Asplos 22 Session 9b Taurus A Data Plane Architecture F In this work, we present the design and implementation of taurus, a data plane for line rate inference. taurus adds custom hardware based on a flexible, parallel patterns (mapreduce) abstraction to programmable network devices, such as switches and nics; this new hardware uses pipelined simd parallelism to enable per packet mapreduce operations (e.g., inference). In a taurus enabled data center, the control plane gathers a global view of the network and trains ml models to optimize security and switch level metrics. meanwhile, the data plane uses these models to make per packet, data driven decisions. unlike traditional sdn based data centers, the control.
Control Plane Vs Data Plane Taurus is a new data plane architecture for switches and network interface controllers (nics) that runs ml models at line rate and uses their output for forwarding decisions. figure 1 — an ml enabled network with a taurus switch. training happens in the control plane while the taurus switch runs per packet inference. Taurus adds custom hardware based on a flexible, parallel patterns (mapreduce) abstraction to programmable network devices, such as switches and nics; this new hardware uses pipelined simd parallelism to enable per packet mapreduce operations (e.g., inference). our evaluation of a taurus switch asic supporting several real world models. Ml inference happens in the data plane data plane packet forwarding (match action) decision making (ml inference) control plane policy creation (flow rules ml training) packets in packets out packet digest ml model weights flow rule software defined network with taurus 10 ml inference is on critical path. Asplos'22: the 27th international conference on architectural support for programming languages and operating systemssession 9b: smart networkingsession cha.
Taurus A Data Plane Architecture For Per Packet Machine Learning Ml inference happens in the data plane data plane packet forwarding (match action) decision making (ml inference) control plane policy creation (flow rules ml training) packets in packets out packet digest ml model weights flow rule software defined network with taurus 10 ml inference is on critical path. Asplos'22: the 27th international conference on architectural support for programming languages and operating systemssession 9b: smart networkingsession cha. In this paper, we propose taurus, a domain specific architecture for per packet ml in data planes. taurus extends the protocolindependent switch architecture (pisa) [15, 63] with a new compute block based on a parallel patterns abstraction (i.e., mapreduce) [88], which supports data parallelism common in modern ml applications [28]. Taurus: a data plane architecture for per packet ml. february 2022. doi: 10.1145 3503222.3507726. conference: asplos '22: 27th acm international conference on architectural support for programming.
юааdataюаб юааplaneюаб Vs Control юааplaneюаб Whatтащs The Difference In this paper, we propose taurus, a domain specific architecture for per packet ml in data planes. taurus extends the protocolindependent switch architecture (pisa) [15, 63] with a new compute block based on a parallel patterns abstraction (i.e., mapreduce) [88], which supports data parallelism common in modern ml applications [28]. Taurus: a data plane architecture for per packet ml. february 2022. doi: 10.1145 3503222.3507726. conference: asplos '22: 27th acm international conference on architectural support for programming.
Comments are closed.