Amazon Technologies, Inc.

Leah Shalev

Erez Izenberg

Said Bshara

Avigdor Segal

Jonathan Cohen

Evgeny Schmeilin

Hani Ayoub

Anna Rom-Saksonov

Nitzan Zisman

Shadi Ammouri

Shay Agroskin

Ghanim Fodi

Gilad Broner

Talor Kovary

A different network address is assigned to each port of a first device and a second device that are configured to exchange traffic over a multi-path connection of a transport layer protocol. When a timeout is detected by the first device indicating that a packet sent to a first network address assigned to a first port of the second device was not received, the first device can regenerate the packet with a second network address assigned to a second port of the second device, and send the regenerated packet to the second network address of the second device. The first device can send heartbeat packets to the first port in an exponential backoff manner and, upon receiving a heartbeat acknowledgement from the first port, resume sending the packets to the first port.

Multi-port load balancing using transport protocol

An integrated circuit for distributing processing tasks includes a pre-selector circuit and a scheduler circuit. The pre-selector circuit is configured to receive a processing task, determine a category of the processing task, and select, from a set of task distribution techniques and based at least in part on the category of the processing task, a task distribution technique for distributing the processing task to a group of processing units. The scheduler circuit is configured to implement the selected task distribution technique to select, from the group of processing units, a target processing unit for performing the processing task.

Processing unit selection mechanism

A network device can be used on a client side and a server side to accelerate request/response protocols for object transfers between the client side and the server side. The network device can interface with an application executing on the host device, which can allow the network device to expose different operations that can enable acceleration of different higher level object transfer commands. The network device can expose operations to pre-post anonymous receive buffers and client buffers even before the server side responds to the object transfer commands, which can allow the client buffers to be available to the server early on. Small transfers can be performed without remote direct memory access (RDMA) and can be received in the anonymous receive buffers at the server side. Bulk object data can be sent via RDMA for direct data placement into the client buffers to enable zero copy.

Accelerating request/response protocols

Growing community of inventors

Zikhron Ya'Aqov, Israel

Leah Shalev