F5 Networks, Inc.

F5, Inc.

Nginx, Inc.

Nikolay Shadrin

Igor Sysoev

Valentin Bartenev

Maxim Romanov

Data is dynamically shared from a first process to a second process by creating a shared memory segment, obtaining a file descriptor referencing the shared memory segment, and mapping the shared memory segment in an address space of a first process. The file descriptor is sent to a second process. Responsive to receiving the file descriptor, the shared memory segment is mapped in an address space of the second process. Via the shared memory segment, data from the first process is shared to the second process.

Using shared memory to transport data between server processes

Filter-based request processing includes generating first data corresponding to a request. A first queue node is generated for processing the first data. The first queue node references a first buffer and a filter subroutine. The first buffer references the first data and a completion handler for performing completion tasks associated with the filter subroutine. The first queue node is executed. The executing includes processing the first data using the filter subroutine to generate a second buffer referencing second data. A second queue node is generated that includes the completion handler. The second queue node is executed. The executing includes processing the completion handler to perform the completion tasks. A response is transmitted corresponding to the request. The response includes the second data referenced by the second buffer.

Filter-based request processing in a web server

Pipelined request processing using shared memory includes writing, by a first process, data associated with a request and an identifier referencing the data to a shared memory segment. The first process transmits, to a second process, the identifier referencing the data. The second process compares the transmitted identifier to the identifier in the shared memory segment. Responsive to the transmitted identifier matching the identifier in the shared memory segment, the second process updates the identifier in the shared memory segment to indicate that the data has been retrieved by the second process. The comparison and update is performed using an atomic compare-and-swap operation. Using the identifiers prevents race conditions between the different processes in trying to access the data. The second process processes the data to generate a response and transmits the response to the first process.

Pipelined request processing using shared memory

A server is dynamically reconfigured by storing a plurality of server configurations in a configuration store. Requests, received by the server, are routed to one of a plurality of workers for processing the requests. Each request is associated with a current configuration of the plurality of configurations that a worker uses to process the request. The number of workers using each configuration of the plurality of configurations is counted. Responsive to the counting, it is determined that a prior configuration of the plurality of configurations is not being used by the workers. The prior configuration is deleted from the configuration store responsive to the determination that the prior configuration is not being used.

Multi-threaded server architecture supporting dynamic reconfiguration

Passing a reference to a first socket from a first process to a second process within a computing environment includes opening, by the first process, a second socket within the computing environment. The second socket is represented by a file in a first file system of the first process. The first process transmits a reference to the first socket to the second socket. A second file system of the second process is mapped to include the file representing the second socket. The second process reads the reference to the first socket from the file representing the second socket mapped to the second file system. The second process references the first socket using the reference to the first socket read from the file.

Growing community of inventors

San Francisco, CA, United States of America

Nikolay Shadrin