公开(公告)号：US06799207B1

公开(公告)日：2004-09-28

申请号：US09546343

申请日：2000-04-10

申请人： Everett Arthur Corl, Jr. ,  Anthony Matteo Gallo ,  Seeta Hariharan ,  Colin Beaton Verrilli

发明人： Everett Arthur Corl, Jr. ,  Anthony Matteo Gallo ,  Seeta Hariharan ,  Colin Beaton Verrilli

IPC分类号： G06F15173

CPC分类号： H04L45/00 ,  H04L29/06 ,  H04L45/563 ,  H04L67/34 ,  H04L69/22 ,  H04L69/329

摘要： A method and system for downloading software managed trees (SMTs) in a network processing system provides dynamic update of frame classifiers while maintaining proper network protocol processing. The network processing system includes a general purpose processor acting as control point processor and a plurality of network processors. The new SMT is built by an application on the control point processor and downloaded to one or more of the network processors. The new SMT is placed in a separate memory location accessible to the network processors, rather then overwriting the existing SMT. The active tree pointers are then changed to transfer control to the new SMT.

摘要翻译： 在网络处理系统中下载软件管理树（SMT）的方法和系统提供帧分类器的动态更新，同时保持适当的网络协议处理。 网络处理系统包括充当控制点处理器的通用处理器和多个网络处理器。 新的SMT由控制点处理器上的应用程序构建并下载到一个或多个网络处理器。 新的SMT放置在网络处理器可访问的单独内存位置，而不是覆盖现有的SMT。 然后更改活动树指针，将控件转移到新的SMT。

公开(公告)号：US06484171B1

公开(公告)日：2002-11-19

申请号：US09540333

申请日：2000-03-31

申请人： Everett Arthur Corl, Jr. ,  Anthony Matteo Gallo ,  Clark Debs Jeffries ,  Colin Beaton Verrilli

发明人： Everett Arthur Corl, Jr. ,  Anthony Matteo Gallo ,  Clark Debs Jeffries ,  Colin Beaton Verrilli

IPC分类号： G06F1730

CPC分类号： H04L47/10 ,  H04L45/00 ,  H04L47/20 ,  Y10S707/99937 ,  Y10S707/99943 ,  Y10S707/99945

摘要： This process accepts rule domination declarations and subjects rules to a computer program which either finds a type of administrative error (cyclic domination) or assigns a priority number to each rule so that any two rules which intersect (some key fits both rules) have necessarily different priority numbers. In the case that priority numbers are assigned, the process goes on to check for a second type of administrative error, namely inclusion of a first rule in a second (every key which fits the first rule also fits the second), and with the second having higher priority (so that the first is never referenced). If neither error occurs, then the number of different priority numbers is minimized. Every key when tested by such a rule set with neither error must fit either no rules at all or must fit exactly one rule with highest priority. In the latter case, the action of the unambiguously determined rule can then be applied to the packet represented by the key.

摘要翻译： 该过程接受规则控制声明和主体规则到计算机程序，该计算机程序找到一种管理错误（循环统治）或为每个规则分配一个优先级号，以便相交的任何两个规则（一些关键符合两条规则）必然不同 优先级数字。 在分配优先级编号的情况下，进程继续检查第二种类型的管理错误，即在第二种类型中包括第一规则（适合第一规则的每个密钥也适合第二规则），并且与第二类 具有较高的优先级（使得第一个从未被引用）。 如果没有发生错误，则不同优先级号码的数量被最小化。 每个密钥通过既不具有错误的规则进行测试时，也不能完全符合任何规则，也不能完全符合一个具有最高优先级的规则。 在后一种情况下，明确确定的规则的动作可以应用于由密钥表示的分组。

公开(公告)号：US06298340B1

公开(公告)日：2001-10-02

申请号：US09312148

申请日：1999-05-14

申请人： Jean Louis Calvignac ,  Everett Arthur Corl, Jr. ,  Anthony Matteo Gallo ,  Marco C. Heddes ,  Clark Debs Jeffries ,  Piyush Chunilal Patel ,  Mark Anthony Rinaldi ,  Colin Beaton Verrilli

发明人： Jean Louis Calvignac ,  Everett Arthur Corl, Jr. ,  Anthony Matteo Gallo ,  Marco C. Heddes ,  Clark Debs Jeffries ,  Piyush Chunilal Patel ,  Mark Anthony Rinaldi ,  Colin Beaton Verrilli

IPC分类号： G06F1730

CPC分类号： G06F17/30961 ,  Y10S707/99931 ,  Y10S707/99933 ,  Y10S707/99943 ,  Y10S707/99944

摘要： A classification system includes a software managed tree testing bits from a key which labels an item. The bits are chosen by application of the Choice Bit Algorithm to the Rules in a Database of Rules. A controller including logic parses an unknown Key for bits to be tested in the decision nodes of a binary tree. Tests dictated by the tree are conducted in a predetermined way until all but one Rule from the database or all but a few Rules from the database are eliminated from consideration, whereupon the Key is fully tested by the one remaining Rule or in a lattice constructed of the remaining plurality of Rules, to determine an action to enforce on the item. Certain compare tests are used in the binary tree for the case that otherwise identical or similar rules are applied to integer ranges of key values which do not fall upon power of 2 boundaries. Furthermore, some very frequently occurring rules in such final tests might be designated as secondary rules, the remaining rules designated as primary rules, the entire decision tree recalculated using only primary rules, and the primary rules then connected to secondary rules only when logically necessary by means of a system of pointers making use of relative priorities of rules.

摘要翻译： 分类系统包括从标签项目的键的软件管理树测试位。 通过将选择位算法应用于规则数据库中的规则来选择位。 包含逻辑的控制器在二叉树的决策节点中解析要测试的位的未知密钥。 由树进行的测试以预定的方式进行，直到从数据库中除了一个规则之外的所有除了数据库中的所有规则或从数据库中除了少数几个规则之外的所有测试都被消除，由此Key被完整的一个规则或由 剩余的多个规则，以确定对该项目执行的操作。 在二叉树中使用某些比较测试，否则相同或相似的规则应用于不落在2边界的幂的关键值的整数范围。 此外，这些最终测试中的一些非常频繁出现的规则可能被指定为次要规则，剩余的规则被指定为主要规则，仅使用主要规则重新计算的整个决策树，然后仅在逻辑上必要时连接到次级规则的主要规则 使用指针的相对优先级的指针系统的手段。

公开(公告)号：US06981027B1

公开(公告)日：2005-12-27

申请号：US09546653

申请日：2000-04-10

申请人： Anthony Matteo Gallo ,  Seeta Hariharan ,  Marco C. Heddes ,  Sridhar Rao ,  Colin Beaton Verrilli ,  Gail Irene Woodland

发明人： Anthony Matteo Gallo ,  Seeta Hariharan ,  Marco C. Heddes ,  Sridhar Rao ,  Colin Beaton Verrilli ,  Gail Irene Woodland

IPC分类号： G06F15/167 ,  G06F9/44 ,  G06F12/00 ,  G06F12/02 ,  G06F12/06 ,  G06F13/14 ,  H04L12/24 ,  H04L12/44 ,  H04L12/46 ,  H04L12/56 ,  H04L29/06

CPC分类号： H04L49/35 ,  H04L49/3009 ,  H04L49/90 ,  Y10S707/99953 ,  Y10S707/99956

摘要： A method and system for memory management in a network processing system provides for allocation of physical memory areas within network processors coupled to a control point processor by a bus. The allocation scheme provides a memory management layer without requiring a complete operating system interface and supports asynchronous completion of the allocation requests. Multicast allocation is supported allowing an allocation to be simultaneously requested on multiple network processors. The allocation mechanism returns a token, which may then be used to access the memory location via a protocol over the bus, and a single token may refer to an allocation made on several network processors where the actual physical addresses and memory configurations are different.

摘要翻译： 用于网络处理系统中的存储器管理的方法和系统提供通过总线分配耦合到控制点处理器的网络处理器内的物理存储器区域。 分配方案提供了一个内存管理层，而不需要完整的操作系统接口，并支持异步完成分配请求。 支持组播分配，允许在多个网络处理器上同时请求分配。 分配机制返回令牌，然后可以通过总线通过协议来访问存储器位置，并且单个令牌可以指在实际物理地址和存储器配置不同的多个网络处理器上进行的分配。

公开(公告)号：US06760776B1

公开(公告)日：2004-07-06

申请号：US09546462

申请日：2000-04-10

申请人： Anthony Matteo Gallo ,  Natarajan Vaidhyanathan ,  Colin Beaton Verrilli ,  Gail Irene Woodland

发明人： Anthony Matteo Gallo ,  Natarajan Vaidhyanathan ,  Colin Beaton Verrilli ,  Gail Irene Woodland

IPC分类号： G06F1516

CPC分类号： H04L69/10 ,  H04L69/32

摘要： A method and apparatus for processing network frames by embedding control information achieves an efficient frame processing system within a network processor (NP). The layer type of the frame can be quickly determined by the layer processing components of picocode running on the NP by examining control information that is written by ingress processing layers to produce a modified frame format. The frames are routed to appropriate layer processors and processing for certain layers may be bypassed if the picocode determines that no processing is required at that layer. The frame may also be discarded completely by any of the layer processors.

摘要翻译： 通过嵌入控制信息来处理网络帧的方法和装置在网络处理器（NP）内实现有效的帧处理系统。 通过检查由入口处理层写入的控制信息以产生修改的帧格式，可以通过在NP上运行的picocode的层处理组件快速确定帧的层类型。 帧被路由到适当的层处理器，并且如果皮卡代码确定在该层不需要处理，则可以绕过某些层的处理。 帧也可以被任何层处理器完全丢弃。

公开(公告)号：US06738376B1

公开(公告)日：2004-05-18

申请号：US09543676

申请日：2000-04-07

申请人： Richard John Blasiak ,  Gregory Scott Burgin ,  Anthony Matteo Gallo ,  Colin Beaton Verrilli ,  Gail Irene Woodland

发明人： Richard John Blasiak ,  Gregory Scott Burgin ,  Anthony Matteo Gallo ,  Colin Beaton Verrilli ,  Gail Irene Woodland

IPC分类号： H04L1228

CPC分类号： H04L45/00 ,  H04L12/185 ,  H04L12/1854 ,  H04L12/1886 ,  H04L45/16

摘要： A system for controlling multicast traffic in a multiprocessing system is disclosed. The multiprocessing system includes a general-purpose processor, a plurality of network processors and a switch fabric. The system comprises a multicast identifier (MID) manager to manage a distributed database of MIDs. The distributed database includes at least one compound MID. The at least one compound MID structure further comprises a plurality of MIDs representing at least two multicast groups. The system also includes a MID processor which traverses the at least one compound MID and resolves it into its associated simple MIDs. Accordingly, the present invention provides for a system and method for efficiently controlling multicast transmissions when different types of multicast groups are to receive the multicast transmissions. In this system and method a new type of multicast identifier (MID) is provided, a compound MID. The compound MID is a single MID but references multiple MIDs. The system includes a MID manager which manages a distributed database of MIDS. Finally, the system and method includes at least one MID processor which accesses the appropriate compound MIDs and resolves them into the simple member MIDs to allow a router to provide the information to an appropriate bridge.

摘要翻译： 公开了一种用于在多处理系统中控制多播流量的系统。 多处理系统包括通用处理器，多个网络处理器和交换结构。 该系统包括用于管理MID的分布式数据库的多播标识符（MID）管理器。 分布式数据库包括至少一个复合MID。 所述至少一个复合MID结构还包括表示至少两个多播组的多个MID。 该系统还包括一个MID处理器，它遍历至少一个复合MID并将其解析成其相关联的简单MID。 因此，本发明提供了一种用于当不同类型的多播组要接收多播传输时有效地控制多播传输的系统和方法。 在该系统和方法中，提供了一种新型的组播标识符（MID），即复合MID。 复合MID是单个MID，但是引用了多个MID。 该系统包括管理MIDS的分布式数据库的MID管理器。 最后，该系统和方法包括至少一个MID处理器，其访问适当的复合MID并将其解析成简单成员MID，以允许路由器将信息提供给适当的桥。

公开(公告)号：US06700883B1

公开(公告)日：2004-03-02

申请号：US09543145

申请日：2000-04-05

申请人： Anthony Matteo Gallo ,  Brahmanand Kumar Gorti ,  Donald Newland Jones ,  Natarajan Vaidhyanathan ,  Colin Beaton Verrilli

发明人： Anthony Matteo Gallo ,  Brahmanand Kumar Gorti ,  Donald Newland Jones ,  Natarajan Vaidhyanathan ,  Colin Beaton Verrilli

IPC分类号： H04L1256

CPC分类号： H04L29/06 ,  H04L69/32

摘要： A controllable mechanism for by-passing Layer 4 (L4) classification is based on the insertion into a set of Layer 3 (L3) rules in an L3 lookup tree set of Layer 4 (L4) Classification Required Flags. The state of the L4 classification flag is set by comparing the L4 classification rule to an IP (Internet Protocol) lookup rule. Routing is accomplished by selecting which rule to apply to the data packet and reading the state of the corresponding L4 Classification Required Flag. In response to a first state of the corresponding L4 Classification Required Flag, an L4 classification is performed followed by a routing of the data packet. In response to a second state of the corresponding L4 Classification Required Flag performing a routing of said data packet. In a second embodiment, the method inserts into a set of L3 rules in L3 lookup means a set L4 Classification Required Flags and Global Flags. A first state of the L4 Classification Required Flag is set when a new rule is added to L4 classification means, the new rule being correlatable to a single entry in L3 lookup means. Routing is accomplished by selecting which rule to apply to the data packet and reading the state of the corresponding L4 classification required flag. In response to a first state of the corresponding L4 classification flag, a L4 classification is performed followed by a routing of the data packet. In response to a second state of the corresponding L4 classification flag, the state of the Global Flag is read, and in response to a first state of the Global flag, a L4 classification is performed followed by a routing of the data packet. In response to a second state of the Global Flag, the data packet is routed.

摘要翻译： 用于旁路第4层（L4）分类的可控机制是基于在第4层（L4）分类要求标志的L3查找树集中插入到一组第3层（L3）规则中。 通过将L4分类规则与IP（因特网协议）查找规则进行比较来设定L4分类标志的状态。 通过选择应用于数据包的规则并读取相应的L4分类要求标志的状态来完成路由。 响应于对应的L4分类要求标志的第一状态，执行L4分类，然后进行数据分组的路由。 响应于相应的L4分类要求标志的第二状态来执行所述数据分组的路由。 在第二实施例中，该方法在L3查找中插入一组L3规则，即集合L4分类要求标志和全局标志。 当将新规则添加到L4分类装置时，设置L4分类要求标志的第一状态，新规则可与L3查找装置中的单个条目相关。 通过选择应用于数据包的规则并读取相应的L4分类所需标志的状态来完成路由。 响应对应的L4分类标志的第一状态，执行L4分类，然后进行数据分组的路由。 响应对应的L4分类标志的第二状态，读取全局标志的状态，并且响应于全局标志的第一状态，执行L4分类，然后进行数据分组的路由。 响应全局标志的第二状态，数据包被路由。

公开(公告)号：US06792423B1

公开(公告)日：2004-09-14

申请号：US09723717

申请日：2000-11-28

申请人： Clark Debs Jeffries ,  Anthony Matteo Gallo ,  Natarajan Vaidhyanathan ,  Colin Beaton Verrilli

发明人： Clark Debs Jeffries ,  Anthony Matteo Gallo ,  Natarajan Vaidhyanathan ,  Colin Beaton Verrilli

IPC分类号： G06T1730

CPC分类号： G06F17/30985 ,  H04L29/12801 ,  H04L61/6004 ,  Y10S707/99936

摘要： A method and system for finding a longest matching prefix for an input keyword from among multiple prefixes. The prefixes are data strings of varying lengths wherein prefixes of length n or greater are probabilistically a longest prefix match. The method of the present invention begins by mapping the prefixes of length greater than or equal to n1, that is, in the interval [n1, L], into a first lookup system. Remaining prefixes of length less than n1 but greater than or equal to n2, that is, in the interval [n2, n1−1], are mapped into a second index utilizing a second hash function, wherein n2 is less than n1. Further lookup systems on prefixes having lengths in the intervals [n3, n2−1], [n4, n3−1], and so on, may also be utilized, as determined by optimization studies and the statistics of routing tables.

摘要翻译： 一种用于从多个前缀中为输入关键字找到最长匹配前缀的方法和系统。 前缀是具有不同长度的数据串，其长度为n或更大的前缀概率地是最长前缀匹配。 本发明的方法首先将长度大于或等于n1的前缀，即间隔[n1，L]映射到第一查找系统中。 长度小于n1但大于或等于n2的剩余前缀，即在间隔[n2，n1-1]中，使用第二散列函数映射到第二索引，其中n2小于n1。 还可以利用在间隔[n3，n2-1]，[n4，n3-1]等中具有长度的前缀上的进一步查找系统，如通过优化研究和路由表的统计确定的。

公开(公告)号：US06654372B1

公开(公告)日：2003-11-25

申请号：US09543144

申请日：2000-04-05

申请人： Francis Arts ,  Olivier Didier Duroyon ,  Anthony Matteo Gallo ,  Brahmanand Kumar Gorti ,  Donald Newland Jones ,  Natarajan Vaidhyanathan ,  Colin Beaton Verrilli

发明人： Francis Arts ,  Olivier Didier Duroyon ,  Anthony Matteo Gallo ,  Brahmanand Kumar Gorti ,  Donald Newland Jones ,  Natarajan Vaidhyanathan ,  Colin Beaton Verrilli

IPC分类号： H04L1256

CPC分类号： H04L45/00 ,  H04L45/302

摘要： A controllable mechanism for by-passing Layer 4 (L4) classification is based on the insertion into a set of MAC rules in SA MAC lookup means a set of Layer 4 (L4) Skip Classification Flags. Routing is accomplished by selecting which rule to apply to the packet and reading the state of the corresponding L4 Skip Classification Flag. In response to a first state of said corresponding L4 Skip Classification Flag, performing an L4 classification followed by a routing of the data packet. In response to a second state of said corresponding L4 Skip Classification Flag, reading the state of a Global Classification Flag. In response to a first state of said Global Classification Flag, performing an L4 classification followed by a routing of said data packet. In response to a second state of said Global Classification Flag performing a routing of the data packet. The L4 Skip option change does not use the option change of reading the L4 Skip Classification Flag from the port table, whereas the second embodiment uses this option. The third and fourth embodiments are similar to the first and second embodiments but with the. addition of inserting into a set of Layer 3 (L3) rules in L3 lookup means a set of Layer 4 (L4) Classification Required Flags. The third embodiment does not use the option of reading the L4 Skip Classification Flag from the port table, whereas the fourth embodiment uses this option.

摘要翻译： 用于旁路第4层（L4）分类的可控机制是基于在SA MAC查找中插入一组MAC规则的方法，即一组第4层（L4）跳过分类标志。 通过选择哪个规则应用于分组并读取相应的L4跳过分类标志的状态来完成路由。 响应于所述对应的L4跳过分类标志的第一状态，执行L4分类，然后进行数据分组的路由。 响应于所述对应的L4跳过分类标志的第二状态，读取全局分类标志的状态。 响应于所述全局分类标志的第一状态，执行L4分类，然后进行所述数据分组的路由。 响应于所述全局分类标志的第二状态执行数据分组的路由。 L4跳过选项更改不使用从端口表读取L4跳过分类标志的选项更改，而第二个实施例使用此选项。 第三和第四实施例类似于第一和第二实施例，但是与。 在L3查找中插入到一组第3层（L3）规则中的添加意味着一组第4层（L4）分类要求标志。 第三实施例不使用从端口表读取L4跳过分类标志的选项，而第四实施例使用该选项。

公开(公告)号：US06898179B1

公开(公告)日：2005-05-24

申请号：US09544896

申请日：2000-04-07

申请人： Brian Mitchell Bass ,  Jean Louis Calvignac ,  Anthony Matteo Gallo ,  Marco C. Heddes ,  Mark Anthony Rinaldi ,  Michael Steven Siegel ,  Colin Beaton Verrilli ,  Fabrice Jean Verplanken

发明人： Brian Mitchell Bass ,  Jean Louis Calvignac ,  Anthony Matteo Gallo ,  Marco C. Heddes ,  Mark Anthony Rinaldi ,  Michael Steven Siegel ,  Colin Beaton Verrilli ,  Fabrice Jean Verplanken

IPC分类号： G06F11/00 ,  H04B7/216 ,  H04L1/16 ,  H04L12/26

CPC分类号： G06F15/17

摘要： The transport protocol for communicating between general purpose processors acting as contact points and network processors in a packet processing environment such as Ethernet is provided. In such an environment, there is at least one single control point processor (CP) and a plurality of network processors (NP), sometimes referred to as blades. A typical system could contain two to sixteen network processors, and each network processor connects to a plurality of devices which communicate with each other over a network transport, such as Ethernet. The CP typically controls the functionality and the functioning of the network processors to function in a way that connects one end user with another, whether or not the end user is on the same network processor or a different network processor. There are three types of communication provided; first, there is communication generally referred to as control services and normally there will be only one pico processor which operates as a GCH (guided cell handler) and only one that operates as a guided tree handler (GTH). A path is provided for the controls to the GCH and the GTH commands, and a separate path is provided for the data frames between the GDH's (general data handler) and the CP.

摘要翻译： 提供了用于在诸如以太网的分组处理环境中用作接触点的通用处理器和网络处理器之间进行通信的传输协议。 在这样的环境中，存在至少一个单个控制点处理器（CP）和多个网络处理器（NP），有时称为刀片。 典型的系统可以包含两到十六个网络处理器，并且每个网络处理器连接到通过诸如以太网的网络传输彼此通信的多个设备。 CP通常控制网络处理器的功能和功能，以使终端用户与另一终端用户相连的方式起作用，无论终端用户是否在同一个网络处理器或不同的网络处理器上。 提供三种通讯方式; 首先，通常被称为控制服务的通信，并且通常将只有一个微微处理器作为GCH（引导的单元处理器）操作，并且只有一个作为引导树处理器（GTH）操作。 为GCH和GTH命令的控制提供路径，并为GDH（通用数据处理程序）和CP之间的数据帧提供单独的路径。