Red Hat, Inc.

Bela Ban

Vladimir Blagojevic

Manik Surtani

Sanne Grinovero

An example method includes: identifying, by a new node, an address of a shared data store comprising information on a current membership in a peer-to-peer system, wherein the shared data store is shared by a plurality of nodes that are current members of the peer-to-peer system, wherein the shared data store is a container for storing data in a storage cloud; sending, by the new node, a first message comprising an address of the new node to the shared data store; requesting, by the new node, at least one membership data structure from the shared data store; receiving, by the new node, a second message comprising the at least one membership data structure; generating, by the new node, a new membership data structure comprising the address of the new node and the plurality of addresses for the plurality of nodes identified in the at least one membership data structure; sending, by the new node, a third message comprising the new membership data structure to the shared data store; and joining, by the new node, the peer-to-peer system, wherein the joining comprises using the new membership data structure to identify nodes of the plurality of nodes to receive a plurality of messages from the new node.

Using a shared data store for peer discovery

A method for light-weight fork channels for clustering is disclosed. The method includes receiving, by a processing device, a message at a main channel used for group communication between processes executed by the processing device, identifying a fork channel identifier (ID) and a fork stack ID in a header of the message, processing the message with a fork stack corresponding to the fork stack ID in the header of the message, the fork stack comprising a subset of protocols of a main protocol stack of the main channel, and providing the message to a fork channel corresponding to the fork channel ID, wherein the fork channel to utilize the fork stack to separate messages for the fork channel from the main channel.

Light-weight fork channels for clustering

After acquiring a network address, a computing device accesses a shared data store and writes the network address to the shared data store. The computing device additionally reads a plurality of network addresses from the shared data store, wherein the plurality of network addresses are for a plurality of nodes that are members of a peer-to-peer system. The computing device then joins the peer-to-peer system based on communicating with the plurality of nodes using the plurality of network addresses.

A method and apparatus for processing message is described. In one embodiment, an application programming interface is configured for receiving and sending messages. A multiplexer receives messages from different servers. A service name is coupled to each message with the corresponding destination service. A single shared channel is formed. The messages are processed over the single shared channel.

Message handling multiplexer

Systems and methods are disclosed for replicating data stored in an in-memory data cache to a remote site. An example system includes an in-memory data cache and an in-memory keys cache. The system also includes a key insert module that detects a modification to the in-memory data cache, identifies one or more keys of the plurality of keys based on the modification, and inserts the identified one or more keys into the in-memory keys cache. The system further includes an update module that retrieves from the in-memory keys cache a set of keys, retrieves from the in-memory data cache modified data associated with the set of keys, and transmits to a remote site a modification list including the set of keys and the modified data associated with the set of keys.

Replication between sites using keys associated with modified data

A replication count of a data element of a node of a cache cluster is defined. The data element has a key-value pair where the node is selected based on a hash of the key and a size of the cache cluster. The data element is replicated to at least one other node of the cache cluster based on the replication count.

Caching of nodes in cache cluster

A method and apparatus for processing message is described. In one embodiment, messages are received over a plurality of channels from a plurality of applications in a virtual machine. An identifier is coupled to each message. The identifier refers to the application originating the corresponding message. A shared transport is formed and associated with the channels. The messages are processed with the shared transport with the identifier.

Shared transport

An apparatus and a method for processing messages. A transport protocol stack receives messages from senders. The transport protocol stack comprises a concurrent stack having an out of band thread pool and a non-out of band thread pool. A channel is formed with a thread from the non-out of band thread pool of the concurrent stack for each sender. Non-out of band messages are processed in parallel for each scope of an application with a non-out of band thread. Non-out of band messages pertains to processes of the application from a same sender. Processes from a scope of the application are unrelated to processes from another scope of the application. Non-out of band messages within a same scope are processed in a predefined order.

Concurrent delivery for messages from a same sender

A method and apparatus for processing message is described. In one embodiment, an application programming interface is configured for receiving and sending messages. A message credit account of a sender is maintained. The message credit account of the sender is debited when a message is received from the sender to a receiver. The message credit account of the sender is credited when a replenishment message is sent to the sender in response to the receiver processing the received message. The sender is prevented from sending any messages to the receiver when a balance in the message credit account of the sender becomes negative.

Flow control protocol

A method and apparatus for processing messages is described. In one embodiment, an application programming interface provides for a flush protocol to force members of a group to send all of their pending messages prior to a predetermined event. A client sends a request to a coordinator of the group to join the group and to transfer a state of the group to the client. The application programming interface performs a single flush operation on the group in response to the request.

Combined join and state transfer

A method and apparatus for processing message includes forming a logical address upon a creation of a channel of a group communication protocol at a logical address generator of a server. The logical address includes a unique identifier (UUID) associated with a logical name. The UUID created upon connecting the channel, and deleted upon disconnecting the channel. A mapping between UUIDs and physical addresses of members of the group is stored in a cache of a member of the group.

Mechanism for processing messages using logical addresses

A method and apparatus for interfacing a drive with a tree-structured cluster as a library for a data grid storage system is described. In one embodiment, the tree-structured cluster of cache nodes is provided for storing data. The cluster is mounted as a drive or file system based on data stored in the cache nodes transparent to anyone using the drive or file system. A file system interface interfaces with the tree-structure cluster of cache nodes to presents data stored in the cache nodes.

Grid based file system

A method and apparatus for processing message is described. In one embodiment, an application programming interface is configured for receiving and sending messages. A building block layer is coupled to the application programming interface. A channel layer is coupled to the building block layer. A transport protocol stack is coupled to the channel layer for implementing properties specified by the channel layer. The transport protocol stack has a concurrent stack consisting of an out of band thread pool and a regular thread pool. The transport protocol layer is to process messages from each sender in parallel with the corresponding channel for each sender.

Queuing for thread pools using number of bytes

A method and apparatus for processing message is described. In one embodiment, a channel layer is configured to form a channel with a corresponding process. The channel is associated with a group. A plurality of sockets is formed in a transport layer of a group communication protocol. Traffic of messages to and from the processes is distributed across the plurality of sockets in the transport layer.

Growing community of inventors

Kreuzlingen, Switzerland

Bela Ban