Asymmetric Coded Distributed Computation for Resilient Prediction Serving Systems

With the surge of AI services, prediction serving systems (PSSes) have been widely deployed. PSSes are often run on many workers and thus are prone to stragglers (slowdowns or failures), so it is critical to design straggler-resilient PSSes for low latency of prediction. A traditional way is replication that assigns the same prediction job to multiple workers, while incurring significant resources overheads due to its replicated redundant jobs. Recently, coded distributed computation (CDC) has become a more resource-efficient way than replication, as it encodes the prediction job into parity units for prediction reconstruction via decoding. However, we find that state-of-the-art CDC methods either trade accuracy for low latency with the encoder and decoder both simple, or trade latency for high accuracy with the encoder and decoder both complicated, leading to unbalance between accuracy and latency due to the above symmetry between the encoder and decoder. Our insight is that the encoder is always used in CDC-based prediction, while the decoder is only used when stragglers occur. In this paper, we first propose a new asymmetric CDC framework based on the insight, called AsymCDC, composed of a simple encoder but a complicated decoder, such that the encoder’s simplicity makes a low encoding time that reduces the latency largely, while the decoder’s complexity can be helpful for accuracy. Further, we design the decoder’s complexity in two steps: i) an exact decoding method that leverages an invertible neural network’s (INN) invertibility to make the decoding have no accuracy loss, and ii) a decoder compacting method that reshapes INN outputs to utilize knowledge distillation effectively that compacts the decoder for low decoding time. We prototype AsymCDC atop Clipper and experiments show that the prediction accuracy of AsymCDC is approximately the same as state-of-the-arts with the encoder and decoder both complicated, while the latency of AsymCDC only exceeds that of state-of-the-arts with the encoder and decoder both simple by no more than \(2.6\%\) .