Full bandwidth packet handling with server systems including offload processors

A distributed system server system for a packet processing without top of rack (TOR) units, comprising:

• a plurality of servers, each server having
• at least one host processor, and
• a plurality of offload processor modules, each offload processor module being connected to memory bus on the server and having an input-output (IO) port and multiple offload processors, each offload processor module also configured to receive network packets from one of the servers through a memory bus and the IO port on the offload processing module;
• and
• a memory controller configured to send network packet data directly to at least one offload processor module via the memory bus to which the offload processor module is attached;
• wherein
• a first of the offload processor modules is connected directly to a second of the offload processor modules through the IO ports of the first and second processor modules, the first and second offload processor modules configured to provide bidirectional network packet flow between at least the first and second offload processor modules and at least one host processor without use of a TOR unit for an inter-server connection.