Method and apparatus for load balancing based on packet header content

[0001] This disclosure relates generally to communication of packets within a network. More particularly but not exclusively, the present disclosure relates to techniques for making load balancing decisions based on packet header content, such as hypertext transfer protocol (HTTP) header content, and for creating and evaluating logical rules to make such load balancing decisions.

[0002] The most basic unit of data transmission in Transmission Control Protocol/Internet Protocol (TCP/IP) or Internet networking is a packet (sometimes referred to as a “datagram”). A packet is a small piece of information coded at a source, marked with the source address (SA), and directed to a destination address (DA). The SA and DA are typically placed in a header of the packet, while data is placed in a payload of the packet. Traditional IP networks and systems rely exclusively on IP addressing to route the packet from one IP network to another, until arriving at the destination address specified in the packet. Routers, switches (such as Ethernet switches), hubs, or other network devices operate to forward packets to their ultimate destination.

[0003] A packet can have several headers. For instance, a packet can have a Layer 2 header and a Layer 2 payload. The Layer 2 payload in turn can include a Layer 3 header and a Layer 3 payload. The Layer 3 payload can in turn include a Layer 4 header and a Layer 4 payload, and so on. The Layer 4 payload includes an HTTP header and the HTTP payload. Within the HTTP header, there are several fields whose contents are defined by Request for Comments (RFC) standards. Examples include HTTP header fields having information present in hostname, uniform resource locator (URL), and cookie fields.

[0004] For purposes of load balancing network traffic, packets are routed among different servers to ensure that one server does not service all incoming client requests. One technique to load balance is based on the content present in the RFC-defined hostname, URL, and cookie fields of an HTTP header. For example, a packet can be routed to a particular server on the basis of a hostname or host identification (ID) specified in the destination hostname field. As another example, a packet can be directed to a particular server based on a prefix, suffix, or pattern in a URL string of a GET request. As yet another example, cookie switching allows load balancing based on a cookie value, where the cookie value is present as a name value pair in the HTTP header. Accordingly, since the content of these HTTP fields are used for load balancing decisions, this type of routing of packets can be viewed as involving “content switching.”

[0005] While use of the information in these three fields does provide load balancing capability, such use is rather primitive and limited in functionality. For instance, the syntax of these rules can be likened to a simple “if-then” statement (e.g., “if the hostname=X, then route the packet to server 1”). The simple structure and primitive evaluation process of these rules is inadequate for use in more complex load balancing scenarios.

[0006] One aspect of the present invention provides a method that includes defining a plurality of first rules made up of relatively less complex second rules. The method includes defining a policy having at least some of the plurality of first rules and defining corresponding actions to undertake that are related to communication of a packet within a network. The first rules are converted into minterm representations. A bit mask is generated for each of the second rules based on their presence in the minterm representations. The method uses the generated bit masks and content in a header of the packet to evaluate the plurality of first rules in the policy and to determine a corresponding action to undertake.

[0012] Embodiments of techniques to load balance based on packet header content by using and evaluating logical rules are described herein. In the following description, numerous specific details are given to provide a thorough understanding of embodiments. One skilled in the relevant art will recognize, however, that the invention can be practiced without one or more of the specific details, or with other methods, components, materials, etc. In other instances, well-known structures, materials, or operations are not shown or described in detail to avoid obscuring aspects of the invention.

[0013] Reference throughout this specification to “one embodiment” or “an embodiment” means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Thus, the appearances of the phrases “in one embodiment” or “in an embodiment” in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0014] As an overview, one embodiment provides a technique to load balance using content switching. Packets can be routed to a particular server or otherwise processed based on the header content of the packets. In an embodiment, the HTTP header contents of the packets are used in such processing. Rather than making load balancing decisions based solely on the hostname, URL, or cookie fields in the HTTP header, the embodiment can support content switching decisions based on any field in the HTTP header, including both known and unknown fields.

[0015] According to one embodiment, a plurality of content switching policies is provided. A policy comprises a set of rules and actions associated with these rules. Each rule in turn can comprise a simple rule, a rule based on a method type, a nested rule including multiple simple rules, and other rules based on content of fields in the HTTP header. Various embodiments provide example syntax to define these rules. With regards to nested rules, such rules can be complex rules since they contain multiple rules to evaluate. Accordingly, one embodiment simplifies the evaluation of these nested rules by converting the nested rules in to their sum of products or minterm representations, and uses a bit mask technique in conjunction with the minterm representations to determine which nested rules are valid.

[0016] FIG. 1 is a block diagram of an example system 100 in which one embodiment may be implemented. For the sake of simplicity and brevity, not all of the possible components that may be present in the system 100 are shown in FIG. 1 or described in detail herein. Only certain components that are helpful in understanding operation of an embodiment are shown and described.

[0017] The system 100 includes a plurality of client terminals 102 that are communicatively coupled to a network 104. The network 104 is illustrated in FIG. 1 as an Internet, and it is appreciated that other embodiments can be implemented in conjunction with other types of networks, such as local area networks (LANs), virtual private networks (VPNs), and the like.

[0018] One or more routers 106 are coupled to the network 104. Each router is in turn coupled to one or more switches 108. An example of a switch 108 with which an embodiment may be implemented is the ServerIron® product available from Foundry Networks, Inc. of San Jose, Calif. The switch 108 includes one or more processors 110, and one or machine-readable storage media 112. In an embodiment, the storage medium 112 can store software, code, or other machine-readable instructions executable by the processor 110, wherein such machine-readable instructions are used in connection with performing load balancing decisions based on packet header content. In one embodiment, the storage medium 112 stores code that defines rules and load balancing actions to be undertaken based on the rules, including code that executes an algorithm to evaluate the rules and to perform actions based on results of the evaluation of the rules. Additionally, the storage medium 112 can store data (such as in databases, look-up tables, files, or other data structure) that is used in connection with performing the load balancing algorithms and decisions.

[0019] Each of the switches 108 is coupled to a plurality of servers 114. The servers 114 can form part of a server cluster, or can be identified by group identifications (IDs), in some implementations. The servers 114 can provide access to applications, web sites, files and data, or other resources requested by the client terminals 102. A URL, destination address, port number, IP address, hostname, domain name, or other suitable identification mechanism can identify these resources and/or the servers 114. Based on load balancing decisions and other criteria, the switch 108 can decide which particular server 114 to route a request sent from a client terminal 102.

[0020] In one embodiment, the communication between the various components in the system 100 can be performed using packet-based or packet-switching communication techniques. Example protocols are TCP/IP and HTTP, and it is appreciated that some embodiments may be implemented in systems 100 that can use other types of protocols (or combinations thereof).

[0021] FIG. 2 is a diagrammatic illustration of a packet 200, such as a TCP packet, that can be evaluated by one embodiment when making content-based load balancing decisions. The packet 200 includes a header 202, a payload 204, and a trailer 206. The payload 204 can in turn include additional layers headers and payloads. For instance, a Layer 2 payload can contain the Layer 3 header and payload, the Layer 3 payload can contain the Layer 4 header and payload, and so on.

[0022] One of these embedded layers corresponds to an application layer that contains an HTTP packet 208 having an HTTP header 210 and an HTTP payload 212. The HTTP header 210 includes a plurality of fields 214 containing certain pieces of information. For example, some of the fields 214 include a field having a URL string (of a destination for the packet 208), a field having a hostname associated with the destination, and a cookie field. There may be other fields that contain information associated with a server group ID, a destination port, method, control information, and the like. Other fields (and their contents) in the HTTP header 210 may be defined or undefined. The defined fields can be fields whose contents are defined by standards or protocols. The undefined fields have no specific requirement as to their contents or formats, and are generally available for insertion of any type of information deemed useful by users.

[0023] According to an embodiment, decisions as to a particular action to perform on a packet (such as where to forward, whether to drop, persist, redirect, and the like) can be based on information contained in any of the fields 214. Thus, rather than making a load balancing decision (with regards to which server 114 to route a packet) based solely on information contained in just one of the URL, hostname, and cookie fields, an embodiment performs load balancing decisions based on content of any one of the fields 214. These load balancing decisions are not necessarily restricted to content of just one field, but may be based on contents of any multiple and combination of fields, as will be explained in further detail below.

[0024] According to an embodiment, a plurality of content switching (CSW) policies is provided. A CSW policy comprises a set of rules and actions associated with these rules. A CSW policy can have an example command line interface (CLI) syntax of:

[0025] csw-policy <policy-name>: match <rule-name><action>[<action-target>],

[0026] wherein <policy-name> is the name of the particular CSW policy, <rule-name> is the rule being evaluated in the CSW policy, <action> is the action to be taken if the rule is true or matches, and <action-target> is the identification of a destination to which the action is to be applied. The table below lists possible values of <action>, the corresponding possible <action-target> definitions, a description of the action and/or target definition, and corresponding example CSW policies.

[0027]

Action	Target Definition	Description	Examples
Forward	<server-id/group-id>	Forwards the packet to the group-id	match ruleA forward 1
		specified in the action target.
Persist	<offset> <length> <type>	Persist action type is valid for	match ruleD persist 10 20
		header-name rule type.	group-server-id
		Offset - Offset of the content to be	* group-server-id is in the 20
		persisted on	bytes located at the offset 10
		Length - Length of the content to be	from the header-name
		persisted on	specified in ruleD.
		Type - Persistence type. This can be	match ruleD persist 10 20
		group-server-id, hashing or server-	server-name
		name.	server-name is in the 20
		group-server-id: packets are sent to	bytes located at the offset 10
		the group-server-id specified at the	from the header-name
		offset.	specified in ruleD.
		server-name: packets are sent to the	match ruleD persist 1020
		server-name specified at the offset.	hashing
		Hashing: packets with the same hash
		content specified at the offset are
		sent to the same server.
Redirect	<new-domain> [<new-url>]	Redirects the packet to the specified	match ruleB redirect “foo.com”
		domain-name. Optionally one can
		specify the new-url also.
Reply-error	—	Sends an error page to the client	match ruleE reply-error
Reset-client	—	Send a reset to the client	match ruleD reset-client

[0028] To briefly describe some of the entries in the table above, the CSW policy “match ruleA forward 1” in the first table row states that if there is a match of ruleA, then forward the packet to the servers having a group ID of 1. The CSW policy “match ruleB redirect ‘foo.com’” in the third table row states that if there is a match of ruleB, then redirect the packet to the domain “foo.com.” The other example CSW policies, actions, and target definitions are self-explanatory from the table above, and therefore, for the sake of brevity, will not be described in further detail.

[0029] The above table and examples provided <rule-name> examples of “ruleA,” “ruleB,” etc. An example CLI syntax to define any particular rule is:

[0030] Rule: rule-name rule-type <rule-data>,

[0031] wherein the “rule-type” specifies the type of rule involved with that particular <rule-name>, and <rule-data> specifies the fields of an HTTP packet (or other location of the packet) that has data to be evaluated by the rule. Example rule types are nested rule, method, URL, version, header, header-name, XML tag, and other suitable rule types.

[0032] The table below provides more details for the rule types, syntax for <rule-data>, description of the rule type, and examples according to an embodiment.

[0033]

Rule-Type	Rule-Data Syntax	Description	Examples
Method	<eq | ne> <method-string>	This type of rule is used for content switching	csw-rule ruleA
		based on the method type in the http header.	method eq GET
		The two operators supported are eq (equals)
		and ne (not equals) to test the equality of the
		method string.
		Method strings can be one of the following:
		OPTIONS|GET|HEAD|POST|PUT|DELETE|
		TRACE|CONNECT
Nested-rule	<rule-string>	Nested rules are used for defining complex	csw-rule ruleB
		rules with logical operators. The syntax for	“X&(Y|Z|(!W))”
		representing these rules is shown below:	W, X, Y, Z represent
	expr	−> expr & term	simple rules.
		−> expr | term
		−> term
	term	−> !factor
		−> factor
	factor	−> (expr)
		−> ruleId
		In the above example:
		ruleId—name of the simple rule. Nesting of
		complex rules is not allowed in one
		embodiment.
		&—logical “and”
		|—logical “or”
		!—logical “complement”
url	<prefix|suffix|pattern> <type-	This type of rule uses the url in the http	csw-rule ruleC prefix
	value-string>	header to make content switching decision.	“/home”
		The following three operators are supported
		Prefix—looks for the longest prefix specified
		by the type-value-string.
		Suffix—looks for the longest suffix specified
		by the type-value-string.
		Pattern—looks for the pattern specified by
		the type-value-string.
Header-name	<header-name> exists	This type of rule is used for determining	csw-rule ruleF
		whether a given header name exists in the	header-name
		http headers.	“Language” exists.
		Only this type of rule can accept the persist
		rule action in one embodiment.
Header	<header-name>	This type of rule is used for switching based	csw-rule ruleG header
	<equals|prefix|suffix|pattern>	on the value of the http headers. For	“host” prefix
	<header-value>	example, it is possible to switch based on the	“www.yahoo”
		host name in the http header.
xml-tag	<tag-name>	Content switching decision is based on the	csw-rule ruleH
	<prefix|suffix|pattern>	xml-tag present in the packet.	xml-tag “tag1” prefix
	<tag-value>		“test”

[0034] To briefly describe some of the entries in the table above, the first table row defines ruleA based on the method type specified in the HTTP header. Thus in the example, ruleA is defined for packets where the method in the HTTP header is equal to (or contains) a GET string.

[0035] The second table row specifies that nested rules are used for defining complex rules (e.g., ruleB). The complex rules are made up of simple rules (e.g., W, X, Y, Z) that strung together through logical operators. The third table row applies the rule to the URL specified in the HTTP header. The URL may be examined with respect to the prefix, suffix, or other pattern in the URL string.

[0036] The fourth table row defines a rule in terms of whether a given header name exists in the HTTP header. The fifth table row defines a rule based on the content of the HTTP header, which can include a header name, a string (such as prefix, suffix, or pattern), or a header value. For example, the rule might be applied to a packet having a string or host prefix of “www.yahoo.”

[0037] The sixth table row involves a rule wherein content switching is based on the extensible markup language (XML) tag present in the packet. XML information may be present in the packet in a location different from the HTTP header. An embodiment of a technique to use content switching based on XML content in a packet is disclosed in U.S. patent application Ser. No. 10/731,979, entitled “METHOD AND APPARATUS FOR LOAD BALANCING BASED ON XML CONTENT IN A PACKET,” filed concurrently herewith, with inventor Anilkumar Gunturu, assigned to the same assignee as the present application, and incorporated herein by reference in its entirety.

[0038] With regards to nested rules, multiple rules can be combined to define a single rule, using parenthesis and logical operators in an embodiment. The table below explains these logical operators and their use within an expression of a nested rule:

[0039]

csw-rule “rule-name” nested-rule expr
	The syntax of expr is discussed below
----------------------------------------------------------------------
left associative grammar
NOT(!) has higher precedence than AND (&) and OR (|)
----------------------------------------------------------------------
expr	−> expr && term
	−> expr ∥ term
	−> term
term	−> !factor
	−> factor
factor	−> (expr)
	−> rule-id (rule-id is a predefined rule)
Example:
csw-rule: “n_rule” nested-rule “(!ruleA ∥ ruleB) && (ruleC ∥ ruleD)”
ruleA, ruleB, ruleC, ruleD can be either simple rule or multiple rules.

[0040] The table below provides an illustration of the manner in which simple rules may be individually defined or configured at the switches 108 with CLI command, and then combined into complex (nested) rules. An example policy (policy A) that defines content switching actions based on the results of the evaluation of the rules is also provided below.

[0041]

	csw-rule r_url_prefix_1 url prefix “/home”
	csw-rule r_url_prefix_2 url prefix “/news”
	csw-rule r_url_suffix_1 url suffix “.html”
	csw-rule r_url_suffix_2 url suffix “.gif”
	csw-rule r_url_pattern_1 url pattern “cnn”
	csw-rule r_url_pattern_2 url pattern “cbs”
	csw-rule r_url_host_1 header hostname EQ “cnn.com”
	csw-rule r_url_host_2 header hostname EQ “cbs.com”
	csw-rule r_cookie_1 header cookie-name EQ “cookie1”
	csw-rule r_cookie_1 header cookie-name EQ “cookie1”
	csw-nested-rule ruleA (r_url_prefix_1 AND r_cookie_2) OR
	r_url_pattern_1
	csw-nested-rule ruleB (r_url_host_1 AND r_cookie_2)
	OR r_url_pattern_1
	csw-nested-rule ruleC (r_url_prefix_1 AND r_suffix_2) OR
	r_url_pattern_1
	csw-policy policyA
	match rule ruleA forward 1
	match rule ruleB reply-error
	match rule ruleC reset-client
	default 5

[0042] In the above example, the simple rules are defined with regards to content of the URL string (prefix, suffix, and pattern) specified in the HTTP header. Other simple rules are defined with regards to values in the hostname and cookie specified in the HTTP header. Complex nested rules A, B, and C are in turn defined using these simple rules that are expressed using logical operators and parentheses. Finally, the policy A is defined, which specifies the particular action for the switch 108 to undertake based on whether there is a match of any one of the nested rules A, B, or C.

[0043] FIG. 3 is a flowchart 300 of a technique to use complex rules to evaluate contents of a packet in connection with content switching according to an embodiment. Elements of the flowchart 300 (and other flowcharts depicted herein) may be embodied in software, code, or other machine-readable instruction stored on a machine-readable medium. For example, operations represented by the flowchart 300 may be embodied in software stored at the storage medium 112, which can be executed by the processor 110. It is appreciated that the operations depicted in the flowcharts need not necessarily occur in the exact order shown, and that various operations may be added, removed, modified, or combined.

[0044] At a block 302, a plurality of both simple and complex rules is defined. These complex and simple rules may be defined using the techniques illustrated and described with respect to the tables shown above, such as via use of logical operators, parentheses, rule types, and so forth. For load balancing purposes, the rules and policies may be defined in a manner that packet traffic to the servers 114 can be optimally distributed. At the block 302, the CSW policies may also be defined to specify the actions to be taken given one or more rules that are satisfied.

[0045] In an embodiment, the rules and policies may be generated via CLI commands programmed into the storage medium 112 or otherwise stored at a block 304. Any suitable technique may be used to store the rules and policies at the block 304, including but not limited to, storage in databases, file systems, lookup tables, or other data structure. The rules and policies may also be stored as static or dynamic code. For the sake of simplicity of explanation, the rules and policies will be described hereinafter in terms of being stored in a database format, and it is appreciated that the invention is not limited to databases.

[0046] At a block 306, a packet sent from one of the clients 102 is received at the switch 108 (or at some other network location that has the capability to perform content switching). At this point, the switch 108 needs to evaluate the contents of the packet and determine which server 114 to route the packet (or to perform some other action).

[0047] For example at a block 308, the switch 108 can evaluate the contents of the packet by reading the data in its HTTP header fields. Various techniques may be used to read the data in the HTTP header fields. In one embodiment, all fields can be read in sequence. In another embodiment, only certain fields are read and may further be read in some specific order.

[0048] Using the contents of the HTTP header (or other portions of the received packet), the applicable rules are evaluated at a block 310. For example, if the HTTP header includes a GET method string and if a URL string of “foo.com” is present in the HTTP header, then a particular complex rule is evaluated using results of simple rules that are defined using contents of method and URL fields in an HTTP header. More detailed descriptions of techniques for evaluating rules at the block 310 are provided later with respect to FIG. 4 and accompanying tables.

[0049] At a block 312, one or more actions are performed on the packet based on the evaluated rules, in accordance with the applicable CSW policy. For example, the packet may be forwarded to a particular server or redirected. Other actions include persist, error message, reset, and the like.

[0050] As described above with reference to the block 310, the switch 108 determines the action to be performed on an incoming packet after inspecting contents of the incoming packet. The packet inspection process involves matching the contents to the rules defined in a policy database.

[0051] Since, the rules in the policy database may differ widely (e.g., host header rules and URL rules) and the nested nature of the rules (e.g., ruleC=ruleA AND ruleB), an efficient searching strategy is used by an embodiment for the evaluation of rules. An implementation that evaluates one rule after another might not be sufficiently scalable in some instances. For example, consider the following CSW policy configuration:

[0052] if((rule₁ and rule₂) or rule_A)

[0053] {

• • [0054] forward 1;

[0055] }

[0056] else if (rule₃ or rule₄)

[0057] {

• • [0058] forward 2;

[0059] }

[0060] .

[0061] .

[0062] .

[0063] else (rule_n−1 and rule_n)

[0064] {

• • [0065] forward n;

[0066] }

[0067] If n becomes large, evaluating the rules using the above logic will become inefficient. An embodiment of an evaluation algorithm avoids this sequential rule analysis by using bit masks that are calculated during the configuration time.

[0068] Before the details of such an algorithm that handles multiple headers and nested rules are described, a first embodiment of a packet-scanning method for CSW policies comprising rules that belong to only one HTTP header and simple rules is first discussed. In this first embodiment for instance, URL switching makes the load balancing decisions based on the URL field in the HTTP header. A typical example is described below:

[0069] if(url has a prefix “a₁b₁c₁”)

[0070] {

• • [0071] forward 1

[0072] }

[0073] else if(url has a prefix “a₂b₂c₂”)

[0074] {

• • [0075] forward 2

[0076] }

[0077] .

[0078] .

[0079] .

[0080] else if(url has a prefix “a_nb_nc_n”)

[0081] {

• • [0082] forward n

[0083] }

[0084] Such rules are not evaluated one after the other in this first embodiment. Instead, all the patterns (“a₁b₁c₁”, “a₂b₂c₂”, . . . , “a_nb_nc_n”) are stored in a data structure, and the input text from the HTTP header is scanned only once to determine the matching action. This URL switching technique uses finite automaton and trie data structures, for example.

[0085] However, as content switching is extended for more than one HTTP header and nested rules, the scheme of this first embodiment becomes less efficient. Consider the following example in which “&” represents AND, “|” represents OR, and “!” represents NOT:

• • [0086] if((url has a prefix “a₁b₁c₁”) & (host has a pattern “x₁y₁z₁”))
  • [0087] {
    • [0088] forward 1
  • [0089] }
  • [0090] else if((url has a prefix “a₂b₂c₂”) | (cookie has a prefix “x₂y₂z₂”))
  • [0091] {
    • [0092] forward 2
  • [0093] }
  • [0094] .
  • [0095] .
  • [0096] .
  • [0097] else if(((url has a prefix “a_nb_nc_n”) & !(host has a prefix “x_ny_nz_n”))| (cookie has a prefix “x₂y₂z₂”))
  • [0098] {
    • [0099] forward n
  • [0100] }

[0101] Note that in this example, there are multiple operators AND, OR, and NOT, and there is also the presence of different headers url, cookie, and host. This policy may be evaluated sequentially one rule after another, and/or each specific simple rule and complex rule may be evaluated in accordance with the first embodiment. However, such techniques contain multiple rule evaluation steps, which require increased processing capacity, and which therefore may increase latency and delay in the processing (e.g., forwarding) of packets.

[0102] Accordingly, an embodiment provides a more efficient technique for evaluating complex rules and policies. A flowchart 400 of FIG. 4 depicts an embodiment of the technique from the block 310 of FIG. 3 to evaluate CSW policies having complex rules. At a block 402, simple rules are grouped into different databases depending on the rule type. That is, rules of the same type for string comparisons are placed in the same database. For example, all the rules that operate on hostname values are grouped into one database, and all the rules that operate on cookie values are grouped into a different database. Each of these databases can be later evaluated/searched (in a block 408 below) by scanning the packet contents only once, similar to the method used by URL switching technology described earlier, and then searching for matches in each database. Fast pattern matching algorithms, finite automaton, and dynamic prefix trie are some example techniques that can be used to search the databases.

[0103] At a block 404, conversion of nested rules into their sum of products or “minterm” representation is performed. In this minterm form, a nested rule is broken down into a set of minterms summed together. Although the nested rule may have both “&” and “|” operators, each minterm will have only the “&” operator. For example if r_n is a nested rule, and r_a, r_b, and r_c are simple rules satisfying the relation r_n=(r_a|r_b) & r_c, the nested rule r_n can also be represented as r_n=(r_a & r_c)| (r_b & r_c). The minterms of the nested rule r_n are thus (r_a & r_c) and (r_b & r_c). These minterms may be placed in a minterm table or other suitable data structure.

[0104] Since minterms have only the “&” operator, if any of the rules in the minterm is false, the entire minterm is false. This property and the finite automaton searching strategy described earlier, coupled with the bit mask scheme, described later are used to efficiently evaluate the CSW policy. By converting a complex rule into a set of minterms, simpler AND operations can be performed, thereby resulting in a reduction of the processing requirements and operations to evaluate the complex rule.

[0105] It is noted that the operations depicted in the blocks 402-404 and in a subsequent block 406 need not necessarily occur at the block 310 of FIG. 3. The configuration performed during these blocks may be performed at the blocks 302-304 of FIG. 3, for example.

[0106] During the configuration, a bit mask is generated for every rule in each rule type database. The bit length of the bit mask is equal to the total number n of minterms. If n^th bit in the bit mask is binary 0, then the n^th minterm is false—this means that if a particular simple rule is not present in that minterm, that simple rule has a “false” value in that minterm, which makes the entire minterm false. For all the minterms in which the given simple rule is present, the corresponding bit in the bit mask is set to binary 1 or otherwise enabled. All other bits are set to zero or otherwise disabled, unless the corresponding minterm does not have any rules that belong to the same rule type database, in which case the corresponding bit positions are enabled.

[0107] When the packet arrives, its header contents are matched with each of these databases at a block 408. Each database search matches one simple rule. For example, the host field in the HTTP header may specify a particular host pattern or string. The host field database is searched for the specific simple rule that matches that host pattern. As another example, for a URL prefix specified in the HTTP header, the URL database is searched to locate a simple rule that matches that URL prefix. This process is repeated at the block 408 for the other contents of the HTTP header, to locate matching simple rules.

[0108] At a block 410, the previously generated bit masks corresponding to the matching simple rules from the block 408 are obtained. At a block 412, an AND operation is applied to these obtained bit masks to generate a new bit mask. The first bit index in the new bit mask whose value is binary 1 (i.e., non-zero) is located at a block 414 and represents the index of the matched rule from the minterm table. This matched rule from the minterm table corresponds to a set of rules in the policy that is valid, and therefore the corresponding defined action needs to be undertaken by the switch 108.

[0109] To further illustrate the operations depicted in the flowchart 300 of FIG. 3 and in the flowchart 400 of FIG. 4, the following example is provided:

[0110] A. Block 402: The following rule type databases are loaded with definitions for a plurality of simple rules R_u1−3, R_h1−2, and R_c1−2.

[0111] URL Rules

[0112] R_u1=url has a prefix “a₁b₁c₁”

[0113] R_u2=url has a prefix “a₂b₂c₂”

[0114] R_u3=url has a prefix “a₃b₃c₃”

[0115] Host Rules

[0116] R_h1=host has a pattern “x₁y₁z₁”

[0117] R_h2=host has a pattern “x₂y₂z₂”

[0118] Cookie Rules

[0119] R_c1=cookie has a pattern “e₁f₁g₁”

[0120] R_c2=cookie has a pattern “e₂f₂g₂”

[0121] B. Blocks 302-304: The following nested rules and policy are defined.

[0122] Nested Rules

[0123] R_n1=(R_u1 & R_h1)

[0124] R_n2=(R_u2 & R_c1)

[0125] R_n3=(R_u3 | R_h2) & R_c2

[0126] Policy

[0127] Match R_n1 forward 1

[0128] Match R_n2 forward 2

[0129] Match R_n3 forward 3

[0130] C. Block 404: The nested rules are converted into minterms. After the conversion, the policy is represented as below:

[0131]

	Nested Rule	Minterm	Action
	Rn1	M1 = (Ru1 & Rh1)	Forward 1
	Rn2	M2 = (Ru2 & Rc1)	Forward 2
	Rn3	M3 = (Ru3 & Rc2)	Forward 3
	Rn3	M4 = (Rh2 & Rc2)	Forward 3

[0132] D. Block 406: The total number minterms in this example is 4, and hence the bit mask length is 4 bit positions. The following tables explain the creation of bit masks.

[0133]

		Bit
	Minterm	Value	Explanation
Ru1 Mask Bit
Index
1	M1 = (Ru1 & Rh1)	1	Ru1 is present in minterm
2	M2 = (Ru2 & Rc1)	0	Ru1 is not present in minterm
3	M3 = (Ru3 & Rc2)	0	Ru1 is not present in minterm
4	M4 = (Rh2 & Rc2)	1	No rule from URL database
Ru2 Mask Bit
Index
1	M1 = (Ru1 & Rh1)	0	Ru2 is not present in minterm
2	M2 = (Ru2 & Rc1)	1	Ru2 is present in minterm
3	M3 = (Ru3 & Rc2)	0	Ru2 is not present in minterm
4	M4 = (Rh2 & Rc2)	1	No rule from URL database
Ru3 Mask Bit
Index
1	M1 = (Ru1 & Rh1)	0	Ru3 is not present in minterm
2	M2 = (Ru2 & Rc1)	0	Ru3 is not present in minterm
3	M3 = (Ru3 & Rc2)	1	Ru3 is present in minterm
4	M4 = (Rh2 & Rc2)	1	No rule from URL database
Ru3 Mask Bit
Index
1	M1 = (Ru1 & Rh1)	1	Rh1 is present in minterm
2	M2 = (Ru2 & Rc1)	1	No rule from host database
3	M3 = (Ru3 & Rc2)	1	No rule from host database
4	M4 = (Rh2 & Rc2)	0	Rh1 is not present in minterm
Rh2 Mask Bit
Index
1	M1 = (Ru1 & Rh1)	0	Rh2 is not present in minterm
2	M2 = (Ru2 & Rc1)	1	No rule from host database
3	M3 = (Ru3 & Rc2)	1	No rule from host database
4	M4 = (Rh2 & Rc2)	1	Rh2 is present in minterm
Rc1 Mask Bit
Index
1	M1 = (Ru1 & Rh1)	1	No rule from cookie database
2	M2 = (Ru2 & Rc1)	1	Rc1 is present in minterm
3	M3 = (Ru3 & Rc2)	0	Rc1 is not present in minterm
4	M4 = (Rh2 & Rc2)	0	Rc1 is not present in minterm
Rc2 Mask Bit
Index
1	M1 = (Ru1 & Rh1)	1	No rule from cookie database
2	M2 = (Ru2 & Rc1)	1	Rc2 is not present in minterm
3	M3 = (Ru3 & Rc2)	0	Rc2 is present in minterm
4	M4 = (Rh2 & Rc2)	0	Rc2 is present in minterm
Ru1 bit mask = 1001
Ru2 bit mask = 1010
Ru3 bit mask = 1100
Rh1 bit mask = 0111
Rh2 bit mask = 1110
Rc1 bit mask = 0011
Rc2 bit mask = 1101

[0134] E. Blocks 408-414: The following pseudo code for the overall search scheme represents the manner in which each database is searched for a simple rule that matches contents of the HTTP header, and then an AND operation is applied to the bit masks of these matching simple rules to obtain a new bit mask, the first bit position index of which corresponds to a valid matching nested rule from the minterm table of section C above.

[0135] for(i=0; i<num_of_databases; ++i)

[0136] {

• • [0137] rule=search_each_database(i);
  • [0138] new_mask &=rule->mask;

[0139] }

[0140] All of the above U.S. patents, U.S. patent application publications, U.S. patent applications, foreign patents, foreign patent applications and non-patent publications referred to in this specification and/or listed in the Application Data Sheet, are incorporated herein by reference, in their entirety.

[0141] The above description of illustrated embodiments of the invention, including what is described in the Abstract, is not intended to be exhaustive or to limit the invention to the precise forms disclosed. While specific embodiments of, and examples for, the invention are described herein for illustrative purposes, various equivalent modifications are possible within the scope of the invention and can be made without deviating from the spirit and scope of the invention.

[0142] For example, embodiments have been described herein in terms of making load balancing decisions on the basis of HTTP header content. It is appreciated that one embodiment of the invention can also be used to make load balancing decisions on the basis of other header content. With regards to this other header content, logical rules similar to those described herein can be created, and then such rules can be evaluated using the techniques described herein. Accordingly, it is understood that the invention is not necessarily limited to use solely with HTTP header content.

[0143] Moreover, several example formats and syntax for rules, policies, software commands, and the like have been provided herein. It is appreciated that these examples are merely for purposes of explanation and illustration. Other embodiments can use different formats and syntax.

[0144] These and other modifications can be made to the invention in light of the above detailed description. The terms used in the following claims should not be construed to limit the invention to the specific embodiments disclosed in the specification and the claims. Rather, the scope of the invention is to be determined entirely by the following claims, which are to be construed in accordance with established doctrines of claim interpretation.