五、發明說明⑴ ~ " ~ 本發明係知供一種適應型多層位址雜散表管会方法， 特係指用於網路交換器設計的一種適應型多層位址雜散 表管理方法_。 按，一般的區域網路交換器係以封包的 定傳輪路徑，如第-圖所示，它有個過渡資料庫“^ 有連結的終端機之M A c位址及其連結的交換埠碼 一二；斗，依據該過瀘資料庫i i的資料，交換器即可藉由 過程1 2來傳送封包或是將其過濾、，此外，該過據 =;、庫1 1内的資料係可藉由一來源位址學習過程來 更改。 .. ^ 承上所述’用以作為過濾資料庫Λ i者，例如多岸 址雜散表，其係為一有效率的過濾資 = 比單層位址 12或是來源位址學習過管是前送過程 ^ ^ 巧担丄d搜哥—個目的位址戋熹奴 個來源位址，均須讀（RAM)N欠 义=于 須搜尋整個N層位址雜散表，因 換σ之，即是 寬，且延後了傳送訊息的^ = Μ报大的RAM頻 有鑑於上述缺點，是以，本良 驗，積極從事研究，終有本‘明f年從事該行業之經 管理方法』之產生。 、應型多層位址雜散表 本發明之主要目的係提供一 雜散表中搜尋位址的適應型夕a 有政率地在多層位址 *，本發明之主要特徵；雜：表管理方法。 、藉由一來源位址學習過 Ί* C：\#^\87〇359.ptd 五、發明說明（2) -- 程、及一過時剔除過程來自動調整一 N層位址雜^表的有 效層數nn (即存有位址資料的層數，N>=nn )，使— 位址學習過程、前送過程、及過時剔除過程只在存有位址 資料的nn個有效層中搜尋特別的位址，而無需於所 位址雜散表中搜尋特別的位址，因此可有效 : 址雜散表中搜尋位址。 在夕層位 麦是’為達到上述之目的，本發明大體上包括 一 ί 過程、一前送過程、及一過時剔除過程，該 ^ ，、/、在存有位址資料的ηη個有效層中搜尋別 址，而該多層層），址雜散表的有效層數二：； 位址資料的層數，Ν> = ηη)係由上述來源位:^子有 係包含有： 動調整其中，該來源位址學習過輕 1 )當上述網路交換器由埠收到一筆資料時， Γ位ΐί雜散映射，而對應到⑽個R AM的地址？且^ 出； 地將母地址中的位址資料讀 2 收到來源位址分別與上述所讀出位址斗 右，、係付合所讀出位址資料中之一，’對， 料;若不符合時，則表示其係為一新的；：;立：的時間資 該新的來源位址寫入位址雜散表中，’、思因此將 雜散對應位置為空時，該新的來源i址二！：之〜 址，否則則將其窝n 則寫入該層位 多層位址雜I主 雜政對應位置，此時、+、 層位址雜散表的有效層數即為nn + 1 . 1上迭V. Description of the invention ~ " ~ The present invention is known for an adaptive multi-level address stray table management method, and particularly refers to an adaptive multi-level address stray table management method for network switch design. . Press, the general LAN switch is a fixed route of the packet, as shown in the figure-it has a transition database "^ MA c address of the connected terminal and its connected switch port code One or two; bucket, according to the data in the database ii, the switch can send packets or filter them through the process 12; in addition, the data = ;, the data in the library 1 1 can be Changed by a source address learning process .. ^ Continued from the above, 'used as a filtering database Λ i, such as a multi-site address stray table, which is an efficient filtering resource = than a single layer The address 12 or the source address learning process is a forward process ^ ^ Cleverly 丄 d search brother-a destination address 戋 熹 slave source address, all must read (RAM) N unsense = you must search the entire The N-level address stray table is changed to σ, that is, wide, and delayed ^ = Μ reported by the large RAM frequency. In view of the above shortcomings, this good experience, actively engaged in research, and finally have The emergence of the 'Managed Method of Engagement in the Industry in the Ming Dynasty'. The Application of Multi-level Address Spur Table The main purpose of the present invention is to provide a The adaptive type of searching for addresses in a hash table is a multi-layered address *, which is the main feature of the present invention; Miscellaneous: a table management method.. Learned from a source address. * C: \ # ^ \ 87〇359.ptd V. Description of the invention (2)-Process and an outdated elimination process to automatically adjust the effective number of layers of an N-level address miscellaneous table nn (that is, the number of layers where address data is stored, N > = nn), so that-the address learning process, forward process, and outdated removal process only search for special addresses in the nn effective layers where address data is stored, without searching for special addresses in the address hash table. Therefore, the address can be effective: search the address in the address stray table. In order to achieve the above purpose, the present invention generally includes a process, a forward process, and an obsolete elimination process. The ^ ,, /, searches for other addresses in the ηη effective layers where address data is stored, and the multi-layer layer), the effective number of address stray tables is two :; the number of layers of address data, N > = ηη ) It is from the above source bit: ^ The sub-system contains: It is adjusted automatically, the source address is too light to learn 1) When the above network exchange When the converter receives a piece of data from the port, the Γ bit is stray map and corresponds to the address of one RAM? And ^ out; the address data in the parent address is read 2. The received source address is different from the above. The read address is right, and it is one of the address data read by Fuhe, 'Yes, data; if it does not match, it means that it is a new one; The source address is written into the address stray table, and when the corresponding position of the spur is empty, the new source i is the second address:! ~~ address, otherwise its nest n is written into the layer Corresponding position of the main multi-level address miscellaneous I, at this time, the number of effective levels of the hash table of the +, multi-level address is nn + 1.1.
<^:\專利\870369.卩七(1 第5頁 五、發明說明（3) 該前送過程係包含有 1 )當上述網路交換器由埠 技次"士 的位址經雜散映射，而對應2資料時’冑其中的目 自⑽層中的每一層，一一 的地址’且分別 出； 將母—地址中的位址資料讀 2 )將所收到目的位址分別與上述 若其係符合所讀出位址資料中之—， 」比對， )由記錄在符合的位址資料中的交換埠傳出、若：j封包 時，則表示其係為一未知的目位、 ’ 付合 播出去或是丟棄、或傳予c p U ’因此將該筆資料廣 而該過時剔除過程係包含有： 1 )分別自η η層中的每一層，一一 =出，並檢視其是否過期，若以 —1 ，亦即於夕尨七、昨双衣的有效層數即為nn 再搜尋第nn層。，"、位址學習過程、前送過程中無須 發明之適應型多層位址雜散表管理方法 决速有效地搜尋多層位址雜散表。 乃疋 =關本發明為達上述目的、特徵所採用 其功效，茲例舉較佳音 的技術手長及 第-圖係-般2 圖式說明如下： 瓜乂換盗的學習過程與前送過程的示意圖。. 五、發明說明（4) =二圖係本發明之實施例的簡單方塊說明圖 ί二圖係表示本發明之N層雜散表的示意圖。 第四圖係為本發明之來源位址學習過程2 1的流程圖 f五圖係為本發明之前送過程2 2的流程圖。 第六圖係為本發明之過時剔除過程2 3的部份流程圖 第七圖係為本發明之最大有效層數nn的增滅調整流释 圖號說明 11 :過 遽 資 料 庫 12 前 送 過 程 13 來 源 位 址 學 習過程 2 1 來 源 位 址 學 習過程 2 2 前 送 過 程 2 3 過 時 剔 除 過 程 3 雜 散 映 射 首先，&第一圖所#，本發明大體上係包括有：-來 源位址學習過程21、-前送過程22、1!過時剔除過 程2 3，該等過程俜口 a + ^ . # & 糸/、在存有位址資料的nn個有效層中搜 卞’二J :址’ =β亥多層（N層I位 (即存有位址資料的層數，狀衣的'外 習過程、及過時剔除過係由上述來源位址學 學習過程21係當上述網ϋ調整’其中，該來源位址 到-筆資料（S Α/d广=換器（未示於圖中）由谭收< ^: \ Patents \ 870369. 卩 VII (1 page 5 5. Description of the invention (3) The forwarding process includes 1) When the above network switch is operated by the port technology " Hash mapping, and corresponding to the 2 data, 'the target in the layer from each layer in the layer, one by one address' and separately; read the address data in the parent-address 2)) respectively received destination address If it is in accordance with the above, if it is in accordance with the read address data,-"" comparison,) is sent from the exchange port recorded in the corresponding address data, if: j packet, it indicates that it is an unknown The position, 'pay out or discard, or pass to cp U'. Therefore, the process of widening the data and the outdated rejection process includes: 1) from each of the η η layers, one by one = out , And check whether it is expired. If -1 is used, that is, the number of effective layers of Yu Xiqiu and Yesterday's double clothing is nn and then search for the nn layer. &Quot; There is no need to invent an adaptive multi-level address stray table management method in the process of address learning and forwarding. It is necessary to efficiently search for multi-level address stray tables. Nai = The function of the present invention to achieve the above-mentioned objectives and characteristics, for example, the technical skills of the better sound and Figure-General 2 schematic illustrations are as follows: schematic diagram. V. Invention Description (4) = The second diagram is a simple block diagram of an embodiment of the present invention. The second diagram is a schematic diagram showing the N-layer spur table of the present invention. The fourth diagram is a flowchart of the source address learning process 21 of the present invention f The fifth diagram is a flowchart of the process 22 before the present invention is sent. The sixth diagram is a partial flowchart of the obsolete elimination process 2 3 of the present invention. The seventh diagram is the increase and decrease adjustment of the maximum effective layer number nn of the present invention. 13 Source address learning process 2 1 Source address learning process 2 2 Forwarding process 2 3 Out-of-date rejection process 3 Stray mapping First, the present invention generally includes:-Source address learning Process 21,-Forward process 22, 1! Out-of-date rejection process 2 3, these processes pass a + ^. # &Amp; 卞 /, search for 'second J' in the nn effective layers where address data is stored: '= β Hai multi-layer (N layers I bit (ie the number of layers where address data is stored, the' outside learning process, and outdated rejection are determined by the above-mentioned source address science learning process 21 when the above network is adjusted ') Among them, the source address to the data (S Α / d 广 = converter (not shown in the figure) by Tan Shou
Α )時，將其中的來源位址（S AΑ), the source address (S A
C:\專利\870369.ptd 發明說明（5) 五 \〇j — j雜散映射3，而對應到nn個R A M的地址，£分 的每-層，--地將每-地址中的位址資料讀出 甚並=斤收到來源位址分別與上述所讀出位址資料比對， 料所讀出位址資料中之-，貝|!更新其中的時間資 兮靳符合時，則表示其係為一新的來源位址，因此將 Π:ί::址寫入位址雜散表中，此時，若㈣中= 址，否目丨應位置為空時’該新的來源位址則寫入誃展办 多芦：:f將其寫入ηη +1層的雜散對應位置，此：層 夕s位=雜散表的有效層數係為⑽+丄； 匕時上述 料系：ii網路交換器由埠收到-筆資 昧射q )夺將其中的目的位址（D A ) 4 、 戽射3，而對應到nn個R AM的地址，且別）經雜散 每—層，一一地將每一地址中的位址資料讀自卟層中的 ::立址分別與上述所讀出位址資」J巧: 筆資料(封包 盆# Α 土 ★ 換璋傳出；若不符合時，目，丨本- ::、為-未知的目的巍，因此將該筆資 2表不 丟棄、或傳予CPU (未示於圖中）； 、出去或是 :該過時剔除過程2 3係分別自η η層中的 成 :2將母-地址中的位址資料讀出，並檢：盆曰：層，一 右為過期則將其自所屬層中剔除；若 显/士疋否過期， 資料均已剔除，則將該第ηη層歸為備用j : ^所有位址 位址雜散表的有效層數即為n〜^ 上述多層 址學習過程…前送過程22二::以位 C:\ 專利\870369.ptd 第8頁C: \ Patents \ 870369.ptd Description of the invention (5) Five \ 〇j — j stray mapping 3, and corresponding to nn RAM addresses, each minute of the sub-level, and every bit of the-address The read-out of the address data is equal to the comparison between the received source address and the above-mentioned read-out address data. It is expected that when the read-out address data is-, the || Indicates that it is a new source address, so Π: ί :: address is written into the address stray table. At this time, if ㈣ 中 = address, if the destination is empty, the new source should be empty. The address is written in 誃 展 多 :: f to write it to the corresponding position of the spur on the ηη +1 layer. Here: the effective number of s bits = spur table is ⑽ + 丄; Material: ii network switch received by the port-the penny ray q) captures the destination address (DA) 4 and ray 3, and corresponds to nn R AM addresses, and other) through miscellaneous Disperse each layer, and read the address data in each address one by one from the porch layer :: standing address and the address information read out above "Jqiao: pen data (封包 盆 # Α 土 ★ change璋 Outgoing; if not, head, 丨 this-:: , For unknown purpose, the form 2 is not discarded, or passed to the CPU (not shown in the figure); go out or: the outdated rejection process 2 3 is the result from the η η layer. : 2 Read out the address data in the parent-address and check: basin: layer, the one on the right is out of date, then it will be removed from the layer; if the data is removed, it will be deleted. The ηηth layer is classified as a spare j: ^ The effective number of layers for all address address stray tables is n ~ ^ The above-mentioned multi-layer address learning process ... forwarding process 22 :: bit C: \ patent \ 870369.ptd Page 8
五、發明說明（6) 以下將以第三 本發明之來源 過時剔除過程 最大層數係令 每一層的雜散 散對應係設為 示第二層i位 第四圖係 圖。如該圖所 數m令為1、 中s買取一hash 斷該處是否存 中判斷S A是 步驟S 2 1 〇 料，若不符合 外，若資料不 參數X為〇時 中，將參數m 是否大於最大 則於步驟S 2 〇，則將所接 數X並不大於 數nn是否等於 均以被使用（ 圖〜第七圖中所示之流程圖來具體說明有 位址學習過程2 1、一前送過程2 2、及一 2 3的實際運作。在此’多層位址雜散表之 為N ( N > 1 )、最大有效層數係令為nn、 表大小係設為L、每一 s A (或D A )的雜 i。如第三圖所示，hash table(2)(i)即表 址0 為本發明之來源位址學習過程2 1的泠兹 示，其係於步驟S2100中先將=參 且將參數X令為〇 ;之後在步驟S 2 1 〇 1 table(m)(i);並於步驟3 2工〇 2中判 在資料；若資料存在，則於步驟s 2 1 〇 4 否符合所讀出的MA C位址，若符合，則於 5中更新hash tab 1 e( m) (i )中的時間資 ’則進行步驟S 2 1 Ο 6以後之步驟；此” 存在，則進行步驟S 2 1 〇 3，該步驟係當 ’將其值令為m ;之後於步驟s 2 1 〇 6 加一；然後在步驟S 2 1 0 7中判斷參數m 有效層數nn，若其係大於最大有效層數nn， 1 0 8中判斷參數x是否大於〇，若是大於 收的S A窝入hash table(x)(i)中，若參 〇 ’則於步驟S 2 1 1 〇中判斷最大有效層 N ’若是等於N則表示所有的位址雜散表層 步驟S 2 1 1 2 )，若是有效層數nn不等於V. Description of the invention (6) In the following, the source of the present invention is the third. The obsolete elimination process. The maximum number of layers is the order of the spurious correspondence of each layer. As shown in the figure, m is set to 1. Buy a hash in s to judge whether SA is stored in step S 2 1 〇 If not, if the data does not match the parameter X is 0, whether the parameter m If it is larger than the maximum value, then in step S 2 〇, whether the number X received is not greater than the number nn is equal to be used (the flow chart shown in Figure ~ the seventh figure to specifically explain the address learning process 2 1, a The actual operation of the forward process 2 2 and 1 2 3. Here, the multi-level address stray table is N (N > 1), the maximum effective layer number is set to nn, the table size is set to L, each A s A (or DA) is miscellaneous i. As shown in the third figure, hash table (2) (i), that is, table address 0 is a source list of the source address learning process 21 of the present invention, which is based on steps In S2100, first set = parameter and set the parameter X to 0; then, in step S 2 1 0 table (m) (i); and judge the data in step 32 2; if the data exists, then in step s 2 1 〇4 Does it match the read out of the MA C address, if it does, then update the time data in hash tab 1 e (m) (i) in 5 ', then proceed to step S 2 1 〇 6 This "exists, then Step S 2 1 〇3, this step is when the value is set to m; then in step s 2 1 〇6 plus one; then in step S 2 1 07 to determine the effective number of parameters m layer nn, if it If it is greater than the maximum effective number of layers nn, it is determined whether the parameter x is greater than 0. If it is greater than the received SA, it is nested in the hash table (x) (i). If it is 0, it is judged in step S 2 1 1 〇 The maximum effective layer N 'is equal to N, which means all address stray surface steps S 2 1 1 2), if the effective layer number nn is not equal to
CA 專利 \870369.ptd 第9頁 五、發明說明（7) N ’則於步驟S2Ul將SA寫入hash … ϊ驟321 〇7中，若參數瓜並不大於 $ίί ，則重複進行上述步驟s 2 1 0 1以下 第五圖係為本發明之前送過程2 2的流程圖。 ?所不,:ΐ係於步驟S 2 2 0 0中先將所需之參數m 驟3 2 2 〇 1 中讀取一hash table(m) .1 / ? ^ S 2 2 〇 2中判斷該處是否存在資料；若 f料存在，則於步驟s 2 2 α 3中判斷所接收D A是否符 1所讀出的Μ A^位址，若符合，則將該封包由記錄在符 〇的MA C位址資料中的交換槔傳出（步驟s 2 2 〇 7 )，若不符合，則進入步驟s 2 2 〇 4中，並使參數爪= m + 1 ;之後再於步驟s 2 2 〇 5中判別參數爪是否大於 最大有效層數nn，若參數m大於最大有效層數nn，則表―示 所接收D A係為一未知的目的位址（步驟s 2 2 〇 6 )， 此時可將此封包廣播出去或是丟棄、或傳予c p u 數：未大於最大有效層數nn，則再度返回 /驟S 2 2 0 1，並繼續以下之步驟。 ^六圖係為本發明之過時程 圖，其係用以檢視第nn層县不1晚达扯^ =幻司伤伽·程 如該圖所示，其係於步驟： = 的流程圖。 令為1，·之後在步驟sHo先需之參數1 (k) ; s 2 3 0 2 tV^VilSVable(nn) ^〒判斷該處疋否存在資料；若 五、發明說明（8) Πί步程序（步驟S 2 3 〇 3 )，若*料不存 驟3 2^1。，323 04中，令參數1^ = 1^+1;並於步 於步驟S 斷參數k是否大於1，若是大於L，則 ηη-ι)，若是中’令最大有效層數為1111—1 (即^ 繼續”之以於L，則再度返回步驟”301，並 在此，值得一提的是，於上述來源位址學習過程2 1如ί 2 ί剔除過程2 3中，最大有效層數nn的增減調整 = 其中，最大有效層數-之起始值係為1 0)，當最大有效層數nn須增加一層時（步 耸^，則先判斷nil是否等於N (步驟2 4 〇 2 右寺於N，則返回步驟2 4 〇 1 ，若等 n;;n(n；i2(:r4〇3) ，則先判斷nn是否等於1 (步驟2 !，則inn:nn ::返回步驟2401，若不等於， h nn — 1 (步驟 2 4 0 6 )。 層數 nn 學習過程、及過時剔除過程來自動i整故哕ί來源位址 I過程、前送過程、及過時剔除過程可J =位址學 料的nn個有效層申搜尋特別的位 =了，、在存有位址資 型多層位址雜散表管理方法 ^，此，本發明之適應 雜散表。 了决逮有效地搜尋多層位址 综上所述 本發明之 適應型多層位址雜散表管理方CA Patent \ 870369.ptd Page 9 V. Description of Invention (7) N 'writes SA into the hash in step S2Ul ... Step 321 07, if the parameter is not greater than $ ίί, repeat the above steps s The fifth figure below 2 1 0 1 is a flowchart of the sending procedure 2 2 before the present invention. No ,: In step S 2 2 0 0, the required parameter m is first read from a hash table (m) in step 3 2 2 〇. 1 /? ^ S 2 2 〇 2 to determine the Whether the data exists; if the f data exists, then in step s 2 2 α 3, it is judged whether the received DA matches the M A ^ address read from 1. If it matches, the packet is recorded by the MA recorded in the code 0. The exchange address in the C address data is transmitted (step s 2 〇 7). If it does not match, the process proceeds to step s 2 〇 4 and sets the parameter claw = m + 1; then, step s 2 2 〇 In step 5, it is judged whether the parameter claw is greater than the maximum effective layer number nn. If the parameter m is greater than the maximum effective layer number nn, it indicates that the received DA system is an unknown destination address (step s 2 2 06). Broadcast this packet or discard it, or pass it to the number of CPUs: if it is not greater than the maximum effective number of layers nn, then return / step S 2 2 0 1 again, and continue the following steps. ^ The six pictures are the obsolete time chart of the present invention, which is used to view the nn-level county not arriving at the airport for one night. ^ = Magical Division injury ga · Cheng As shown in the figure, it is a flowchart of steps: =. Let it be 1, and then the parameter 1 (k) required in step sHo; s 2 3 0 2 tV ^ VilSVable (nn) ^ 〒 judge whether there is data at this place; if five, description of the invention (8) Πίstep procedure (Step S 2 3 〇3), if * is not expected, step 3 2 ^ 1. In 323 04, let the parameter 1 ^ = 1 ^ + 1; and then, in step S, determine whether the parameter k is greater than 1, and if it is greater than L, then ηη-ι); if it is medium, let the maximum effective layer number be 1111-1 (That is, ^ continued "is the same as L, then return to step 301 again, and it is worth mentioning that in the above source address learning process 2 1 such as ί 2 ί rejection process 2 3, the maximum number of effective layers nn increase / decrease adjustment = Among them, the maximum effective number of layers-the initial value is 10), when the maximum effective number of layers nn must be increased by one layer (step ^, first determine whether nil is equal to N (step 2 4 〇 2 If the right temple is at N, return to step 2 4 〇1, if you wait for n ;; n (n; i2 (: r4〇3), then first determine whether nn is equal to 1 (step 2 !, then inn: nn :: return Step 2401, if not equal, h nn — 1 (step 2 4 0 6). The number of layers nn learning process, and out-of-date rejection process to automatically rectify the source address I process, forward process, and out-of-date rejection process. However, J = search for special bits of nn effective layers of address materials, the management method of multi-level address stray tables with existing address data types ^, so the invention adapts to stray tables. Effectively address the multilayer summary adaptive search of the present invention the multilayer side address management table stray