尋方式的缺 點’本發明提找表的電路架構，—方面可以充分利 9 1258282 用交換器裡用於存放查找表的記憶空間，另—方面則可以確 保每-次的搜尋都—定不會超過—定的時限而造成封包的衝 撞。 本發明的最主要精神不在追求最短的搜尋時間，而是大 里運用平4的*構’以保證即使在最差的情形其搜尋時間也 必小於-定上限的可預測性。藉由這樣的設計，依據本發明 • ㈣路交換^可以經由計算而決定採用效能適當的硬體元 件，而不至於像部份習知技術，因為搜尋時間的不確定性而 不付不採用可能完全沒有必要的高效能硬體元件。依據本發 明的網路交換器因此可以具有合理的成本。 本毛明的另一目的係基於所提出的電路架構的搜尋方 式’可以用硬體方式實施。硬體化的結果，可以使得搜尋速 度更為快速、成本降低、減輕其他元件的運算負擔。 # 本發明所提出的電路_，是將存放纽表的記憶空間 劃分為N個區塊，每-健塊可以存放L筆記錄（所以總共 可以存放NxL筆記錄）。這N個記錄區塊㈣和網路交換器 的Μ個埠口各自的搜尋比對裝置以分別的匯流排連接（所以 總共有Ν個匯流排）。也就是說，經由這些匯流排，一個記錄 區塊的一筆記錄可以同時、平行輸出給所有Μ個搜尋比對裝 置，而一個搜尋比對裝置可以同時收到Nf分別來自所有Ν 個§己錄區塊的記錄。 10 1258282 本發明所提出的電路架構還包含有一個定址裝置，這個 定址裝置在被-個搜尋比對裝置觸發後，會依序連續產生地 址0、1、…、L-1信號同時給所有N個記錄區塊。每收到一 個地址彳§ 5虎，一個記錄區塊就經由它的匯流排將那個地址的 記錄平行輸出給全部Μ個搜尋比對裝置。因此一個搜尋比對 裝置會依序收到所有存放在Ν個記錄區塊地址〇的Ν筆記 • 錄、然後是所有存放在Ν個記錄區塊地址1的Ν筆記錄、然 後是所有存放在Ν個記錄區塊地址筆記錄，依此類推， -個搜尋比對裝置會在定址裝置產生完地址〇、i、…、W u虎的彳固循％後’就比對過所有NxL筆記錄。因此在最差 的情形，-個封包的搜尋查找表時間最多不會超過定址裝置 產生L個地址信號的一個循環的時間。 兹配合所附圖示、實施例之詳細說明及中請專利範圍， • 將上述及本發明之其他目的與優點詳毅後。然而，當可了 而設，不當視為本發明 ’請參照所附之申請專 解所附圖示純係為解說本發明之精神 範噚之定義。有關本發明範疇之定義 利範圍。 【實施方式】 本發明所提出的電路架構適用於網超 的交換裝置裡做位址記錄的儲存與搜尋。 網路第二層（含）以上 .裝置」包括一 般的第二層的區域網路交換器 這裡所謂的「交換 器，也包含具備第 1258282 二層以上交換功能的交換器、路由器、或是類似的裝置。這 衣置最大的4寸欲在於具有多個網路璋口，並能夠決定從一 们阜口收到的封包應該是丢棄、還是轉發到哪些其他的蜂 口以及疋否需要記錄發出封包的運算裝置是位於哪個蜂口 所連接的網段等相關資訊。 第2圖係本發明一較佳實施例應用於一第二層之區域網 路又換之不⑦圖。本發明所提出的電路架構以及此架構的 運作方式，可以完全以硬體方式實施，也可以部份硬體、部 伤早刃體或疋全部韋刃體的方式實施。在此實施例裡，主要是 、第^的MAC 1立址為主要搜尋的依據，然請了解本發明之 架構當可適用於其他洲MAC位址以外的位址為路經決定 依據的網路交換裝置。 第圖所示僅是一第二層區域網路交換器内部有關本發 明的部份’交換器的其他模組都與習知者相同，故於此省略。 對於熟習相關技藝人士，#無礙於其了解本發明之主要精神 斤在如第2圖所示，交換器具有8個網路埠口（編號 P〇 P7) ’在交換器内部這8個璋口的界面電路之後各連結有 —個對應的搜尋比對裝置。每-個璋口收到封包時，就 曰將封包透過連線21〇傳送給對應的搜尋比對裝置。請 注意到各個埠口以及對應的搜尋比對裝置200都是平行運 就疋況，比如，埠口 2正在接收其所連接網段的封包 12 1258282 日守，埠口 5也可能同時在接收其所連結網段的封包。 搜♦比對U 2GG收到_個封包以後，會擷取封包裡的 重要資訊，像是發出此物運算麵㈣位址，以及這 個封包目的地的MAC位址等。搜尋比對裝置2〇〇接著透過匯 ⑽排220對一個定址裝置3〇〇送出一個搜尋信號。請注意到， 因為每―埠口是平行運作，紐裝置則可能在同—時間會 收到來自-個以上的搜尋比對裝置所發出的搜尋信號， 而且這些搜尋信號不會彼此干擾。 定址裝置300在收到搜尋信號的觸發後，就會連續循序 產也址彳s號輸出到匯流排3 1 〇上。在本實施例裡，定址裝 置j〇〇會循序產生地址0、地址丨、…、地址127共128個地 址^號。如第3a圖所示的匯流排31〇的時序圖，在時間丁卜 定址裝置300收到埠口 0的搜尋比對裝置2〇〇發出的搜尋信 號，就會依序連續發出128個位址信號，而在時間τ2，定址 裝置300收到埠口 }的搜尋比對裝置200發出的搜尋信號， 同樣自動依序連續發出128個位址信號。如第3b圖所示的匯 流排310的時序圖，在時間T1，定址裝置3〇〇收到埠口 〇的 搜尋比對裝置200發出的搜尋信號，就會依序連續發出ι28 個位址信號，而在時間T2 (當定址裝置300發出地址65的 信號時），定址裝置300收到埠口 1的搜尋比對裝置200發出 的搜尋信號，定址裝置300會記錄下來當時所發出的地址 13 1258282 (65)，然後在對應於埠口 0的叫固地址發送完成後，繼續 發送連續的地址信號直到所記錄下來的地址65為止，所以對 璋口 1而言’定址裝置300同樣為它的搜尋比對裝置2〇〇依 序連續發出128個位址信號，只是位址信號是從^到127、 再從0到65，而不是從〇〜127。 在本貫施例裡，存放查找表的記憶空間被分為二個區塊 • 彻、·，每個區塊可以容納128筆記錄（所以總共可以容 納2x128=256筆記錄）。每一筆記錄包含有多個搁位，而每一 個欄位包含有多個位元，如第4圖所示。其中，控制位元爛 位表示這筆記錄是不是「有效」的，也就是這筆記錄是空的 (尚未使用）、還是記錄有某_運算裝置的資料。控制位元也 可以用來圮錄這筆記錄是不是一筆常駐記錄，也就是它是否 會水遠停留在查照表裡。埠口號碼欄位則是記錄這個運算裝 • 置所在的網段是連接於這個交換裝置的哪一個埠口。MAC位 址攔位則是記錄這個運算裝置的MAC位址。時間註記（time stamp )欄位則是表示這筆記錄已經存在於查找表的時間長 短，通常當一筆有效的記錄停留的時間過久、可是又長時間 沒有使用到（可能是因為運算裝置故障、沒有開機、或是根 本離開了這個區域網路），交換器會將這筆記錄的控制位元修 改為「沒有使用」，以將這筆記錄的空間釋放出來存放其他、 新的運算裝置的資料。 14 1258282 請再參閱第2圖。定址裝置3〇〇的地址匿流排3i〇同時、 平行將輸出的地址信號供給記錄區塊彻和则，所以記錄 區塊400 # 口 500 f同時收到相同的地址錢。當收到地址信 號時，記錄區塊400 #口 500會平行地將其記錄區塊内、相對 應位址的記錄内容輪出到其各自的輸出匯流棑4丨〇和51258282 IX. Description of the invention: [Technical field to which the invention pertains] The present invention relates to network switching, and the singularity of the singer and the singer, especially for the network switching device to search for and determine the packet forwarding path. Circuit Architecture [Prior Art] Non-wired or wireless local area network 'Because of the length of the riding line and the coverage of the radio wave, there are theoretical and practical limitations. The regional network switch (four) is connected with each other. The wired or wireless network segment becomes the hub of the complete regional network. With the advancement of technology, the Wei of the (4)· is also increasing, in addition to supporting the other day, the other m integrated optical fiber, wireless transmission Interfaces and media, and even higher-level packet processing (such as load balancing provided by layer 3 s), are becoming more and more popular in the market. # <不什么疋的父Converter, its most basic function is still the so-called network packet forwarding (f〇rwardjng). Figure 1 is a schematic diagram of the general regional network architecture. As shown in the figure, it is changed to 10 The ports 0, 2, and 5 respectively connect three network segments 20, 30, and 40. The computer on the three network segments starts communication and exchanges data, and the hard or built-in initial body of the switch i is received according to it. The source and destination addresses in the packets sent by these computers (that is, the so-called Μ. bits, v of the second layer of the network establish a lookup table. This lookup table (not ") In the network address of each computer in the local area network, it is located in 6 1258282 which is connected to the network segment _ there is a network network layer 3 address (such as ίρ address), and other related Assume that the computer 21 wants to transmit data to the power supply (4) ^ and change the packet sent from the network segment 20 _ 21 to the computer 31 as the destination. By looking up the table, the exchange U knows that the computer 31 is located at the heart 2 The network segment 曰 ^ these packets are forwarded to the network segment 30 (instead of the network segment 2 〇 and 4 〇). Similarly, 〇 is sent to the computer 22, the switch H) is received from the network segment 2 由 by the computer 21, the data packet destined for the computer U, by looking up the table 'reader H), the computer 22 is located at the source of the source h _ network segment 2G'_1G knows that electricity (four) will also make _ packets, 疋 疋 abandon these packets, do not forward these packets out. With such a switch ^ 〇 使 网 〇 〇 〇 〇 〇 〇 〇 There will be no unnecessary packets in the 4th to affect the performance of the network. It should be noted that the above packet forwarding action of the switch 10 is time-limited, and the switch H) must be in the mouth. Before the next bit of the next-package, the completion of this pass is now intelligent, the packet (four) sends '(four) may have _ the situation will happen, the situation is: for the time is not ready for this currently received packet Disposal, and this: the package of my loss 'another kind of situation is to complete the disposal of this packet, and: the loss of the next - a few bits in front of the packet, will also cause the loss of the next - ^. Such a packet H, under the current conventional communication protocol, will cause the computer that sends the packet to resend the same-packet after a period of time, and the retransmitted 1258282 packet will consume the network bandwidth and the transmission time is extended. It also reduces the speed of data transmission. It is conceivable that the overall performance of the local area network will be significantly reduced when there are many situations in which the front and back packets "crash". 'As the switch network 埠σ, the number of connected computers increases, and the speed of the network is upgraded from HK)Mblts/see to delete (10), even the basic packet forwarding function of the 1' switch is also facing considerable challenge. Because the number of connected computers increases, the content recorded in the lookup table increases, and the time for searching the lookup table is inevitably prolonged. As the network transmission speed increases, the time for the switch to perform forwarding and the like is further shortened. Therefore, how to effectively search the lookup table becomes an important issue for (4)||R&D. The packet path search methods of the conventional switch mainly include three types: nyear search, hashing, and content addressable memory (CAM). The records of the lookup table using the linear search are arranged in a certain order, and the name can be imagined. The linear search is (4) the order of the comparison table (4) records of each computer until the MAC address of the destination of the packet is found. The brain records so far. This method is very simple, and the memory space used to store the lookup table can also be completely simplified. Linear search f is implemented in the form of a body, so in order to speed up the search to avoid packet collision, f needs a fast processor, which increases the cost of implementation. The hashing method is to use a hash function that is determined by the fact that the 8 1258282 hash function takes as input the M A C address of the packet, and converts the address of the record. When a computer's data is recorded in the lookup table, the hash function is used to determine which record is stored in the lookup table. When a packet is received to determine its forwarding path, the hash function is also used to calculate where the destination computer's record is located in the lookup table. Because only the -th order is needed, and the hash function is often implemented in a hardware manner, the hash search method is very fast. But because the hash function may calculate the same record address for different mac addresses, that is, the so-called hash collision "ashing coUision", this almost inevitable hash collision makes the memory of the lookup table" Can not be fully (some addresses may never be used), and the hash collision will make the records of multiple computers correspond to the same address - (10) in addition to the aforementioned - operations, but also add These indefinite numbers of computer records _ comparisons, causing uncertainty when searching. CAM's approach and hashing are similar to using the contents of the packet to calculate the address of a lookup table. CAM uses a complex hardware addressing method to achieve a unique address to avoid collisions, so the cam simultaneously It has the advantage of searching for Shaanxi speed: 14 charge 77 uses 5 memories, but the disadvantage of CAM is that the cost is too high. SUMMARY OF THE INVENTION The shortcomings of the search method based on the lookup table of the intelligent regional network switching device, the circuit architecture of the present invention, can fully benefit the memory space used for storing the lookup table in the switch. On the other hand, it can ensure that every search will not exceed the set time limit and cause the collision of the packet. The most important spirit of the present invention is not to pursue the shortest search time, but to use the "construction" of the flat 4 to ensure that the search time must be less than the predictability of the upper limit even in the worst case. With such a design, according to the present invention, (4) way switching ^ can be determined by calculation to use hardware components with appropriate performance, and not like some conventional techniques, because of the uncertainty of the search time, it is possible to not use it. High-performance hardware components are completely unnecessary. The network switch in accordance with the present invention can therefore be reasonably cost. Another object of the present invention is that the search method based on the proposed circuit architecture can be implemented in a hardware manner. The result of hardwareization can make the search speed faster, reduce the cost, and reduce the computational burden of other components. # The circuit _ proposed by the present invention divides the memory space of the button table into N blocks, and each block can store L pen records (so a total of NxL pen records can be stored). The N recording blocks (4) and the network switch's respective search ports are connected by separate bus bars (so there are a total of one bus bar). That is to say, through these bus bars, one record of one recording block can be output to all the search and comparison devices simultaneously and in parallel, and one search comparison device can simultaneously receive Nf from all the other § recorded areas. Block record. 10 1258282 The circuit architecture proposed by the present invention further comprises an addressing device. After being triggered by the search-matching device, the addressing device continuously generates addresses 0, 1, ..., L-1 signals simultaneously to all Ns. Record blocks. Each time an address 彳 § 5 is received, a recording block outputs the record of that address in parallel to all of the search aligning devices via its bus. Therefore, a search comparison device will sequentially receive all the notes recorded in the address of one record block, then all the records recorded in address 1 of the record block, and then all stored in the file. Recording block address pen records, and so on, a search comparison device will compare all NxL pen records after the addressing device has generated addresses 〇, i, ..., W u tiger's tamping cycle %. Therefore, in the worst case, the search lookup table time of one packet will not exceed the time of one cycle in which the addressing device generates L address signals. BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, the detailed description of the embodiments, and the scope of However, it is to be understood that the invention is not to be construed as being limited to the scope of the invention. The definition of the scope of the invention is within the scope of the invention. [Embodiment] The circuit architecture proposed by the present invention is applicable to the storage and search of address records in the switching device of the network super. The second layer (inclusive) of the network. The device includes a general second-layer area network switch. The so-called "switch" also includes a switch, router, or the like having the switching function of the 1258282 layer 2 or higher. The device's maximum 4 inch desire is to have multiple network ports and to determine whether the packets received from one of the ports should be discarded or forwarded to other beehives and whether they need to be recorded. The computing device that sends the packet is related information such as the network segment to which the bee port is connected. Fig. 2 is a diagram showing a preferred embodiment of the present invention applied to a second layer local area network. The proposed circuit architecture and the operation mode of the architecture may be implemented completely in a hardware manner, or may be implemented in the form of a part of a hard body, an early blade or a full blade. In this embodiment, mainly, The MAC 1 address of the ^ is the basis of the main search, but please understand the architecture of the present invention. The network switching device can be applied to the address other than the MAC address of other continents. One The other modules of the two-layer area network switch relating to the present invention are the same as those of the prior art, and therefore are omitted here. For those skilled in the art, # does not hinder their understanding of the main spirit of the present invention. As shown in Figure 2, the switch has eight network ports (numbered P〇P7). Each of the eight port interfaces in the switch is connected with a corresponding search and comparison device. When each packet is received, the packet is transmitted to the corresponding search comparison device through the connection 21〇. Please note that each port and the corresponding search comparison device 200 are shipped in parallel. For example, the port 2 is receiving the packet 12 1258282 of the network segment to which it is connected, and the port 5 may also receive the packet of the network segment to which it is connected. Search ♦ After the U 2GG receives the _ packet, it will 撷Take the important information in the packet, such as the address of the object (4), the MAC address of the packet destination, etc. The search comparison device 2 then sends out an address device 3 through the sink (10) row 220. a search signal. Please note that Because each port is operated in parallel, the button device may receive the search signals from more than one search comparison device at the same time, and the search signals do not interfere with each other. The address device 300 is receiving After the triggering of the search signal, the sequential output is also outputted to the busbar 3 1 。. In this embodiment, the addressing device j〇〇 sequentially generates address 0, address 丨, ..., address 127. 128 addresses ^. As shown in the timing diagram of the bus bar 31〇 shown in Fig. 3a, the search signal sent by the search comparison device 2 received by the time address receiving device 300 will be sequentially 128 address signals are continuously transmitted, and at time τ2, the addressing device 300 receives the search signal sent by the search device 200, and automatically sends out 128 address signals in sequence. As shown in the timing diagram of the bus bar 310 shown in FIG. 3b, at time T1, the addressing device 3 receives the search signal from the search device 200 of the port, and sequentially sends 284 address signals in sequence. And at time T2 (when the addressing device 300 issues the signal of the address 65), the addressing device 300 receives the search signal from the search device 200 of the port 1, and the addressing device 300 records the address 13 1258282 issued at that time. (65), then after the completion of the transmission of the fixed address corresponding to the port 0, the continuous address signal is continuously transmitted until the recorded address 65, so for the port 1 the addressing device 300 is also its search. The comparison device 2 successively issues 128 address signals in sequence, except that the address signals are from ^ to 127 and then from 0 to 65 instead of from 〇 to 127. In the present example, the memory space in which the lookup table is stored is divided into two blocks. • Each block can hold 128 records (so a total of 2x128=256 records can be accommodated). Each record contains multiple shelves, and each field contains multiple bits, as shown in Figure 4. Among them, the control bit ruin indicates whether the record is "valid", that is, whether the record is empty (not yet used), or whether the data of a certain _ computing device is recorded. The control bit can also be used to record whether the record is a resident record, that is, whether it will stay in the checklist. The port number field is the one on which the network segment where the computing device is located is connected to the switching device. The MAC address block is the MAC address of the computing device. The time stamp field indicates the length of time that the record already exists in the lookup table. Usually, when a valid record stays for too long, it is not used for a long time (probably because the computing device is faulty, The switch does not switch the control bit of the record to "not used" to release the space of the record to store the data of other, new computing devices. . 14 1258282 Please refer to Figure 2. The address block 3i of the address device 3 〇 simultaneously supplies the output address signals to the recording block in parallel, so that the recording block 400 #口 500 f receives the same address money at the same time. When the address signal is received, the recording block 400 #口 500 will rotate the recorded contents of the corresponding blocks in the recording block to their respective output confluences 4丨〇 and 5 in parallel.
上。匯流排41G和51G同時連接人個埠口的搜尋比對裝置 200，所以八個埠口的搜尋比對裝置2〇〇會同時、平行收到二 筆記錄’-筆來自記錄區塊4〇〇，另一筆來自記錄區塊·。 以第3a圖的時序圖為例。假設璋口 〇的搜尋比對裝置· 收到-個封包而要決定如何處理這個封包，崞口 q的搜尋比 對裝置200在時間τ；[斜 對疋址裝置300發出搜尋信號，定址裝 置300於是開始循序產生〇〜 127個地址信號給記錄區塊4〇〇 200會同時收到記錄區 和500，於是埠口 0的搜尋比對裝置 塊400位於地址〇的記錄、 貢料，然後是同時收到記錄 及§己錄區塊500位於地址1 塊400位於地址2的記錄、 以及記錄區塊5〇〇位於地址〇的 區塊400位於地址1的記錄、以 的資料’然後是同時收到記錄區 以及記錄區塊500位於地址2的 貝料依此犬員推。在收到各個記錄區塊來的記錄後，搜尋比 對衣置2GG會進行二項比對動作，其中之—是㈣包的來源 霞位址（也就是發出此封包的電腦的MAC位址）比對各 筆己錄内的MAC位址，如果經過所有記錄的比對都沒有發現 15 1258282 有付合的，這代表可能有一個新的運算裝置加入到區域網 路，這就需要透過連線23G通知交㈣置的其賴組（未圖 示）’在查找表裡增加-筆記錄。另外—個比對的動作是，用 封包的目的地M A C位址比對各筆記錄内的m A C位址，如果 發現有符合的MAC位址，就可依照比對符合的記錄裡的璋口 號碼攔位，判斷目的地其實就在這個埠口所連接的網段上， 如果是的話，就可透過連線23G通知其他模組，將這個封包 丟棄，不然就依照比對符合的記錄裡的埠口號碼欄位，通知 其他模組將封包轉送到那個埠口所連結的網段。 在這個例子裡，請注意到雖然料"的搜尋比對裝置 2〇〇啟動定址裝置綱來產生地址信號，但是所有其他璋口 的搜尋比對裝置200同樣會平行收到記錄區塊和$⑽所 輸出的所有記錄資料，只是所有其他埠口的搜尋比對裝置· 因為沒有㈣封包所以不會去進行比對。可是若以第儿圖為 例，在埠口 0的搜尋比對裝置還在進行比對時，埠口工 的搜尋比對裝置200在時間T2收到了 _個封包，它因此會對 定址裝置·發出搜尋信號，並且開始對記錄區塊彻和· 所送出的記錄資料加以比對，只是它開始比對時，是從位址 66開始，而不是從〇開始。 對於每-搜尋比對裝置而言，只要它發出搜尋信號 給定址裝置獅後，它就開始接收同時來自所有記錄區塊的 16 I258282 記錄資料’然後就可以對這些收到的記錄進行前述的封包來 源MAC位址、封包目的地MAC位址兩種比對，在最差的情 形，它最多重複128次這樣的收取、比對的動作，而在這樣 128次動作的循環後，搜尋比對裝置2〇〇 一定可以做出（1)封 包來源的MAC位址是否要記錄在查找表中、以及（2)對這個 封包是要丟棄、還是轉發到某一埠口的判斷出來。因此對於 • 實施本發明的區域網路交換裝置而言，它的硬體效能可以很 精確的設計成一定能在一個時限前完成這128次的接收比對 動作，因此可以避免封包衝撞的問題。 對於-·個搜尋比對裝置200而言，如果它在這128次的 接收比對動作完成前，就已經可以做出判斷出來的話，它可 以不待全部128次的接收比對動作結束，就指示其他的模組 先去進行相關的處置。舉例來說，假設它在比對所有記錄區 帽| 塊位址66的記錄時，就發現了有一筆記錄有符合封包目的地 的MAC位址’這時就可以指示其他模組依據這個記錄的内容 丢棄或轉發封包，而接下來就不需要再比對封包目的地的 MAC位址，可是封包來源的MAC位址的比對還是繼續進行。 假设到位址88時，比對發制—筆記錄有符合的mac位址， 由於已經知道這筆記錄的地址，如果有需要的話，搜尋比對 衣置200可以直接調整這筆記錄的内容，像是更改埠口號碼 (比如况，廷個機器換接了另一個網段）、時間註記等。請注 17 1258282 思到由於記錄區塊都還在存取中，搜尋比對裝置200對這筆 5己錄的修正會需要經過記錄區塊的仲裁（arbitration )。接下 來，對於尚未比對的其他地址，一些實施例是容許定址裝置 300知績元成整個循環，而在另外一些實施例裡，搜尋比對 I置200會發出另外的信號停止定址裝置3〇〇繼續產生地址 信號。 φ 對於一個搜尋比對裝置200而言，如果它在這128次的 接收比對動作完成後，仍然比對不到封包來源MAC位址、或 封包目的地MAC位址時，對於後者，交換裝置只有將封包轉 發到該封包進入的埠口以外的其他所有埠口。至於前者，這 代表可此有一個新的運算裝置加入到區域網路，這需要在查 找表裡增加一筆記錄。由於在之前128次的比對過程中，搜 才比對U 200經由記錄的控制位元搁位判&，已經知道記 修 錄區塊裡哪些記錄是閒置的。所以搜尋比對裝置細可以直 接經由記憶區快的仲裁，將資料寫入一筆記錄裡。 由於本發明彳闕大量的平行運算，所以很可能二個甚至 更夕搜寻比對裳置2〇〇同時欲寫入同一筆空閒的記錄。解決 、言衝大的方式不一’比如說，各個埠口可以被賦予某種優 先順序，像是埠° G優先於埠口 1、埠Π 1優先於埠口 2、.··。 阜K的搜尋比對裝置200要寫入一筆記錄時，它會 先確疋埠Π G、1、·..、κ]都沒有要寫人同_記錄後才會進 18 1258282 订更夕的細即以及其他的做法這裡就不予資述。不過從以 上的敘述可以看出’整個記憶空間裡的筆記錄都有可能 ' Q此不會有採用雜凑方式的浪費記憶空間的問題。on. The bus bars 41G and 51G are simultaneously connected to the search and comparison device 200 of the person's mouth, so the search and comparison device 2 of the eight ports will receive two records simultaneously and in parallel '- pen from the recording block 4〇〇 The other one comes from the record block. Take the timing diagram of Figure 3a as an example. Assuming that the search device compares the device to receive a packet and decides how to process the packet, the search device 200 of the port q is at time τ; [the oblique address device 300 sends a search signal, and the addressing device 300 Then, the sequential generation of 〇~127 address signals to the recording block 4〇〇200 will simultaneously receive the recording area and 500, so the search of the port 0 is located at the address 〇 record, the tribute, and then simultaneously The received record and the logged block 500 are located at the address 1 block 400 at the address 2 record, and the record block 5 〇〇 at the address 〇 block 400 is located at the address 1 record, the data 'and then received simultaneously The recording area and the recording block 500 located at address 2 are pushed by the dog. After receiving the records from each record block, the search for the matching 2GG will perform a two-matching action, where is the source address of the (four) packet (that is, the MAC address of the computer that sent the packet) Comparing the MAC addresses in each record, if the comparison of all the records does not find that 15 1258282 has been paid, this means that a new computing device may be added to the local area network, which requires a connection. The 23G notice (4) set its own group (not shown) 'in the lookup table to add - pen records. In addition, the action of the comparison is to compare the destination MAC address of the packet with the m AC address in each record. If a matching MAC address is found, the comparison can be made according to the comparison. The number is blocked, and the destination is actually located on the network segment to which the port is connected. If yes, the other modules can be notified through the connection 23G, and the packet is discarded, otherwise it is in accordance with the matching record. The port number field is used to notify other modules to forward the packet to the network segment to which the port is connected. In this example, please note that although the search " search device 2 initiates the addressing device to generate the address signal, all other ports of the search device 200 will also receive the recorded blocks in parallel and $ (10) All the recorded data is only the search and comparison device of all other ports. Because there is no (4) packet, it will not be compared. However, if the first map is taken as an example, when the search comparison device of the mouthpiece 0 is still being compared, the search and comparison device 200 of the mouthpiece receives _ packets at time T2, and thus it will address the device. A search signal is issued, and the recorded data of the recorded block and the sent data are compared, except that it starts from the address 66, rather than starting from the address. For each-searching device, as soon as it sends a search signal to the addressing device lion, it begins to receive 16 I258282 records from all recorded blocks and then can perform the aforementioned packets on these received records. The source MAC address and the packet destination MAC address are two kinds of comparisons. In the worst case, it repeats up to 128 such collection and comparison actions, and after such a cycle of 128 operations, the search comparison device 2) It is possible to make (1) whether the MAC address of the packet source is to be recorded in the lookup table, and (2) whether the packet is to be discarded or forwarded to a certain port. Therefore, for the regional network switching device embodying the present invention, its hardware performance can be accurately designed to complete the 128 comparison comparison actions before a time limit, thereby avoiding the problem of packet collision. For the search matching device 200, if it can make a judgment before the 128 times of the comparison comparison operation is completed, it can indicate that the reception comparison operation is completed after all 128 times. Other modules go to the relevant disposal first. For example, if it compares the records of all the record area caps|block addresses 66, it finds that there is a record with a MAC address that matches the destination of the packet. Then, it can instruct other modules to rely on the contents of this record. The packet is discarded or forwarded, and then there is no need to compare the MAC address of the packet destination, but the comparison of the MAC address of the packet source continues. Suppose that when the address is 88, the matching fax-pen record has a matching mac address. Since the address of the record is already known, if necessary, the search for the clothing 200 can directly adjust the content of the record, like It is to change the port number (such as the situation, the machine is switched to another network segment), time note, and so on. Please note 17 1258282 that since the recording block is still being accessed, the search comparison device 200 will need to go through the arbitration of the recorded block for the correction of the record. Next, for other addresses that have not been compared, some embodiments allow the addressing device 300 to perform the entire cycle, while in other embodiments, the search for the I set 200 will signal additional addressing to stop the addressing device. 〇 Continue to generate the address signal. φ For a search comparison device 200, if it is still less than the packet source MAC address or the packet destination MAC address after the 128 reception comparison operations are completed, for the latter, the switching device Only forward the packet to all ports other than the port that the packet entered. As for the former, this means that a new computing device can be added to the local area network, which requires adding a record to the lookup table. Since the search was compared to the U 200 via the recorded control bit in the previous 128 comparisons, it is known which records in the record block are idle. Therefore, the search comparison device can directly write the data into a record directly through the arbitration of the memory area. Since the present invention has a large number of parallel operations, it is very likely that two or even the same day search will be placed at the same time to write the same free record. For example, the individual mouths can be given a certain priority order, such as 埠 ° G is preferred over 埠 1, 埠Π 1 is preferred over 埠 2, . When 搜寻K's search comparison device 200 wants to write a record, it will first confirm that G, 1, ..., κ] have no written _ record, then it will enter 18 1258282. Fine and other practices are not covered here. However, from the above description, it can be seen that the pen record in the entire memory space is possible. Q This will not have the problem of wasting the memory space by using the hash method.
上面的解釋並未詳細說明搜尋比對裝置雇、以及定址 裝置300是如何實施的。不過對於熟習相關技藝人士而言， 這些裝置的實施都相當簡單，比如說，定址裝置·基本上 就是-個計數器（_如），而搜尋比對裝置2〇〇可能包含有 夕们曰存mU存放從每—個區塊傳來的記錄資料，而搜尋比 對裝置200可能是以簡單的勤體對這些暫存器來進行線性比 對、或是以完全硬體的方式來進行比對。在另外—些實施例 裡’ «比«置·是與埠σ的界面電路結合—起，而非 分別獨立之電路。 藉由以上較佳具體 述本創作之特徵與精神 施例來對本創作之範疇 涵蓋各種改變及具相等 圍的範®壽内。 貫施例之詳述，係希望能更加清楚描 ，而亚非以上述所揭露的較佳具體實 加以限制。相反地，其目的是希望能 性的安排於本創作所欲巾請之專利範 【圖式簡單說明】 之網路連線示意 第1圖係一習知之第二層區域網路交換器 圖 弟2圖係本發明—較佳實施例應用於_第二層區域網路交換 19 1258282 器之示意圖。 第ja 3b圖‘刀別頌示依據本發明一較佳實施例之定址裝置 於兩種情境下之時序圖。 第4圖係依據本發明一較佳實施例之查照表記錄之示意圖。 【主要元件符號說明】The above explanation does not specify how the search comparison device is employed and how the addressing device 300 is implemented. However, for those skilled in the art, the implementation of these devices is quite simple. For example, the addressing device is basically a counter (_), and the search device 2 may contain a mU. The record data transmitted from each block is stored, and the search comparison device 200 may perform linear comparison on these registers with a simple physical body or in a completely hardware manner. In other embodiments, the ««« is combined with the interface circuit of 埠σ, rather than separate circuits. The scope of this creation encompasses various changes and equal scopes within the scope of the creation and the spirit of the present invention. The detailed description of the embodiments is intended to be more clearly described, and Asian and African are limited by the preferred embodiments disclosed above. On the contrary, the purpose is to hope that the ability to arrange in the creative scope of the patent application [simplified description of the network] shows the first picture is a second-level regional network switch 2 is a schematic diagram of the present invention - a preferred embodiment applied to a second layer regional network exchange 19 1258282. A diagram of the timing diagram of the addressing device in accordance with a preferred embodiment of the present invention in two scenarios is shown in Fig. 3b. Figure 4 is a schematic illustration of a lookup table record in accordance with a preferred embodiment of the present invention. [Main component symbol description]
P0〜7 網路埠口 10 交換器 20 網段 21 電腦 22 電腦 30 網段 31 電腦 40 網段 200 搜尋比對裝置 210 與埠口之間的連線 220 與定址裝置相連的匯流排 230 與其他模組的連線 300 定址裝置 310 與記錄區塊相連的匯流排 400 吕己錄區塊 410 記錄區塊之輸出匯流排 500 記錄區塊 510 記錄區塊之輸出匯流排 T1 時間點 Τ2 時間點 20P0~7 network port 10 switch 20 network segment 21 computer 22 computer 30 network segment 31 computer 40 network segment 200 search connection device 210 and port connection 220 bus bar 230 connected to the addressing device and other Module connection 300 Addressing device 310 Busbar connected to the recording block 400 Lvhexun block 410 Recording block output busbar 500 Recording block 510 Recording block output busbar T1 Time point Τ2 Time point 20