Ankit Singla

Jiawei Jiang, Shaoduo Gan, Yue Liu et al.

The appeal of serverless (FaaS) has triggered a growing interest on how to use it in data-intensive applications such as ETL, query processing, or machine learning (ML). Several systems exist for training large-scale ML models on top of serverless infrastructures (e.g., AWS Lambda) but with inconclusive results in terms of their performance and relative advantage over "serverful" infrastructures (IaaS). In this paper we present a systematic, comparative study of distributed ML training over FaaS and IaaS. We present a design space covering design choices such as optimization algorithms and synchronization protocols, and implement a platform, LambdaML, that enables a fair comparison between FaaS and IaaS. We present experimental results using LambdaML, and further develop an analytic model to capture cost/performance tradeoffs that must be considered when opting for a serverless infrastructure. Our results indicate that ML training pays off in serverless only for models with efficient (i.e., reduced) communication and that quickly converge. In general, FaaS can be much faster but it is never significantly cheaper than IaaS.

Simon Kassing, Hussain Abbas, Laurent Vanbever et al.

Substantial efforts are invested in improving network security, but the threat landscape is rapidly evolving, particularly with the recent interest in programmable network hardware. We explore a new security threat, from an attacker who has gained control of such devices. While it should be obvious that such attackers can trivially cause substantial damage, the challenge and novelty are in doing so while preventing quick diagnosis by the operator. We find that compromised programmable devices can easily degrade networked applications by orders of magnitude, while evading diagnosis by even the most sophisticated network diagnosis methods in deployment. Two key observations yield this result: (a) targeting a small number of packets is often enough to cause disproportionate performance degradation; and (b) new programmable hardware is an effective enabler of careful, selective targeting of packets. Our results also point to recommendations for minimizing the damage from such attacks, ranging from known, easy to implement techniques like encryption and redundant requests, to more complex considerations that would potentially limit some intended uses of programmable hardware. For data center contexts, we also discuss application-aware monitoring and response as a potential mitigation.

Chen Yu, Hanlin Tang, Cedric Renggli et al.

Most of today's distributed machine learning systems assume {\em reliable networks}: whenever two machines exchange information (e.g., gradients or models), the network should guarantee the delivery of the message. At the same time, recent work exhibits the impressive tolerance of machine learning algorithms to errors or noise arising from relaxed communication or synchronization. In this paper, we connect these two trends, and consider the following question: {\em Can we design machine learning systems that are tolerant to network unreliability during training?} With this motivation, we focus on a theoretical problem of independent interest---given a standard distributed parameter server architecture, if every communication between the worker and the server has a non-zero probability $p$ of being dropped, does there exist an algorithm that still converges, and at what speed? The technical contribution of this paper is a novel theoretical analysis proving that distributed learning over unreliable network can achieve comparable convergence rate to centralized or distributed learning over reliable networks. Further, we prove that the influence of the packet drop rate diminishes with the growth of the number of \textcolor{black}{parameter servers}. We map this theoretical result onto a real-world scenario, training deep neural networks over an unreliable network layer, and conduct network simulation to validate the system improvement by allowing the networks to be unreliable.