The Open Systems Interconnection (OSI) Reference Model defines seven network protocol layers (L1-L7) used to communicate over a transmission medium. The upper layers (L4-L7) represent end-to-end communications and the lower layers (L1-L3) represent local communications.
Networking application aware systems need to process, filter and switch a range of L3 to L7 network protocol layers, for example, L7 network protocol layers such as, HyperText Transfer Protocol (HTTP) and Simple Mail Transfer Protocol (SMTP), and L4 network protocol layers such as Transmission Control Protocol (TCP). In addition to processing the network protocol layers, the networking application aware systems need to simultaneously secure these protocols with access and content based security through L4-L7 network protocol layers including Firewall, Virtual Private Network (VPN), Secure Sockets Layer (SSL), Intrusion Detection System (IDS), Internet Protocol Security (IPSec), Anti-Virus (AV) and Anti-Spam functionality at wire-speed.
Improving the efficiency and security of network operation in today's Internet world remains an ultimate goal for Internet users. Access control, traffic engineering, intrusion detection, and many other network services require the discrimination of packets based on multiple fields of packet headers, which is called packet classification.
Internet routers classify packets to implement a number of advanced Internet services such as routing, rate limiting, access control in firewalls, virtual bandwidth allocation, policy-based routing, service differentiation, load balancing, traffic shaping, and traffic billing. These services require the router to classify incoming packets into different flows and then to perform appropriate actions depending on this classification.
A classifier, using a set of filters or rules, specifies the flows, or classes. For example, each rule in a firewall might specify a set of source and destination addresses and associate a corresponding deny or permit action with it. Alternatively, the rules might be based on several fields of a packet header including layers 2, 3, 4, and 5 of the OSI model, which contain addressing and protocol information.
On some types of proprietary hardware, an Access Control List (ACL) refers to rules that are applied to port numbers or network daemon names that are available on a host or layer 3 device, each with a list of hosts and/or networks permitted to use a service. Both individual servers as well as routers can have network ACLs. ACLs can be configured to control both inbound and outbound traffic.
According to one embodiment, an apparatus may comprise a memory configured to store a Rule Compiled Data Structure (RCDS). The Rule Compiled Data Structure (RCDS) may represent a set of rules for packet classification. The apparatus may further comprise a processor coupled to the memory. The processor may be configured to receive an incremental update for the RCDS, perform an active search of the Rule Compiled Data Structure (RCDS) to classify received packets, and update the Rule Compiled Data Structure (RCDS) atomically from the perspective of the active search being performed. The apparatus may further comprise an interface. The interface may be configured to receive a change list; the change list may include the incremental update. The processor may be a search processor. The Rule Compiled Data Structure (RCDS) may include a compiled set of rules and an updated Rule Compiled Data Structure (RCDS) may include the compiled set of rules and one or more applied incremental updates. The updated Rule Compiled Data Structure (RCDS) achieves a same performance as a pre-complied version of the updated RCDS.
According to one embodiment, an apparatus may comprise a memory and a processor coupled to the memory. The processor may be configured to include an incremental update module. The incremental update module may be configured to receive an add, delete, or modify rule operation and create a change list in the memory. The change list may be created to atomically update a Rule Compiled Data Structure (RCDS) from the perspective of an active perspective of an active search process utilizing the RCDS. The apparatus may further comprise an interface and the incremental update module may be further configured to communicate the change list over the interface. The apparatus may further comprise a compiler module coupled to the memory. The compiler module may be configured to receive a set of rules and compile the set of rules into a binary format of a decision tree. The compiler module may be further configured to store the binary format of a decision tree in the memory and to communicate the binary format of a decision tree over the interface.
According to one embodiment a non-transient computer-readable medium has encoded thereon a sequence of instructions which, when executed by a processor, causes the processor to receive an incremental update for a Rule Compiled Data Structure (RCDS). The Rule Compiled Data Structure (RCDS) may represent a set of rules for packet classification. The Rule Compiled Data Structure (RCDS) may be utilized for packet classification by an active search process. The processor may atomically update the Rule Compiled Data Structure (RCDS) based on the incremental update received. The Rule Compiled Data Structure (RCDS) may be atomically updated from the perspective of the active search process utilizing the Rule Compiled Data Structure (RCDS).
According to one embodiment a method may receive an incremental update for a Rule Compiled Data Structure (RCDS). The Rule Compiled Data Structure (RCDS) may represent a set of rules for packet classification. The Rule Compiled Data Structure (RCDS) may be utilized for packet classification by an active search process and atomically updated based on the incremental update received, from the perspective of the active search process utilizing the RCDS.
The method may restrict a state of the Rule Compiled Data Structure (RCDS) to a before state and an after state. The before state may be a state of the Rule Compiled Data Structure (RCDS) before receiving the incremental update for the RCDS, the after state may be a state of the Rule Compiled Data Structure (RCDS) after a series of one or more modifications to the Rule Compiled Data Structure (RCDS) has been completed. The series of one or more modifications may have been completed based on the incremental update received. The series of one or more modifications may be visible to the active search process based on performing one update to the Rule Compiled Data Structure (RCDS) being searched.
The method may further include atomically adding a new rule to the Rule Compiled Data Structure (RCDS) based on the incremental update being an add rule operation, atomically deleting a rule from the Rule Compiled Data Structure (RCDS) based on the incremental update being a delete rule operation, and atomically modifying a rule in the Rule Compiled Data Structure (RCDS) based on the incremental update being a modify rule operation. Modifying the rule includes at least one of: modifying a priority of the rule or modifying at least one field of the rule.
The method may further include identifying a priority fit conflict based on a change in priority of the rule being inconsistent with a current priority ordering of the rule and one or more other rules. The method may atomically modify the priority of the rule based on the priority fit conflict not being identified, and atomically modify the priority of the rule and priority of another rule based on the conflict being identified.
The method may further include determining whether one or more rules need to be added or deleted, and adding or deleting the one or more rules. Adding or deleting the one or more rules is atomic.
The method may further include atomically invalidating a specified rule in the Rule Compiled Data Structure (RCDS) based on the incremental update being a delete operation specifying the rule. The active search process skips the specified rule invalidated.
The method may further include representing the Rule Compiled Data Structure (RCDS) as a tree of the set of rules. The tree may be a binary data structure including one or more nodes and one or more leaves. The method may include representing at least one of the one or more nodes as a parent node and linking the parent node to one or more children. The one or more children may be a node or a leaf Linking the parent node to the one or more children may include pointing the parent node to a sibling list. The sibling list may include the one or more children. The method may link nodes of the tree to one or more nodes and one or more leaves of the tree. The method may link leaves of the tree to one or more buckets. Each bucket may represent a subset of the set of rules. Each bucket may include one or more bucket entries corresponding to the subset of the set of rules. The bucket entries may be ordered by increasing or decreasing rule priority. The method may further include storing the set of rules in a rule table. The rules within the rule table may be ordered or unordered.
The Rule Compiled Data Structure (RCDS) may be a performance tree. The method may further include maintaining a housekeeping tree. The housekeeping tree may be an augmented representation of the performance tree. The housekeeping tree may include field ranges of the rules and lists of the rules at each node in the tree. Updating the performance tree atomically may include utilizing the housekeeping tree such that a series of one or more modifications to the performance tree is made visible to the active search process based on one update to the performance tree being searched.
The method may further include creating a change list specifying the one or more modifications to the performance tree.
The incremental update may be an add, delete, or modify operation, and the method may further include: including a cover list of rules for each rule in the housekeeping tree. A change list may be created specifying one or more rules to add, delete, or modify based on the cover list. The method may include updating the cover list based on the change list determined.
The method may further include maintaining in each leaf a pointer to a bucket from among the one or more buckets and a bucket rule counter. The bucket rule counter may track a number of rules included in the bucket. The bucket rule counter may be incremented based on a rule being added to the bucket. The bucket rule counter may be decremented based on a rule being deleted from the bucket.
The method may include tracking at each node a total number of incremental updates. The method may determine at a given node a ratio of the total number of incremental updates tracked for the given node to a number of rules represented by the given node. The method may adaptively adjust the performance tree by recompiling a subtree based on the ratio being greater than a given threshold value.
The method may include atomically adding a new rule to the tree. The method may split a leaf of the tree into one or more new nodes and add the rule to a bucket associated with one or more leaves of the one or more new nodes.
Each bucket may be a data structure and the one or more bucket entries may be a rule, an index to a rule, a pointer to a rule, a pointer to a set of rules, or a pointer to another bucket.
The method may further include atomically adding a new rule to the tree, and identifying a destination bucket from among the one or more buckets to include the new rule. The method may append the new rule to the end of the destination bucket based on determining a space fit and priority fit of the new rule in the destination bucket. Appending the new rule to the end of the destination bucket takes one update.
A space fit may be based on a current number of rules in the destination bucket being less than a maximum number of rules for the destination bucket.
A priority fit may be based on the priority associated with the new rule being consistent with a priority ordering of current rules in the destination bucket.
The method may further include atomically adding a new rule to the tree, identifying a destination bucket from among the one or more buckets to include the new rule, and creating a new bucket based on determining the priority associated with the new rule being inconsistent with a priority ordering of rules in the destination bucket. The active search process is unaffected by the new bucket created. The method may further include: including the set of rules of the destination bucket in the new bucket, including the new rule in the new bucket, and adjusting an order of the set of rules and the new rule based on increasing or decreasing priority order. The method may update a link of a leaf in the tree. The leaf may have a link to a destination bucket from among the one or more buckets. The link update may include pointing the leaf to the new bucket. The link update takes one update.
The method may further include atomically adding a new rule to the tree by identifying a destination bucket from among the one or more buckets to include the new rule, creating a new bucket, the active search process being unaffected by the new bucket created, including the new rule in the new bucket, and updating a bucket entry of the destination bucket to point to the new bucket. The bucket entry update takes one update.
The method may further include atomically adding a new rule to the tree by identifying a destination bucket from among the one or more buckets to include the new rule, and creating a subtree based on determining lack of space in the destination bucket for the new rule. The lack of space in the destination bucket may be determined based on a maximum number of rules set for the destination bucket. The subtree may include one or more new leaves or new nodes. The active search process is unaffected by the subtree created. The method may further include adding the new rule to one or more buckets of the subtree, adding one or more rules of the destination bucket to one or more buckets of the subtree and linking the subtree to the tree by updating a link of a leaf in the tree. The leaf may have a link to the destination bucket from among the one or more buckets. The link update may include pointing the leaf to the subtree, converting the leaf to a node. The link update takes one update.
The method may further include atomically adding a new leaf or a new node to a parent node in the tree, wherein the parent node in the tree is linked to a current sibling list. The current sibling list may include one or more leaves or one or more nodes. The method may further include creating a new sibling list. The active search process is unaffected by the new sibling list created. The method may further include adding the new leaf or the new node to the new sibling list. The new sibling list may include the current sibling list. The method may further include: including the new sibling list in the tree by updating a link of the parent to point the parent node to the new sibling list. The link update takes one update.
The method may further include reserving a defragmentation portion of the memory space. The defragmentation portion of the memory space may be a designated defragmentation area. The defragmentation area may be a contiguous portion of the memory space reserved at a designated area of the memory space. The method may further include identifying a dead area of the RCDS. The dead area may be a portion of the memory space being occupied by at least one unreferenced leaf, node, bucket, or sibling list. The method may further include defragmenting the Rule Compiled Data Structure (RCDS) by recovering the dead area. The active search process is unaffected by the dead area recovered, the Rule Compiled Data Structure (RCDS) is defragmented atomically. The method may further include recovering the dead area including relocating a used portion of memory. The used portion may include one or more nodes or leaves located adjacent to the dead area identified to the designated defragmentation area. The method may further include moving a new sibling list to a recovered memory space. The recovered memory space may include the dead area identified and the used portion of memory relocated. The method may further include moving the used portion of memory relocated to the defragmentation area out of the defragmentation area to an end portion of the memory space reserved for the RCDS.
The method may further include atomically deleting a specified rule from the tree, wherein a bucket entry of a bucket includes the specified rule. The method may include linking the bucket to a leaf by a leaf bucket pointer and invalidating the specified rule in the bucket entry. The method may further include updating the link by setting the leaf bucket pointer to null if all bucket entries in the bucket are invalidated. Invalidating the specified rule and updating the leaf bucket pointer to null take one update. The active search process skips a bucket if the leaf bucket pointer is null.
The method may further include atomically deleting a specified rule from the tree, wherein a bucket entry of a bucket includes a rule pointer pointing to the specified rule. The method may include linking the bucket to a leaf by a leaf bucket pointer. The method may further include setting the rule pointer to null and updating the link by setting the leaf bucket pointer to null if all rule pointers in the bucket are null. Setting the rule pointer to null and setting the leaf bucket pointer to null take one update. The active search process skips null leaf bucket pointers and skips null rule pointers.
The method may further include atomically deleting a specified rule from the tree, wherein a bucket entry of a bucket includes a rule chunk pointer pointing to a set of one or more rules including the specified rule. The method may include linking the bucket to a leaf by a leaf bucket pointer, invalidating the specified rule and setting the rule chunk pointer to null if all rules in the set of one or more rules are invalidated. The method may further include updating the link by setting the leaf bucket pointer to null if all bucket entries in the bucket are invalidated. Invalidating the specified rule, setting the rule chunk pointer to null, and setting the leaf bucket pointer to null take one update. The active search process skips null leaf bucket pointers, null rule chunk pointers, and invalidated rules.
The method may further include atomically adding one or more rules based on deleting a specified rule from the tree. The method may include identifying a destination bucket from among the one or more buckets to include the new rule. The method may further include creating a subtree based on determining lack of space in the destination bucket for the new rule. Lack of space in the destination bucket may be determined based on a maximum number of rules set for the destination bucket. The subtree may include one or more new leaves or new nodes. The active search process is unaffected by the subtree created. The method may further include adding the new rule to one or more buckets of the subtree, adding one or more rules of the destination bucket to one or more buckets of the subtree, and linking the subtree to the tree by updating a link of a leaf in the tree. The leaf may have a link to the destination bucket from among the one or more buckets. The link update includes pointing the leaf to the subtree, converting the leaf to a node. The link update takes one update.
The method may further include atomically adding one or more rules based on deleting a specified rule from the tree. The method may include identifying a destination bucket from among the one or more buckets to include the new rule and appending the new rule to the end of the destination bucket based on determining a space fit and priority fit of the new rule in the destination bucket. Appending the new rule to the end of the destination bucket takes one update.
According to one embodiment, the incremental update is an add rule operation, a delete rule operation, or a modify rule operation. Updating the Rule Compiled Data Structure (RCDS) based on the incremental update received may further include atomically adding, deleting, modifying, or any combination thereof, one or more rules to/from the RCDS.
According to another embodiment, a method may comprise receiving one or more incremental updates for a Rule Compiled Data Structure (RCDS) representing a plurality of rules. The plurality of rules may have at least one field. The RCDS may represent the plurality of rules as a decision tree for packet classification. The method may determine one or more updates for the RCDS based on the one or more incremental updates received. The method may employ at least one heuristic for maintaining quality of the RCDS. At a given one of the one or more incremental updates received, the method may identify a section of the RCDS and trigger recompilation of the identified section based on the one or more updates determined, altering the RCDS shape or depth in a manner detected by the at least one heuristic employed.
Maintaining quality may include enabling a balanced shape of the RCDS, increasing a number of decision tree paths in the RCDS, or controlling a number of decision tree levels in the RCDS by recompiling the identified section.
The RCDS may include a plurality of nodes, each node may represent a subset of the plurality of rules. The manner detected by the at least one heuristic employed may include identifying a node of the plurality of nodes having a number of children less than a given child threshold value as a result of an update to the identified node for the given one of the one or more incremental updates received, the identified section being the identified node. The method may recompile the identified node to increase the number of children of the identified node providing an increased number of decision tree paths in the RCDS.
The RCDS may include a plurality of nodes, each node may represent a subset of the plurality of rules. The manner detected by the at least one heuristic employed may include at each node, tracking a total number of node updates since a last recompilation of the node. The method may compute a ratio of the total number of incremental updates tracked for the node to a number of rules represented by the node at a last recompilation of the node. The method may recompile the node if the ratio is greater than a given update threshold value as a result of an update to the node caused by the given one of the one or more incremental updates received, the identified section being the node.
Tracking the total number of node updates may include maintaining a counter, incrementing the counter maintained each time a new rule is added to the node, incrementing the counter maintained each time an existing rule is modified or deleted from the node, and resetting the counter maintained if the node is recompiled.
Each node may be cut into a number of child nodes on a selected one or more fields of the at least one field and a selected one or more bits of the selected one or more fields. The node may include a stride value, the stride value may indicate a set of one or more bits used for cutting of the node. The given update threshold value may be a function of a number of the one or more bits in the set.
The given update threshold value may be a function the node's depth, the node's depth being based on a number of nodes included in a direct path between the node and a root node of the RCDS, the root node representing the plurality of rules.
The RCDS may include a plurality of nodes, each node may represent a subset of the plurality of rules, and the identified section may be a node of the plurality of nodes.
The identified section may be the entire RCDS.
The RCDS may include a plurality of nodes, each node may represent a subset of the plurality of rules, and the identified section may be a node of the plurality of nodes having been cut into a number of other nodes, leaves, or a combination thereof. Recompiling may include selecting one or more fields of the at least one field and re-cutting the node into a new number of new other nodes, new other leaves, or a new combination thereof.
The method may further comprise updating the RCDS in a manner enabling the RCDS to incorporate the one or more incremental updates determined and the recompiled identified section atomically from the perspective of an active search process utilizing the RCDS for packet classification.
Another example embodiment disclosed herein includes an apparatus corresponding to operations consistent with the method embodiments described above.
Further, yet another example embodiment may include a non-transitory computer-readable medium having stored thereon a sequence of instructions which, when loaded and executed by a processor, causes the processor to complete methods consistent with the method embodiments described above.
The foregoing will be apparent from the following more particular description of example embodiments of the invention, as illustrated in the accompanying drawings in which like reference characters refer to the same parts throughout the different views. The drawings are not necessarily to scale, emphasis instead being placed upon illustrating embodiments of the present invention.
A description of example embodiments of the invention follows.
The teachings of all patents, published applications and references cited herein are incorporated by reference in their entity.
Complexity of network applications is increasing due to the increasing number of network applications being implemented in network devices. Packet classification is widely used for various kinds of applications, such as service-aware routing, intrusion prevention and traffic shaping. Packet classification solutions are required to handle the exponentially increasing traffics on edge, access, and core devices.
Embodiments of a system, apparatus, and method are disclosed herein for adding, deleting, and modifying rules in one update from the perspective of an active search process using the rules for packet classification. To perform packet classification, a search processor may search for one or more rules that match keys generated from received packets. While an active search is in process, there may be a need to add, delete, or modify rules. Embodiments disclosed herein enable adding, deleting, and modifying rules in one update from the perspective of an active search process using the rules for packet classification.
The rules may be presented by a Rule Compiled Data Structure (RCDS) that may be implemented as a decision tree. The RCDS may also be referred to herein as a “tree” or a “decision tree.” As incremental updates are received, such as incremental updates for adding, deleting, and modifying rules, the tree may be modified based on the incremental updates received and as a result, a shape or depth of the tree may become unbalanced. According to embodiments disclosed herein, heuristics may be employed to determine when to recompile the tree and to identify a section of the tree to recompile, enabling quality of the tree to be maintained. According to embodiments disclosed herein, a more balanced distribution of nodes and leaves of the tree may be achieved by recompilation of a section of the tree, or even the entire tree itself.
The core routers 104a-h may be configured to operate in the Internet core 102 or Internet backbone. The core routers 104a-h may be configured to support multiple telecommunications interfaces of the Internet core 102 and may further be configured to forward packets at a full speed of each of the multiple telecommunications protocols.
The edge routers 106a-f may be placed at the edge of the Internet core 102. Edge routers 106a-f may bridge access routers 108a-e outside the Internet core 102 and core routers 104a-h in the Internet core 102. Edge routers 106a-f may be configured to employ a bridging protocol to forward packets from access routers 108a-e to core routers 104a-h and vice versa.
The access routers 108a-e may be routers used by an end user, such as a home user or an office, to connect to one of the edge routers 106a-f, which in turn may connect to the Internet core 102 by connecting to one of the core routers 104a-h. In this manner, the edge routers 106a-f may connect to any other edge router 106a-f via the edge routers 106a-f and the interconnected core routers 104a-h.
The processors described herein may reside in any of the core routers 104a-h, edge routers 106a-f, and access routers 108a-e. The search processor described herein, within each of these routers, may be configured to analyze (e.g., classify) Internet Protocol (IP) packets based on a set of rules and forward the IP packets along an appropriate network path. Packet classification must be intelligent to handle diverse types of rule sets without significant loss of performance. In addition, new technologies, such as multi-core processors provide unprecedented computing power, as well as highly integrated resources. Thus, packet classification solutions must be well suited to advanced hardware and software technologies.
Existing packet classification methods trade memory for time. Although the tradeoffs have been constantly improving, the time taken for a reasonable amount of memory is still generally poor. Because of problems with existing methods, vendors use Ternary Content-Addressable Memory (TCAM), which uses brute-force parallel hardware to simultaneously check packets against all rules. The main advantages of TCAMs over existing methods are speed and determinism (TCAMs work for all databases).
A TCAM is a hardware device that functions as a fully associative memory. A TCAM cell stores three values: 0, 1, or ‘X,’ which represents a don't-care bit and operates as a per-cell mask enabling the TCAM to match rules containing wildcards, such as a kleene star ‘*’. In operation, a whole packet header can be presented to a TCAM to determine which entry (rule) it matches. However, the complexity of TCAMs has allowed only small, inflexible, and relatively slow implementations that consume a lot of power. Therefore, a need continues for efficient methods operating on specialized data structures.
Current methods remain in the stages of mathematical analysis and/or software simulation (observation based solutions). Proposed mathematic solutions have been reported to have excellent time/special complexity. However, methods of this kind have not been found to have any implementation in real-life network devices because mathematical solutions often add special conditions to simplify a problem and/or omit large constant factors which might conceal an explicit worst-case bound. Proposed observation based solutions employ statistical characteristics observed in rules to achieve efficient solution for real-life applications. However, these methods generally only work well with specific type of rule sets. Because packet classification rules for different applications have diverse features, few observation based methods are able to fully exploit redundancy in different types of rule sets to obtain stable performance under various conditions.
Packet classification may be performed using a packet classifier, also called a policy database, flow classifier, or simply a classifier. A classifier is a collection of rules or policies. Packets received are matched with rules, which determine actions to take with a matched packet. Generic packet classification requires a router to classify a packet on the basis of multiple fields in a header of the packet. Each rule of the classifier specifies a class that a packet may belong to according to criteria on ‘F’ fields of the packet header and associates an identifier (e.g., class ID) with each class. For example, each rule in a flow classifier is a flow specification, in which each flow is in a separate class. The identifier uniquely specifies an action associated with each rule. Each rule has ‘F’ fields. An ith field of a rule R, referred to as R[i], is a regular expression on the ith field of the packet header. A packet P matches a particular rule R if for every i, the ith field of the header of P satisfies the regular expression R[i].
Classes specified by the rules may overlap. For instance, one packet may match several rules. In this case, when several rules overlap, an order in which the rules appear in the classifier determines the rules relative priority. In other words, a packet that matched multiple rules belongs to the class identified by the identifier (class ID) of the rule among them that appears first in the classifier.
Techniques disclosed herein may employ a decision tree that is used to match received packets with rules. A decision tree is a decision support tool that uses a tree-like graph or model of decisions and their possible consequences, including chance event outcomes, resource costs, and utility. Decision trees are commonly used in operations research, specifically in decision analysis, to help identify a strategy most likely to reach a goal. Another use of decision trees is as a descriptive means for calculating conditional probabilities.
As described herein, a processor, such as a search processor, may use a decision tree to select to match a received packet with a rule in a classifier table to determine how to process receive packets. Before runtime, rules may be compiled off-line and downloaded into a search processor. During runtime, packets flow through the search processor. The search processor may generate keys from the packets, search for one or more rules matching the keys, and return results of a match found or not found. While the search processor searches for one or more rules that match keys (e.g., the search processor is performing an active search process), there may be a need to add, delete or modify rules—an incremental update. According to techniques as disclosed herein, the search processor may add, delete or modify rules without affecting the ability of the search processor to search for one or more rules that match keys, in terms of both performance (e.g., how many packets are searched per a unit of time) and functionality.
The control plane apparatus 224 may include a control plane processor 202 that may include a compiler module 216, an incremental update module 210, and may be coupled to control plane memory 208. The control plane memory 208 may include a binary format of a decision tree 218, a housekeeping tree 220, and a change list 222. The compiler module 216 may be configured to receive a rule file 230 and to compile the received rule file into a decision tree, such as a binary format of a decision tree 218 (e.g., a binary data structure).
Each bucket is a data structure that may include one or more bucket entries. A bucket entry may be a rule, an index to a rule, a pointer to a rule, a pointer to a set of rules, or a pointer to another bucket. A bucket may include entries including any combination thereof. For example, a bucket may have one entry that is a pointer to a rule and one entry that is a pointer to a set of rules, etc.
Each bucket may include bucket entries which may contain rule or chunk pointers 296. The rules 286 may include chunks of rules 294. A chunk 294 (of rules) can be a sequential group of rules, or a group of rules scattered throughout the memory, either organized by a plurality of pointers or by recollecting the scattered chunk 294 (e.g., using a hash function).
The binary format of a decision tree 218 may be downloaded to the data plane apparatus 226 over an interface 232. The interface 232 may be a Peripheral Component Interconnect Express (PCIe) interface, Intelligent Interface Controller (I2C) interface, or any other suitable interface that may be known to one skilled in the art.
The data plane apparatus may include a packet processing module 234 that includes a data plane processor 204 coupled to a data plane memory 236. The packet processing module 234 may be coupled to a classifier module 206 that may also be included in the data plane apparatus 226. The classifier module 206 may include a search processor 238 that may be coupled to a search processor memory 240. The data plane processor 204 may be coupled to the search processor 238 by an interface, such as an Interlaken look aside interface, or any other suitable interface that may be known by one skilled in the art.
The binary format of a decision tree 218 may be downloaded to the classifier module 206 over the interface 242 and stored as the Rule Compiled Data Structure (RCDS) 214 that may be included in the search processor memory 240. Alternatively, the binary format of a decision tree 218 may be downloaded over interface 244 to the search processor 238 and stored as the Rule Compiled Data Structure (RCDS) 214. The interface 244 may be an Intelligent Interface Controller (I2C) interface, or any other suitable interface that may be known to one skilled in the art.
The incremental update module 210 may receive a rule, or rule operation, for adding, deleting or modifying rules for the Rule Compiled Data Structure (RCDS). The incremental update module 210 may use the binary format of a decision tree 218 and housekeeping tree 220 to create a change list 222 for atomically updating the Rule Compiled Data Structure 214. The housekeeping tree 220 may be an augmented representation of the Rule Compiled Data Structure (RCDS) 214 including additional information of the tree in order to determine one or more updates for the tree.
The housekeeping tree 220 may include and maintain information for each rule in the tree, such as a cover list. A cover list may include a set of rules that are covered by the rule, or not added to the tree because of a higher priority rule. Covered rules are “useless” rules because they are not matched. By maintaining a cover list at a node for each rule covering one or more other rules of the node, the incremental update module 210 may determine when to add and delete rules. For example, if a rule is to be deleted, the incremental update module 210 may determine to add one or more rules that were not added previously because they were covered by the rule now being deleted. If the covering rule is deleted, the previously “useless” rules may now be useful as there is a possibility that they may be matched on. Cover lists are one example of rule information that may be included and maintained in the housekeeping tree, other rule information may also be included in order to assist the incremental update in determining tree updates.
The change list may specify one or more commands for atomically updating the Rule Compiled Data Structure (RCDS) 214 stored in the search processor memory 240. On the other hand, it may be that an Rule Compiled Data Structure (RCDS) 214 is not stored because the compiler module 216 has not compiled the binary format of a decision tree 218. In that case, the incremental update module 210 may create a change list 222 that creates the Rule Compiled Data Structure (RCDS) 214. The change list 222 may be communicated to the data plane apparatus 226 over the interface 232 and then communicated to the search processor 238 over the interface 242. Alternatively, the change list 222 may be communicated to the search processor 238 over the interface 244. The Rule Compiled Data Structure 214 may be used by an active search process 212 to classify received packets 246.
The packet processing module 234 may be configured to transmit packets 252 and receive packets 246. The data plane processor 204 may send lookup requests 248 to the search processor 238 and receive results 250 from the search processor 238. The search processor 238 may be configured to find one or more rules (matching rules) that match a packet by utilizing the Rule Compiled Data Structure (RCDS) 214. For example, a packet may be broken down into parts, such as a header, payload, and trailer. The header of the packet (or packet header) may be further broken down into fields. The search processor 238 may be configured to find one or more rules that match one or more parts of a received packet.
The lookup request 248 may includes a packet header and group identifier (GID). The GID may index an entry in a global definition/description table (GDT). Each GDT entry may include n number of table identifiers (TID), a packet header index (PHIDX), and a key format table index (KFTIDX). Each TID may index an entry in a tree location table (TLT). Each TLT entry may identify a lookup engine (e.g., search processor) to look for the one or more matching rules. In this way, each TID may specify both who will look for the one or more matching rules and where to look for the one or more matching rules.
Each table identifier (TID) may also index an entry in a tree access table (TAT). A TAT may be used in the context in which multiple lookup engines, grouped together in a cluster, look for the one or more matching rules. Each TAT entry may provide the starting address, in memory, of a collection of rules (or pointers to rules) called a table or tree of rules. The terms table of rules or tree of rules (or simply table or tree) are used interchangeably throughout the disclosure. The TID identifies which collection or set of rules, such as the Rule Compiled Data Structure 214, in which to look for one or more matching rules.
The packet header index (PHIDX) may index an entry in a packet header table (PHT). Each entry in the PHT may describe how to extract n number of keys from the packet header. The key format table index (KFTIDX) may index an entry in a key format table (KFT). Each entry in the KFT may provide instructions for extracting one or more fields (e.g., parts of the packet header) from each of the n number of keys, which were extracted from the packet header.
Each of the extracted fields, together with each of the TIDs, all of which were derived starting with the lookup request, may used to look for subsets of the rules. Each subset contains rules that may possibly match each of the extracted fields. Each rule of each subset may be compared against an extracted field. Rules that match may be provided in responses (e.g., lookup responses) as results 250.
The lookup request and its enumerated stages, as described above, are being provided merely to present concepts. These concepts may be implemented in numerous ways. For example, according to example embodiments of the present invention, these concepts may be implemented by a search processor, such as search processor 238.
As shown in
The search cluster 410 receives a key 402 from the LUF 304a-b (
Having provided (in reference to
Regarding functionality, while the search processor performs a search, the search processor should find the state of the rules (or rule table) to be either “before” or “after” a rule is added, deleted or modified. The search processor should not find the state of the rules to be some intermediate state in between “before” and “after” a rule is added, deleted or modified. For example, it should not be the case that while searching, the search processor finds a rule that matches a key some of the time because of rule updating. The addition, deletion or modification of rules without affecting functionality or performance, as described above, is referred to as an “atomic” update, or “atomically” updating.
The challenge to performing an atomic update (or atomically updating) is that any addition, deletion or modification of rules may take more than one update to complete. Some rules cannot be added, deleted or modified in a single update (e.g., a rule cannot be added, deleted or modified by simply changing a bit in that rule).
According to techniques disclosed herein, adding, deleting, and modifying a rule appear to take one update from the perspective of a search processor performing an active search. The Rule Compiled Data Structure (RCDS), or tree of rules, or tree, represents a set of rules. The tree is a binary data structure having nodes and leaves. Each leaf of the tree points to a subset of the rules, called a bucket of rules, or bucket. Each of the buckets represents a subset of the rules. Each bucket is a data structure (e.g., an array) containing rules, pointers to rules, pointers to chunks of rules, or any combination thereof, which may be stored in a rule table. Rules (or pointers to rules) within a bucket are ordered by priority (e.g., in increasing or decreasing priority). A rule table is a data structure (e.g., an array) containing the rules. Rules within the rule table may be ordered or unordered.
A rule has, among other things, a priority and one or more fields. In this way, modifying a rule includes modifying a priority and/or one or more fields. To describe modifying a priority of a rule, according to one embodiment, the following example is provided.
A network router maintains, in what may be called a “white list,” rules for allowing traffic from particular networks. The white list may also includes a rule for dropping traffic from all networks, called a “default route” or “default rule.” The rules may be prioritized such that the router compares traffic against the highest priority rule first, and if no match is found, compares the traffic against the next highest priority rule second. The router may continues comparing, working down the list of rules, until a match is found or until the router reaches the lowest priority rule, the default route, in which case the traffic is dropped.
The white list may be compromised and the router may be allowing offending traffic. Instead of going through each rule to find the rule that is allowing the offending traffic (which may be time consuming) an operator or administrator of the router may “reprioritize” the rules by changing the priority of the default route from the lowest to the highest priority. Giving the default route the highest priority stops all traffic including the offending traffic.
Modifying a priority of a rule, according to one embodiment, includes determining if changing the priority of the rule conflicts or overlaps a priority of another rule. Using the white list example above, assume the highest priority rule has a priority of 0 and there is no priority higher than 0. There is a conflict if the priority of the default route is changed to 0. In the case of conflicting priority, the priority of the rule and the priority of another rule may be updated. In the case of no conflicting priority (e.g., the highest priority rule has a priority of 1 and the priority of the default route is changed to 0), the priority of the rule is modified without deleting and adding rules. To describe modifying a field of a rule, according to one embodiment, the following example is provided.
If the rule is to be modified, a check may be made to determine if the rule priority is to be modified (646). If the priority is to be modified a check may be made to identify a priority fit conflict based on a change in priority of the rule being inconsistent with a current priority ordering of the rule and one or more other rules (648). If the priority fit conflict is not identified the priority of the rule may be atomically modified (658) and the method ends (662). If the priority fit conflict is identified, the priority of the rule and priority of another rule may be atomically modified (656) After priorities of rules are shifted around, new cover lists may be needed or updates to existing cover lists may be needed based on whether or not rules of the node cover other rules of the node due to the priority changes. As such, additional updates may be needed to remove covered rules from the node or to add rules to the node that have been uncovered due to the priority updates. A check may be made to determine if such additional updates are needed (664). If not, the method thereafter ends (662) in the example embodiment. If yes, the additional updates may be atomically incorporated (666) and the method thereafter ends (662) in the example embodiment. If the priority is not being modified, a check may be made to determine if at least one field of the rule is to be modified (650). If not, the method ends (662). If at least one field is to be modified, a determination may be made for identifying one or more rules that need to be added or deleted (652). In some cases, modifying a field of a rule may involve modifying the field without adding or deleting rules. A check may be made to determine if another field is being modified (654) and iterate to identify the one or more rules that need to be added or deleted (652). If there are no more fields to be modified, the one or more rules determined may be added or deleted and adding or deleting the one or more rules determined is atomic (660) and the method ends (662).
Rules may be added to a bucket until an upper limit is reached. Capping the number of rules that can be added to a bucket avoids making a search of the tree into a linear search of buckets. When the upper limit is reached, it can be said that there is no “space” in a bucket to add a rule. When adding a rule to a bucket that “fits,” in terms of both space and priority, the rule is simply added to the end of the bucket. The rule and its priority are appended to the rule table. A rule may include its priority.
By including a rule priority, it may not be necessary to shuffle (move) rules in memory in order to realize a specific priority order. According to one embodiment, the priority of a rule or rules may be updated instead. Rule priority may be assigned in blocks. Holes (e.g., priority differential) may be present between rules in a bucket. A priority fit may be determined base on determining that holes that are present. If a rule's modified priority fits in a priority hole, other rules may not need to have their priority adjusted. In addition, rules included in a bucket (either directly or indirectly via pointers as disclosed herein) may have a same priority if the rules do not compete for a match. A priority fit may be determined even if the priority of a rule is being modified to a priority that is the same as another rule in the bucket.
For example, according to one embodiment, when adding a rule to a bucket that does not fit priority-wise (e.g., adding the rule to the end of the bucket results in an incorrect priority order), a “new” bucket is created. The new bucket may include the rules from the “old” bucket and the new rule being added in the correct priority order. The link from the tree (leaf) to the old bucket may be replaced with a new link from the tree to the new bucket. Establishing this new link to the new bucket is done in one update (which is one part of adding a new bucket) so as not to affect performance or any active searches being performed on the tree.
When there is no space in a bucket to add a rule, according to one embodiment, a subtree is created, the rule is added to a bucket of the subtree, and a leaf of a tree is updated to point to the subtree.
Creating the subtree may take more time than one update but it does not impact the performance and/or functionality of a search. While the subtree is being created, the subtree and tree are separate and distinct from one another. Therefore, the tree may still be searched without searching the subtree. Once the subtree is created, the leaf is updated to point to the subtree instead of the bucket, making the leaf into a node of the tree. Updating the leaf (which is one part of adding a new subtree) takes one update and is atomic. Because the tree includes the subtree after updating, a search of the tree includes the added rule.
As described earlier, a tree represents a set of rules. When adding a new “area” of a rule to the set (e.g., cutting rules) a leaf, or in some cases a node, is added to the tree. To add a leaf/node to a tree, according to one embodiment, includes creating a sibling list of nodes and leaves including the leaf/node being added and then updating a parent node to point to the sibling list.
When a tree of rules (nodes and leaves) and buckets of rules are first created, they are contiguous in memory, e.g., they occupy a memory layout starting at one memory address and ending at another memory address. Over time, adding and deleting rules, as described above, results in “old” leaves and “old” buckets that are no longer referenced. This creates “holes” or “patches” in the memory layout of unused memory. The memory layout is said to be fragmented. To recover these holes or patches in the memory layout, a defragment or defrag mechanism is used. Like the incremental update mechanisms described above, the defrag mechanism is atomic and appears to take one update from the perspective of a search processor performing an active search.
Memory space 1158 shows contiguous memory units 1154a′-f′ corresponding to nodes 1154a-f. Memory space 1156a and 1156b is allocated for the new sibling list 1170 and the node 1154a link is updated such that it no longer points 1157 to the old sibling list 1154b but instead points 1174 to the new sibling list 1170 located in memory units 1156a-b. The link update takes one update and is atomic. The memory space 1154b′, occupied by the “old” level two of the tree is now unused, and illustrates an example of “dead area.”
To recover the dead area (e.g., four memory units in this example), space for the new sibling list (six memory units) may be made by moving, in memory, one or more nodes after the dead area (viz., two nodes at level three of the tree 1152c, which occupy two memory units 1154c′-d′) to the dead area 1154b′. The node is moved by copying the memory unit to a different location and updating the node link. Updating the node link for each node moved takes one update and is atomic. The nodes 1154e and 1154f may then be moved by copying the memory units 1154e′-f′ to 1154c′-d′ and then updating the corresponding links atomically in one update. The new sibling list 1170 may be moved by copying 1156a and 1156b to a location of a designated defragmentation area 1172, for example, by copying 1156a and 1156b to 1156a′ and 1156b′. The link from 1154a to the new sibling list 1170 may then be updated in one update and is atomic. The new sibling list 1170 now located in the defragmentation may be moved once again by copying 1156a′ and 1156b′ to memory units 1154e′-f′ and a needed two node portion of 1156a. The new sibling list may be moved by updating the node 1154a link to point to the new sibling list 1170 relocated. Thus the dead area is now recovered and the memory is utilized in a contiguous manner. Optionally, the one or more nodes moved to the defrag area may be moved out of the defrag area to the end of the memory layout occupied by the tree.
According to one embodiment, deleting a rule may include invalidating a rule in a rule table.
A leaf of a tree may include a pointer to a bucket of rules (bPTR) and a counter of the number of rules in the bucket (nRule). The counter of the number of rules in the bucket may be included and maintained in the housekeeping tree 220.
When a tree walk engine (TWE) walks a tree, the TWE skips over a leaf with a bucket pointer set to null (e.g., the leaf points to a bucket with zero rules). By skipping over the leaf, a bucket pointed to by that leaf (the bucket associated with the leaf) is not passed to a bucket walk engine (BWE) for processing and the bucket may be deleted by setting the bucket pointer to null.
In adding, deleting, and modifying rules, nodes, as described above, node, leaves and/or buckets of a tree may be added, deleted, split, or merged. The tree may become less than optimal, in terms of depth and storage, compared to the tree when it was first compiled. For example, a tree may represent 100 rules. The tree of 100 rules may be divided into 4 leaves/nodes and result on average, in 25 rules per leaf/node. Rules may be deleted and two or more buckets may be merged as described in embodiments herein. A merge may reduce the number leaves/nodes from 4 to 3 and result on average, in 33 rules per leaf/node. One or more sub-trees may be formed. To “improve” the tree, in terms of depth and storage, after incremental updating, a portion of the tree or subtree may be recompiled.
Recompiling a subtree, according to one embodiment, called “adaptive adjustment of tree” includes tracking the number of updates that occur with a counter at each node of the tree. For example, the housekeeping tree 220 may include and maintain the update counter. For a given node, a ratio may be determined from the number of updates tracked for the given node and the number of rules represented by the given node. When this ratio for a given node is greater than a value (which may be preconfigured or dynamically configured, for example, by feedback mechanism), a subtree of the given node may be recompiled.
As incremental updates are received the tree may be modified based on the incremental updates. As a result, quality of the tree may be reduced as the tree may become inefficient from a storage or performance point of view, or a combination thereof. Inefficiency of the tree may be the result of an altered shape or depth of the tree due to tree updates. For example, the tree's shape may become imbalanced as described further below, or the tree may be wider (e.g., shallower) or deeper than necessary. Inefficiency may result due to an insufficient number of children per node. An insufficient number of children per node may result in a reduced number of decision tree paths and longer search times affecting performance of the tree. According to embodiments disclosed herein, heuristics may be employed to determine when to recompilation in the tree and to identify a section of the tree to recompile, enabling efficiency (e.g., quality) of the tree to be maintained.
Embodiments disclosed herein apply heuristics to identify trigger points, enabling a determination of when and where (e.g., at which particular node) in the tree recompilation may improve tree efficiency (e.g., quality). Recompilation may include recompiling a section of the tree, or even the entire tree itself. Because recompilation takes compute time, a trade-off exists for whether or not to sacrifice compute time in order to improve tree quality.
For example, at one extreme, a decision may be made to recompile the tree for each incremental update received. Recompiling the tree based on each incremental update received may maximize a tree's quality; however, such a maximization of quality may come at the expense of maximizing compute time. At another extreme, an alternative decision may be made to never recompile and to simply alter the tree based on the add, delete, or modify incremental update, as disclosed in example embodiments herein. Such an alternative decision may minimize compute time but may come at the expense of minimizing quality of the tree.
Embodiments disclosed herein apply heuristics to trigger when and where in the tree recompilation may improve tree quality, thus guiding decisions that trade-off compute time for tree quality (e.g., how efficient the tree may be from a storage or performance point of view, or a combination thereof). According to embodiments disclosed herein, heuristics may be employed to identify trigger points that identify a particular node to recompile and when to recompile the particular node.
For example, heuristics may be applied to determine when and where to recompile the tree in order to maintain quality of the tree by providing a more balanced distribution of the nodes and leaves of the tree, such that an efficiency (e.g., quality) of the tree may be maintained.
The recompilation may be triggered at a given one or more incremental updates received. For example, one or more incremental updates 1935 may be received, and at the given incremental update, such as an add rule update 1937, the section of the tree 1933 may be identified and recompilation of the identified section may be triggered based on one or more updates determined for the tree based on the incremental update received.
As described herein, incremental updates received may result in one or more updates to the tree. For example, adding a rule to a leaf may result in converting the leaf to a new node that may be linked to one or more new children, altering a shape or depth of the tree. Thus, incremental updates received may affect shape or depth of the tree resulting in an uneven distribution of the nodes or leaves of the tree. According to embodiments disclosed herein, at least one heuristic may be employed to maintain quality of the tree.
By using the at least one heuristic, a section of the tree, such as the section 1933, may be recompiled or the entire tree 1923 may be recompiled to maintain quality of the tree. It should be appreciated that the balanced tree 1913 and the unbalanced tree 1923 are examples of tree shapes and depths and that balanced and unbalanced trees may be of any suitable shape or depth. By employing at least one heuristic to identify a section of the tree and to trigger recompilation of the identified section, embodiments disclosed herein enable the recompilation to be advantageously performed when and where needed, guiding a decision for when to utilize compute time in order to improve quality (e.g., efficiency) of the tree.
According to embodiments disclosed herein, heuristics may be applied to determine when and where to recompile the tree in order to maintain quality of the tree. Embodiments disclosed herein may determine when and where to recompile a section of the tree based on a given depth of the tree, depths of nodes in the tree, a number of children of a node in the tree, or any other suitable characteristics of the tree that represent shape or depth of the tree. According to one embodiment a heuristic may be based on a node having fewer children than a given threshold of children.
Recompilation of a node that has fewer than the given threshold of children may result in an increase in the number of children of the node, thus providing more paths, enabling faster searches and thereby increasing a tree's efficiency from a performance point of view. The node may have fewer than the given threshold of children based on an incremental update resulting in a child of the node being removed. Thus, a time of recompilation may be when the child of the node is removed based on a given one of one or more incremental updates received. Further, the location for the recompilation may be identified as a section of the tree including the node and a subtree below the node.
According to another embodiment disclosed herein, at least one heuristic may be based on a stride of a node. A stride (also referred to herein as a stride value) of a node is a set of one or more bits used for cutting the node, as described below.
According to one embodiment, a total number of node updates since a last recompilation of the node may be tracked at each node. For example, a counter 1953 may track a total number of updates since a last recompilation of the node 1947a. According to embodiments disclosed herein, a ratio of the total number of incremental updates tracked for a node to a number of rules represented by the node at a last recompilation of the node may be computed. For example, a ratio of the total number of incremental updates tracked by the counter 1953 to the number of rules represented by the node 1947a (e.g., N rules) may be computed. According to embodiments disclosed herein, recompilation of the node 1947a may be triggered if the ratio computed is greater than a given update threshold value.
According to one embodiment, the given update threshold may be adjusted based on the stride value of the node. For example, the stride value 1951 of the node 1947a. According to another embodiment, the given threshold may be adjusted based on the depth of the node in the tree. The node's depth may be based on a number of nodes included in a direct path between the node and a root node of the tree, the root node representing the plurality of rules. For example, the number of nodes included in the direct path between the node 288b and the root node 292 of
As such, employing at least one heuristic may include comparing the ratio computed to the given update threshold, thus, detecting whether or not to trigger recompilation of the node and subtree below, altering the tree shape or depth in a manner detected by the at least one heuristic employed.
Maintaining quality may include enabling a balanced shape of the RCDS, increasing a number of decision tree paths in the RCDS, or controlling a number of decision tree levels in the RCDS by recompiling the identified section.
According to embodiments disclosed herein, the RCDS may include a plurality of nodes, each node may represent a subset of the plurality of rules, and the identified section for recompiling may be a node of the plurality of nodes having been cut into a number of other nodes, leaves, or a combination thereof. Recompiling may include selecting one or more fields of the at least one field and re-cutting the node into a new number of new other nodes, new other leaves, or a new combination thereof. Further, according to embodiments disclosed herein, updating the RCDS in a manner enabling the RCDS to incorporate the one or more incremental updates determined and the recompiled identified section atomically from the perspective of an active search process utilizing the RCDS for packet classification.
The processor routines 2492a-b and data 2494a-b are a computer program product, including a computer readable medium (e.g., a removable storage medium, such as one or more DVD-ROM's, CD-ROM's, diskettes, tapes, etc.) that provides at least a portion of the software instructions for embodiments of the invention system. Computer program product 2492a-b may be installed by any suitable software installation procedure, as is well known in the art. In another embodiment, at least a portion of the software instructions may also be downloaded over a cable, communication and/or wireless connection.
Further, embodiments of the present invention may be implemented in a variety of computer architectures. The general computer of
While this invention has been particularly shown and described with references to example embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the scope of the invention encompassed by the appended claims.
Embodiments may be implemented in hardware, firmware, software, or any combination thereof. Embodiments may also be implemented as instructions stored on a non-transient machine-readable medium, which may be read and executed by one or more procedures. A non-transient machine-readable medium may include any mechanism for storing or transmitting information in a form readable by a machine (e.g., a computing device). For example, a non-transitory machine-readable medium may include read only memory (ROM); random access memory (RAM); magnetic disk storage media; optical storage media; flash memory devices; and others. Further, firmware, software, routines, or instructions may be described herein as performing certain actions. However, it should be appreciated that such descriptions are merely for convenience and that such actions in fact result from computing devices, processors, controllers, or other devices executing the firmware, software, routines, instructions, etc.
As will be appreciated by one skilled in the art, techniques disclosed herein may be embodied as a system, method or computer program product. Accordingly, techniques disclosed herein may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, micro-code, etc.) or an embodiment combining software and hardware aspects that may all generally be referred to herein as a “unit,” “module,” or “system.”
It should be understood that the block diagrams may include more or fewer elements, be arranged differently, or be represented differently. It should be understood that implementation may dictate the block and flow diagrams and the number of block and flow diagrams illustrating the execution of techniques disclosed herein.
This application is a continuation-in-part of U.S. application Ser. No. 13/565,775, filed Aug. 2, 2012, which claims the benefit of U.S. Provisional Application No. 61/514,344, filed on Aug. 2, 2011; U.S. Provisional Application No. 61/514,382, filed on Aug. 2, 2011; U.S. Provisional Application No. 61/514,379, filed on Aug. 2, 2011; U.S. Provisional Application No. 61/514,400, filed on Aug. 2, 2011; U.S. Provisional Application No. 61/514,406, filed on Aug. 2, 2011; U.S. Provisional Application No. 61/514,407, filed on Aug. 2, 2011; U.S. Provisional Application No. 61/514,438, filed on Aug. 2, 2011; U.S. Provisional Application No. 61/514,447, filed on Aug. 2, 2011; U.S. Provisional Application No. 61/514,450, filed on Aug. 2, 2011; U.S. Provisional Application No. 61/514,459, filed on Aug. 2, 2011; and U.S. Provisional Application No. 61/514,463, filed on Aug. 2, 2011. The entire teachings of the above applications are incorporated herein by reference.
Number | Name | Date | Kind |
---|---|---|---|
5107361 | Kneidinger et al. | Apr 1992 | A |
5463777 | Bialkowski et al. | Oct 1995 | A |
5909699 | Sarangdhar et al. | Jun 1999 | A |
6233575 | Agrawal et al. | May 2001 | B1 |
6298340 | Calvignac et al. | Oct 2001 | B1 |
6467019 | Washburn | Oct 2002 | B1 |
6578131 | Larson et al. | Jun 2003 | B1 |
6587466 | Bhattacharya et al. | Jul 2003 | B1 |
6735600 | Andreev | May 2004 | B1 |
6778530 | Greene | Aug 2004 | B1 |
6868414 | Khanna et al. | Mar 2005 | B2 |
6980555 | Mar | Dec 2005 | B2 |
7023807 | Michels et al. | Apr 2006 | B2 |
7039641 | Woo | May 2006 | B2 |
7366728 | Corl, Jr. et al. | Apr 2008 | B2 |
7415472 | Testa | Aug 2008 | B2 |
7441022 | Schuba et al. | Oct 2008 | B1 |
7509300 | Sahni et al. | Mar 2009 | B2 |
7522581 | Acharya et al. | Apr 2009 | B2 |
7536476 | Alleyne | May 2009 | B1 |
7546234 | Deb et al. | Jun 2009 | B1 |
7548944 | Sahita | Jun 2009 | B2 |
7571156 | Gupta et al. | Aug 2009 | B1 |
7937355 | Corl et al. | May 2011 | B2 |
8005869 | Corl et al. | Aug 2011 | B2 |
8156507 | Brjazovski et al. | Apr 2012 | B2 |
8447120 | Ji et al. | May 2013 | B2 |
8477611 | Lim | Jul 2013 | B2 |
8856203 | Schelp et al. | Oct 2014 | B1 |
8934488 | Goyal et al. | Jan 2015 | B2 |
8937952 | Goyal et al. | Jan 2015 | B2 |
8937954 | Goyal et al. | Jan 2015 | B2 |
9137340 | Goyal et al. | Sep 2015 | B2 |
9183244 | Bullis et al. | Nov 2015 | B2 |
9191321 | Goyal et al. | Nov 2015 | B2 |
9195939 | Goyal et al. | Nov 2015 | B1 |
9208438 | Goyal et al. | Dec 2015 | B2 |
9430511 | Billa et al. | Aug 2016 | B2 |
9595003 | Bullis et al. | Mar 2017 | B1 |
10083200 | Goyal et al. | Sep 2018 | B2 |
20020023089 | Woo | Feb 2002 | A1 |
20020124086 | Mar | Sep 2002 | A1 |
20020143747 | Tal et al. | Oct 2002 | A1 |
20030115403 | Bouchard et al. | Jun 2003 | A1 |
20030123459 | Liao | Jul 2003 | A1 |
20030135704 | Martin | Jul 2003 | A1 |
20040095936 | O'Neill et al. | May 2004 | A1 |
20050013293 | Sahita | Jan 2005 | A1 |
20050240604 | Corl et al. | Oct 2005 | A1 |
20060026138 | Robertson et al. | Feb 2006 | A1 |
20060098652 | Singh et al. | May 2006 | A1 |
20060136570 | Pandya | Jun 2006 | A1 |
20060155915 | Pereira | Jul 2006 | A1 |
20060221967 | Narayan et al. | Oct 2006 | A1 |
20060253465 | Willis et al. | Nov 2006 | A1 |
20070168377 | Zabarsky | Jul 2007 | A1 |
20080031258 | Acharya et al. | Feb 2008 | A1 |
20080109392 | Nandy | May 2008 | A1 |
20080120441 | Loewenstein | May 2008 | A1 |
20080140631 | Pandya | Jun 2008 | A1 |
20080177994 | Mayer | Jul 2008 | A1 |
20080310440 | Chen et al. | Dec 2008 | A1 |
20090125470 | Shah et al. | May 2009 | A1 |
20090185568 | Cho et al. | Jul 2009 | A1 |
20090274384 | Jakobovits | Nov 2009 | A1 |
20100034202 | Lu et al. | Feb 2010 | A1 |
20100067535 | Ma et al. | Mar 2010 | A1 |
20110038375 | Liu et al. | Feb 2011 | A1 |
20110137930 | Hao et al. | Jun 2011 | A1 |
20110167416 | Sager et al. | Jul 2011 | A1 |
20110219010 | Lim | Sep 2011 | A1 |
20110270889 | Stevens | Nov 2011 | A1 |
20130036102 | Goyal et al. | Feb 2013 | A1 |
20130039366 | Goyal et al. | Feb 2013 | A1 |
20130060727 | Goyal et al. | Mar 2013 | A1 |
20130070753 | Sahni et al. | Mar 2013 | A1 |
20130085978 | Goyal et al. | Apr 2013 | A1 |
20130166886 | Sasanka et al. | Jun 2013 | A1 |
20130201831 | Tal et al. | Aug 2013 | A1 |
20130218853 | Bullis et al. | Aug 2013 | A1 |
20130232104 | Goyal et al. | Sep 2013 | A1 |
20130238576 | Binkert et al. | Sep 2013 | A1 |
20150117461 | Goyal et al. | Apr 2015 | A1 |
20160071016 | Goyal et al. | Mar 2016 | A1 |
Number | Date | Country |
---|---|---|
1535460 | Oct 2004 | CN |
101351784 | Jan 2009 | CN |
101501637 | Aug 2009 | CN |
2002290447 | Oct 2002 | JP |
WO 2009145712 | Dec 2009 | WO |
WO 2013020002 | Feb 2013 | WO |
WO 2013020003 | Feb 2013 | WO |
Entry |
---|
http://en.wikipedia.org/Access_control_list, downloaded Feb. 4, 2011. |
Baboescu, F., et al., “Packet Classification for Core Routers: Is there an alternative to CAMs?”, Proceedings f the 22nd IEEE Conference on Computer Communications (INFOCOM '03), vol. 1, pp. 53-63 (2003). |
Baboescu, F. and Varghese, G., “Scalable Packet Classification,” Proceedings of the ACMSIGCOMM '01 Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication (SIGCOMM '01), pp. 199-210 (2001). |
Gupta, P. and McKeown, N. “Packet Classification on Multiple Fields,” Proceedings of SIGCOMM '99 Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication (SIGCOMM '99), pp. 147-160 (1999). |
Gupta, P. and McKeown, N. “Classifying Packets With Hierarchical Intelligent Cuttings,” IEEE Micro, 20(1):34-41 (2000). |
Qi, Y., et al., “Packet Classification Algorithms: From Theory to Practice,” Proceedings of the 28th IEEE Conference on Computer Communications (INFOCOM '09), pp. 648-656 (2009). |
Singh, S., et al., “Packet Classification Using Multidimensional Cutting,” Proceedings of the ACMSIGCOMM '03 Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication (SIGCOMM '03), pp. 213-224 (2003). |
International Search Report in PCT/US2012/049406 dated Oct. 18, 2012, 3 pages, entitled “Incremental Update of Rules for Packet Classification”. |
Fong et al., HaRP: Rapid Packet Classification via Hashing Round-Down Prefixes, IEEE Transactions on Parallel and Distributed Systems, IEEE Service Center, v. 22(7), pp. 1105-1119 (2011). |
Gupta, P., “Algorithms for Packet Routing Lookups and Packet Classification,” Dissertation submitted to the Dept. of Comp. Science of Stanford Univ. (Dec. 2000). |
Zhang, B., et al., “On Constructing Efficient Shared Decision Trees for Multiple Packet Filters,” Dept. Computer Science Rice University (2010). |
Abdelghani, M., et al. “Packet Classification Using Adaptive Rule Cutting,” IEEE (2005). |
Yu, L., et al., “A Novel IP Packet Classification Algorithm Based on Hierarchical Intelligent Cuttings,” IEEE 6th Int. Conf. on ITS Telecom. Proceedings 1033-1036 (2006). |
Theiling, Henrik “Generating Decision Trees for Decoding Binaries” ACM 2001 [Online] Downloaded Jul. 14, 2015 http://delivery.acm.org/1 0.1145/390000/384213/p112-theiling.pdf?ip=151.207.250.51&id=384213&acc=ACTIVE%20SERVICE&key=C15944E53DOACA63%2E4D4702BOC3E38B35 %2 E4 D4 702BOC3 E38B35%2E4 D4 702BOC3 E38B35&C FI D=528083660&CFTOKEN= 15678279& acm=1436903293 abc. |
Number | Date | Country | |
---|---|---|---|
20130282766 A1 | Oct 2013 | US |
Number | Date | Country | |
---|---|---|---|
61514344 | Aug 2011 | US | |
61514382 | Aug 2011 | US | |
61514379 | Aug 2011 | US | |
61514400 | Aug 2011 | US | |
61514406 | Aug 2011 | US | |
61514407 | Aug 2011 | US | |
61514438 | Aug 2011 | US | |
61514447 | Aug 2011 | US | |
61514450 | Aug 2011 | US | |
61514459 | Aug 2011 | US | |
61514463 | Aug 2011 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 13565775 | Aug 2012 | US |
Child | 13831191 | US |