An inter-rack server system for a packet processing is disclosed. The system can include a plurality of servers arranged into multiple racks a plurality of top of rack (TOR) units connected to the servers, each TOR unit configured to operate as a TOR switch connecting each of the racks to another of the racks, and a plurality of offload processor modules, each offload processor module having an input-output (IO) port and multiple offload processors, wherein a first offload processor module on a first server on a first rack is connected directly to a second offload processor module on a second server on a second rack, with connection provided through respective IO ports.