Abstract: Systems and methods for configuring and evaluating policies that direct processing of one or more data streams are described. A configuration interface is described for allowing users to specify object oriented policies. These object oriented policies may allow any data structures to be applied with respect to a payload of a received packet stream, including any portions of HTTP traffic. A configuration interface may also allow the user to control the order in which policies and policy groups are executed, in addition to specifying actions to be taken if one or more policies are undefined. Systems and methods for processing the policies may allow efficient processing of object-oriented policies by applying potentially complex data structures to unstructured data streams. A device may also interpret and process a number of flow control commands and policy group invocation statements to determine an order of execution among a number of policies and policy groups.

Abstract: Systems and methods for configuring and evaluating policies that direct processing of one or more data streams are described. A configuration interface is described for allowing users to specify object oriented policies. These object oriented policies may allow any data structures to be applied with respect to a payload of a received packet stream, including any portions of HTTP traffic. A configuration interface may also allow the user to control the order in which policies and policy groups are executed, in addition to specifying actions to be taken if one or more policies are undefined. Systems and methods for processing the policies may allow efficient processing of object-oriented policies by applying potentially complex data structures to unstructured data streams. A device may also interpret and process a number of flow control commands and policy group invocation statements to determine an order of execution among a number of policies and policy groups.

Abstract: Server state objects are identified by an intermediate server among packets transmitted between an application server and a client device on a network based upon a stateless communcation protocol, by monitoring and analyzing the packets transmitted between the application server and the client device. The packets are parsed into a plurality of name-value pairs. The entropy of the name-value pairs having a same name field is computed, and candidate data objects that are likely to be server state objects are selected based upon the computed entropy. Candidate data objects that were transmitted bi-directionally between the application sever and the client device are identified as server state objects.

Abstract: A computer thread utilizes an auxiliary stack resource to execute program code function calls requiring substantial stack memory resources. In a wrapper embodiment, portions of the program code having function calls requiring substantial stack resources are wrapped in a wrapper that calls an auxiliary stack resource for executing the function call and returns the auxiliary stack resource when the function call is completed. In one embodiment, the stack memory allocated to the thread is selected to accommodate a first class of function calls requiring comparatively low amounts of stack memory and the auxiliary stack resource is called to accommodate the requirements of a second class of function calls requiring comparatively large amounts of stack memory. The allocation of stack memory according to the requirements of function calls allows for multiple computing tasks, such as those carried out by Internet websites and services, to be carried out more efficiently.

Abstract: A method and system for storing data in a hash table of a computer system. A hash function specifies a storage location of the hash table for data associated with a given key. But instead of storing the key in this location, a smaller value, denoted a specification value is used. Specification values are produced via a specification function that mathematically captures the information lost by the hashing function when the key is mapped to the hash address. As a result, no two keys can have equal specification values and equal hash values. When examining the data stored at any given hash address, the invention recreates the key's specification value and compares it with the stored specification value. When the compared values are equal, the object is identified. Thus, the invention conclusively establishes whether the data is associated with a given key without storing the key in the hash table.

Abstract: Packets from multiple packet streams are transmitted in an output stream having substantially equally spaced packets using a scheduler which spaces the packets based upon the input stream speeds of the various input streams as well as the bandwidth of the transmission channel. Such a scheduler advantageously spaces the packets so that the maximum distance between two filled slots is never more than twice the minimum distance between any two filled slots. Such a system also advantageously allows the computation of the slot locations to be precomputed when the slots are initially allocated to streams, thereby allowing the scheduler to efficiently compute slot locations while the scheduler is receiving the input streams.

Abstract: The present invention is a system and method for providing server selection for mirrored sites. In one embodiment, a user requests an address via a local name service (LNS), and the local name service requests the address from a server, examples of which include an authoritative name server, and a Optimizer Name Server (ONS). If the ONS has already calculated a best route which applies to this user, and a predetermined time applying to the best route has not yet expired, then the ONS replies to the user with the best route server. In one embodiment, the best route is determined by having the ONS server request from all the mirrored servers the round trip time to the LNS. This is accomplished by having each mirrored server respond to the LNS with a name query. In response, the LNS replies with an error message. The round trip time is measured from the time the name query was sent to the time the error message was received. Each mirrored server then sends the measured round trip time to the ONS.