Amazon Technologies, Inc.

Igor Gorelik

Pablo Puo Hen Cheng

Rosen Ognyanov Baklov

Jesse Aaron Van Beurden

A computing system implemented method is disclosed for distributing data objects in a network of computing nodes. Each node has stored thereon subscription rules defining characteristics of data objects that the particular node may require during processing. The nodes exchange their subscription rules with other nodes in the network. Each node then compares the rules received from other nodes with the attributes of data objects stored on the particular node. Where a rule set matches the attributes of a data object stored thereon, the node marks the particular data object for copying to the node from which the particular rule set was received. The marked data object is then copied to the appropriate node.

Object subscription rule propagation

A computing system implemented method is disclosed for reassigning ownership of a data object between computing nodes. A first computing node having control of a data object transmits a hand-off message indicating control of the data object is being transferred to a second computing node. The first computing node queues requests received at the first computing node relating to the data item. A third computing node that interacts with the data object receives the hand-off message and, in response, delays transmitting requests relating to the data object. When the second computing node receives the hand-off message, it transmits a message claiming control of the data object. In response to the message claiming control, the first computing node and the third computing node transmit their queued requests relating to the data item to the second computing node.

Object ownership migration

Technology for managing state changes of modeled objects in a three dimensional (3D) virtual environment is provided. In one example, a method may include calculating changes to modeled objects in a virtual environment. The modeled objects and the modifications may be stored in a server data store. Modeled objects in the server data store for which the modifications have been received may be flagged and queued for replication of the modifications to the modeled objects from the server data store to a non-transactional data store. The modifications may be replicated by copying modeled object state data from the server data store to the non-transactional data store.

Managing state changes of modeled objects in a three dimensional virtual environment

Computer implemented systems and methods for distributing replica chunks are disclosed. Computing nodes in the network are assigned to zones that correspond to functions performed by the nodes in the network. Replica chunks are likewise associated with the zones that are typically used in processing the replica chunks. When the owner of a replica chunk identifies a new node, it distributes to the new node those replica chunks that are associated with a zone that matches the zone associated with the new node. The owner of the replica chunk is further programmed to receive requests from nodes to subscribe to a data chunk. The node replicates the data chunk to those nodes in the network that have subscribed to the data chunk. When a node no longer requires access to a data chunk, the node unsubscribes from receiving the data chunk.

Selective data replication

A dynamically scalable computing framework used in support of on-line services is disclosed. A network is formed from computing nodes communicatively coupled with communication hubs in a mesh configuration. A load balancing module executing on a computing node monitors processing relating to a plurality of data objects. When a processing threshold is reached, the computing node dynamically reallocates processing of a subset of data objects sharing the same property to a second computing node. Processing nodes in the network receive subscription rules defining characteristics of data objects suitable for copying to another computing node in the network. In response to receiving subscription rules, processing nodes evaluate the rules against data objects on the particular processing node to identifying data objects that should be copied to the other processing nodes corresponding to the received subscription rules.

Scalable network for processing virtual environments

Techniques for distributing processing loads in a multi-node computing environment are described. Ownership of objects of a simulation scenario is assigned to a plurality of computing nodes based on a processing load distribution associated with the simulation scenario. The simulation scenario is executed on the plurality of computing nodes, and processing loads on the computing nodes are monitored. When a processing load of one of the computing nodes has reached a threshold, objects allocated to the one node are reassigned based on properties of a subset of the objects of the simulation scenario. The properties are based on a measure of interaction between the objects in the subset.

Dynamic distribution of simulation load

A computing system is disclosed for reassigning ownership of a data object between computing nodes. A first computing node having control of a data object transmits a hand-off message indicating control of the data object is being transferred to a second computing node. The first computing node queues requests received at the first computing node relating to the data item. A third computing node that interacts with the data object receives the hand-off message and, in response, delays transmitting requests relating to the data object. The third computing node transmits an acknowledgment to the first computing node. Upon receipt of the acknowledgment, the first computing node communicates to the second computing node to assume ownership and transmits any requests that had queued at the first node. When the second computing node receives the message, it transmits a message claiming control of the data object. In response to the message claiming control, the third computing node transmits its queued requests relating to the data item to the second computing node.

Techniques for simulation of objects in a multi-node environment are described herein. Ownership of objects in a simulation scenario is assigned to a plurality of nodes based on a first set of criteria. Simulation authority of the first object is assumed by a second node based on a second set of criteria. Simulation of the first object is performed by the second node without previous acknowledgment, by the first node, of the assumption of simulation authority. Ownership of the first object is maintained by the first node during the time that the second node has simulation authority of the first object.

Growing community of inventors

Irvine, CA, United States of America

Igor Gorelik