Wireless Power Transmission (WPT) has been widely used to replenish energy for Wireless Rechargeable Sensor Networks. However, the charging service model, which is of the essence to commercial WPT, has not emerged so far. In this paper, we present a wireless charging service model from the perspective of cooperative charging economics, and formulate the Cooperative Charging Scheduling (CCS) problem for joint optimization of rechargeable devices' charging cost and moving cost. We first propose two intragroup cost sharing schemes to sustain the cooperation among devices. Then, the approximation algorithm CCSA of the CCS problem is proposed based on greedy approach and submodular function minimization. Furthermore, we model the large-scale CCS problem as a coalition formation game and present a game theoretic algorithm CCSGA. We show that CCSGA finally converges to a pure Nash Equilibrium. We conduct simulations, and field experiments on a testbed consisting of 5 chargers and 8 rechargeable sensor nodes. The results show that the average comprehensive cost of CCSA is 27.3% lower than the noncooperation algorithm and is only 7.3% higher than the optimal solution on average. In field experiments, CCSA outperforms the noncooperation algorithm by 42.9% in terms of comprehensive cost on average. Moreover, CCSGA is much faster than the approximation algorithm and is more suitable for large-scale cooperative charging scheduling.
{"title":"Cooperative Charging as Service: Scheduling for Mobile Wireless Rechargeable Sensor Networks","authors":"Jia Xu, Suyi Hu, Sixu Wu, Kaijun Zhou, Haipeng Dai, Lijie Xu","doi":"10.1109/ICDCS51616.2021.00071","DOIUrl":"https://doi.org/10.1109/ICDCS51616.2021.00071","url":null,"abstract":"Wireless Power Transmission (WPT) has been widely used to replenish energy for Wireless Rechargeable Sensor Networks. However, the charging service model, which is of the essence to commercial WPT, has not emerged so far. In this paper, we present a wireless charging service model from the perspective of cooperative charging economics, and formulate the Cooperative Charging Scheduling (CCS) problem for joint optimization of rechargeable devices' charging cost and moving cost. We first propose two intragroup cost sharing schemes to sustain the cooperation among devices. Then, the approximation algorithm CCSA of the CCS problem is proposed based on greedy approach and submodular function minimization. Furthermore, we model the large-scale CCS problem as a coalition formation game and present a game theoretic algorithm CCSGA. We show that CCSGA finally converges to a pure Nash Equilibrium. We conduct simulations, and field experiments on a testbed consisting of 5 chargers and 8 rechargeable sensor nodes. The results show that the average comprehensive cost of CCSA is 27.3% lower than the noncooperation algorithm and is only 7.3% higher than the optimal solution on average. In field experiments, CCSA outperforms the noncooperation algorithm by 42.9% in terms of comprehensive cost on average. Moreover, CCSGA is much faster than the approximation algorithm and is more suitable for large-scale cooperative charging scheduling.","PeriodicalId":222376,"journal":{"name":"2021 IEEE 41st International Conference on Distributed Computing Systems (ICDCS)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114659459","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-01DOI: 10.1109/ICDCS51616.2021.00020
Haoqiang Fan, S. Bian, Song Wu, Song Jiang, Shadi Ibrahim, Hai Jin
Containers have been widely used in various cloud platforms as they enable agile and elastic application deployment through their process-based virtualization and layered image system. However, different layers of a container image may contain substantial duplicate and unnecessary data, which slows down its deployment due to long image downloading time and increased burden on the image registry. To accelerate the deployment and reduce the size of the registry, we propose a new image format, named Gear image, that consists of two parts: a Gear index describing the structure of the image's file system and a set of files that are required when running an application. The Gear index is represented as a single-layer image compatible with the existing deployment framework. Containers can be launched by pulling a Gear index and on demand retrieving files pointed to by the index. Furthermore, the Gear image enables a file-level sharing mechanism, which helps remove duplicate data in the registry and avoid repeated downloading of identical files by a client. We implement a prototype of the container framework, named Gear, supporting the new image format. Evaluation shows that Gear saves 54 % storage capacity in the registry, speeds up container startup by up to ${5times}$, and reduces 84 % bandwidth demands.
{"title":"Gear: Enable Efficient Container Storage and Deployment with a New Image Format","authors":"Haoqiang Fan, S. Bian, Song Wu, Song Jiang, Shadi Ibrahim, Hai Jin","doi":"10.1109/ICDCS51616.2021.00020","DOIUrl":"https://doi.org/10.1109/ICDCS51616.2021.00020","url":null,"abstract":"Containers have been widely used in various cloud platforms as they enable agile and elastic application deployment through their process-based virtualization and layered image system. However, different layers of a container image may contain substantial duplicate and unnecessary data, which slows down its deployment due to long image downloading time and increased burden on the image registry. To accelerate the deployment and reduce the size of the registry, we propose a new image format, named Gear image, that consists of two parts: a Gear index describing the structure of the image's file system and a set of files that are required when running an application. The Gear index is represented as a single-layer image compatible with the existing deployment framework. Containers can be launched by pulling a Gear index and on demand retrieving files pointed to by the index. Furthermore, the Gear image enables a file-level sharing mechanism, which helps remove duplicate data in the registry and avoid repeated downloading of identical files by a client. We implement a prototype of the container framework, named Gear, supporting the new image format. Evaluation shows that Gear saves 54 % storage capacity in the registry, speeds up container startup by up to ${5times}$, and reduces 84 % bandwidth demands.","PeriodicalId":222376,"journal":{"name":"2021 IEEE 41st International Conference on Distributed Computing Systems (ICDCS)","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122815636","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-01DOI: 10.1109/ICDCS51616.2021.00129
Zheng-Jun Luo, Tian Pan, Enge Song, Houtian Wang, W. Xue, Tao Huang, Yun-jie Liu
SpaceX plans ambitiously to launch approximately 12,000 satellites from 2019 to 2024, expected to be a complement or even competitor to ground networks. However, the mega-scale satellite network is topology-varying and the frequency of inter-satellite link (ISL) handovers increases rapidly as the topology expands, which will further arouse a massive number of route updates with considerable packet travel delay or even packet loss during the route convergence. The classic Dijkstra's algorithm is adopted for space route calculation, however, it always selects the default shortest path from multiple equal-cost shortest paths between two satellite nodes. To reduce the route change as much as possible during the periodical topology change, in this work, we refined the original Dijkstra and propose StableRoute to select the most appropriate route from the equal-cost candidates with the least route updates compared with the routing table last round. In this way, the end-to-end paths can be maintained as far as possible without time-to-time oscillation. Evaluation shows that it reduces 41% of the route updates in a 36 × 36 topology compared with Dijkstra, and the reduction rate will rise persistently with the growth of the satellite constellation.
{"title":"A Refined Dijkstra's Algorithm with Stable Route Generation for Topology-Varying Satellite Networks","authors":"Zheng-Jun Luo, Tian Pan, Enge Song, Houtian Wang, W. Xue, Tao Huang, Yun-jie Liu","doi":"10.1109/ICDCS51616.2021.00129","DOIUrl":"https://doi.org/10.1109/ICDCS51616.2021.00129","url":null,"abstract":"SpaceX plans ambitiously to launch approximately 12,000 satellites from 2019 to 2024, expected to be a complement or even competitor to ground networks. However, the mega-scale satellite network is topology-varying and the frequency of inter-satellite link (ISL) handovers increases rapidly as the topology expands, which will further arouse a massive number of route updates with considerable packet travel delay or even packet loss during the route convergence. The classic Dijkstra's algorithm is adopted for space route calculation, however, it always selects the default shortest path from multiple equal-cost shortest paths between two satellite nodes. To reduce the route change as much as possible during the periodical topology change, in this work, we refined the original Dijkstra and propose StableRoute to select the most appropriate route from the equal-cost candidates with the least route updates compared with the routing table last round. In this way, the end-to-end paths can be maintained as far as possible without time-to-time oscillation. Evaluation shows that it reduces 41% of the route updates in a 36 × 36 topology compared with Dijkstra, and the reduction rate will rise persistently with the growth of the satellite constellation.","PeriodicalId":222376,"journal":{"name":"2021 IEEE 41st International Conference on Distributed Computing Systems (ICDCS)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116730998","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-01DOI: 10.1109/ICDCS51616.2021.00083
Shan Wang, Ming Yang, Yue Zhang, Yan Luo, Tingjian Ge, Xinwen Fu, Wei Zhao
Hyperledger Fabric is a popular permissioned Blockchain framework for a consortium of organizations to develop Blockchain based applications and transact within the consortium. Hyperledger Fabric introduces a fine-grained access control mechanism called the private data collection (PDC), which allows private data to be shared by only a subset of participants. In this paper, we analyze PDC and show three classes of use cases in which misuse of Hyperledger Fabric features may endanger implemented Hyperledger Fabric systems. We present two groups of potential attacks including fake PDC results injection and PDC leakage against the misuse of the policy based consensus protocol. We use prototype systems to validate the discovered attacks. We also collected 6392 Hyprledger Fabric projects on GitHub and built a tool to statically analyse them. We find that 86.51% of the PDC related projects are potentially vulnerable to the fake PDC results injection attacks, and 91.67% have PDC leakage issues. We design new features for the Hyper-ledger Fabric framework to mitigate the attacks and show that the new features have minor impact on the system performance.
{"title":"On Private Data Collection of Hyperledger Fabric","authors":"Shan Wang, Ming Yang, Yue Zhang, Yan Luo, Tingjian Ge, Xinwen Fu, Wei Zhao","doi":"10.1109/ICDCS51616.2021.00083","DOIUrl":"https://doi.org/10.1109/ICDCS51616.2021.00083","url":null,"abstract":"Hyperledger Fabric is a popular permissioned Blockchain framework for a consortium of organizations to develop Blockchain based applications and transact within the consortium. Hyperledger Fabric introduces a fine-grained access control mechanism called the private data collection (PDC), which allows private data to be shared by only a subset of participants. In this paper, we analyze PDC and show three classes of use cases in which misuse of Hyperledger Fabric features may endanger implemented Hyperledger Fabric systems. We present two groups of potential attacks including fake PDC results injection and PDC leakage against the misuse of the policy based consensus protocol. We use prototype systems to validate the discovered attacks. We also collected 6392 Hyprledger Fabric projects on GitHub and built a tool to statically analyse them. We find that 86.51% of the PDC related projects are potentially vulnerable to the fake PDC results injection attacks, and 91.67% have PDC leakage issues. We design new features for the Hyper-ledger Fabric framework to mitigate the attacks and show that the new features have minor impact on the system performance.","PeriodicalId":222376,"journal":{"name":"2021 IEEE 41st International Conference on Distributed Computing Systems (ICDCS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129633699","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-01DOI: 10.1109/ICDCS51616.2021.00110
Simeon Wuthier, Sang-Yoon Chang
Blockchain and the proof-of-work (PoW) distributed consensus protocol rely on peer-to-peer (P2P) networking. We build a PoW P2P simulator for the modeling and analyses of permissionless blockchain networking. Our simulator utilizes a built-in randomness generator for the simulations, has an easy-to-use interface and intuitive visualization, supports dynamic/programmable control and modifications, and can generate simulation data for further processing. We publish our simulator in open source to facilitate its use for blockchain and P2P networking research and especially recommend it for scalability research or preliminary testing. To highlight its features and capabilities, we demonstrate the simulator use in this paper to analyze the recent blockchain security research, including 51% attack, eclipse, partitioning, and DoS attack.
{"title":"Demo: Proof-of-Work Network Simulator for Blockchain and Cryptocurrency Research","authors":"Simeon Wuthier, Sang-Yoon Chang","doi":"10.1109/ICDCS51616.2021.00110","DOIUrl":"https://doi.org/10.1109/ICDCS51616.2021.00110","url":null,"abstract":"Blockchain and the proof-of-work (PoW) distributed consensus protocol rely on peer-to-peer (P2P) networking. We build a PoW P2P simulator for the modeling and analyses of permissionless blockchain networking. Our simulator utilizes a built-in randomness generator for the simulations, has an easy-to-use interface and intuitive visualization, supports dynamic/programmable control and modifications, and can generate simulation data for further processing. We publish our simulator in open source to facilitate its use for blockchain and P2P networking research and especially recommend it for scalability research or preliminary testing. To highlight its features and capabilities, we demonstrate the simulator use in this paper to analyze the recent blockchain security research, including 51% attack, eclipse, partitioning, and DoS attack.","PeriodicalId":222376,"journal":{"name":"2021 IEEE 41st International Conference on Distributed Computing Systems (ICDCS)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130549446","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-01DOI: 10.1109/ICDCS51616.2021.00068
Zhong-qin Wang, Min Xu, Fu Xiao
State-of-the-art battery-free RFID systems attach multiple RFID tags to an object and exploit their RF phase to estimate its three-dimensional (3D) orientation. However, the measured RF phase may be inaccurate because each tag's signal fingerprint (i.e., RSSI and RF Phase) is distorted by multipath interference and electromagnetic interaction between neighboring tags. In this paper, we propose RF-Orien3D that minimizes these interferences for accurate 3D orientation recognition only using two RFID tags. The electromagnet interference modifies the radiation pattern and modulation factor of each tag in the two-element tag array, which can be estimated to compensate for the distortion in RFID fingerprints. To deal with the multipath impact, we simulate multipath noise to generate huge amounts of RFID fingerprints and use them to pre-train a convolutional neural network (CNN). Then we only collect dozens of actual samples to fine-tune the CNN for multipath-tolerant orientation recognition. The experiments show RF-Orien3D recognizes a two-tag labeled object's 2D orientation with the angular error of about 16° and its 3D orientation (azimuth and elevation) with the errors of about 29° and 11° in low/rich multipath scenarios.
{"title":"Recognizing 3D Orientation of a Two-RFID-Tag Labeled Object in Multipath Environments Using Deep Transfer Learning","authors":"Zhong-qin Wang, Min Xu, Fu Xiao","doi":"10.1109/ICDCS51616.2021.00068","DOIUrl":"https://doi.org/10.1109/ICDCS51616.2021.00068","url":null,"abstract":"State-of-the-art battery-free RFID systems attach multiple RFID tags to an object and exploit their RF phase to estimate its three-dimensional (3D) orientation. However, the measured RF phase may be inaccurate because each tag's signal fingerprint (i.e., RSSI and RF Phase) is distorted by multipath interference and electromagnetic interaction between neighboring tags. In this paper, we propose RF-Orien3D that minimizes these interferences for accurate 3D orientation recognition only using two RFID tags. The electromagnet interference modifies the radiation pattern and modulation factor of each tag in the two-element tag array, which can be estimated to compensate for the distortion in RFID fingerprints. To deal with the multipath impact, we simulate multipath noise to generate huge amounts of RFID fingerprints and use them to pre-train a convolutional neural network (CNN). Then we only collect dozens of actual samples to fine-tune the CNN for multipath-tolerant orientation recognition. The experiments show RF-Orien3D recognizes a two-tag labeled object's 2D orientation with the angular error of about 16° and its 3D orientation (azimuth and elevation) with the errors of about 29° and 11° in low/rich multipath scenarios.","PeriodicalId":222376,"journal":{"name":"2021 IEEE 41st International Conference on Distributed Computing Systems (ICDCS)","volume":"115 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114069238","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-01DOI: 10.1109/ICDCS51616.2021.00034
Yudong Huang, Shuo Wang, Tao Huang, Binwei Wu, Yunxiang Wu, Yun-jie Liu
Recent proposals leverage Time-Aware Shaper (TAS) to achieve precise transmission in Time-Sensitive Networking (TSN). However, most of the proposals require the information of all time-triggered flows to be known in advance and synthesize the gate control list of each switch offline, making the mechanisms they designed inapplicable to industrial automation scenarios where the devices are changed dynamically and the flows should be scheduled online. In this paper, we propose an online routing and scheduling mechanism of TAS for time-sensitive networks. In order to maximize the number of schedulable flows and reduce bandwidth waste, we devise the variable time slot mechanism and minimize the sending start time of each flow. Based on these mechanisms, a novel incremental routing and scheduling (IRAS) algorithm is designed to achieve per-flow deployment, with a pre-routing algorithm to reduce synthesis time. The evaluations show that the IRAS algorithm approaches 96.5 % of the optimal solution in scheduling 2000 flows, and has a feasible per-flow computational time from sub-seconds to less than ten seconds.
{"title":"Online Routing and Scheduling for Time-Sensitive Networks","authors":"Yudong Huang, Shuo Wang, Tao Huang, Binwei Wu, Yunxiang Wu, Yun-jie Liu","doi":"10.1109/ICDCS51616.2021.00034","DOIUrl":"https://doi.org/10.1109/ICDCS51616.2021.00034","url":null,"abstract":"Recent proposals leverage Time-Aware Shaper (TAS) to achieve precise transmission in Time-Sensitive Networking (TSN). However, most of the proposals require the information of all time-triggered flows to be known in advance and synthesize the gate control list of each switch offline, making the mechanisms they designed inapplicable to industrial automation scenarios where the devices are changed dynamically and the flows should be scheduled online. In this paper, we propose an online routing and scheduling mechanism of TAS for time-sensitive networks. In order to maximize the number of schedulable flows and reduce bandwidth waste, we devise the variable time slot mechanism and minimize the sending start time of each flow. Based on these mechanisms, a novel incremental routing and scheduling (IRAS) algorithm is designed to achieve per-flow deployment, with a pre-routing algorithm to reduce synthesis time. The evaluations show that the IRAS algorithm approaches 96.5 % of the optimal solution in scheduling 2000 flows, and has a feasible per-flow computational time from sub-seconds to less than ten seconds.","PeriodicalId":222376,"journal":{"name":"2021 IEEE 41st International Conference on Distributed Computing Systems (ICDCS)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127706913","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-01DOI: 10.1109/ICDCS51616.2021.00045
K. Antoniadis, Antoine Desjardins, V. Gramoli, R. Guerraoui, I. Zablotchi
Classical synchronous consensus algorithms are leaderless: processes exchange their proposals, retain the maximum value and decide when they see the same choice across a couple of rounds. Indulgent consensus algorithms are more robust in that they only require eventual synchrony, but are however typically leader-based. Intuitively, this is a weakness for a slow leader can delay any decision. This paper asks whether, under eventual synchrony, it is possible to deterministically solve consensus without a leader. The fact that the weakest failure detector to solve consensus is one that also eventually elects a leader seems to indicate that the answer to the question is negative. We prove in this paper that the answer is actually positive. We first give a precise definition of the very notion of a leaderless algorithm. Then we present three indulgent leaderless consensus algorithms, each we believe interesting in its own right: (i) for shared memory, (ii) for message passing with omission failures and (iii) for message passing with Byzantine failures (with and without authentication).
{"title":"Leaderless Consensus","authors":"K. Antoniadis, Antoine Desjardins, V. Gramoli, R. Guerraoui, I. Zablotchi","doi":"10.1109/ICDCS51616.2021.00045","DOIUrl":"https://doi.org/10.1109/ICDCS51616.2021.00045","url":null,"abstract":"Classical synchronous consensus algorithms are leaderless: processes exchange their proposals, retain the maximum value and decide when they see the same choice across a couple of rounds. Indulgent consensus algorithms are more robust in that they only require eventual synchrony, but are however typically leader-based. Intuitively, this is a weakness for a slow leader can delay any decision. This paper asks whether, under eventual synchrony, it is possible to deterministically solve consensus without a leader. The fact that the weakest failure detector to solve consensus is one that also eventually elects a leader seems to indicate that the answer to the question is negative. We prove in this paper that the answer is actually positive. We first give a precise definition of the very notion of a leaderless algorithm. Then we present three indulgent leaderless consensus algorithms, each we believe interesting in its own right: (i) for shared memory, (ii) for message passing with omission failures and (iii) for message passing with Byzantine failures (with and without authentication).","PeriodicalId":222376,"journal":{"name":"2021 IEEE 41st International Conference on Distributed Computing Systems (ICDCS)","volume":"206 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131919303","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-01DOI: 10.1109/ICDCS51616.2021.00022
Minchen Yu, Zhifeng Jiang, Hok Chun Ng, Wei Wang, Ruichuan Chen, Bo Li
The increased use of deep neural networks has stimulated the growing demand for cloud-based model serving platforms. Serverless computing offers a simplified solution: users deploy models as serverless functions and let the platform handle provisioning and scaling. However, serverless functions have constrained resources in CPU and memory, making them inefficient or infeasible to serve large neural networks-which have become increasingly popular. In this paper, we present Gillis, a serverless-based model serving system that automatically partitions a large model across multiple serverless functions for faster inference and reduced memory footprint per function. Gillis employs two novel model partitioning algorithms that respectively achieve latency-optimal serving and cost-optimal serving with SLO compliance. We have implemented Gillis on three serverless platforms-AWS Lambda, Google Cloud Functions, and KNIX-with MXNet as the serving backend. Experimental evaluations against popular models show that Gillis supports serving very large neural networks, reduces the inference latency substantially, and meets various SLOs with a low serving cost.
{"title":"Gillis: Serving Large Neural Networks in Serverless Functions with Automatic Model Partitioning","authors":"Minchen Yu, Zhifeng Jiang, Hok Chun Ng, Wei Wang, Ruichuan Chen, Bo Li","doi":"10.1109/ICDCS51616.2021.00022","DOIUrl":"https://doi.org/10.1109/ICDCS51616.2021.00022","url":null,"abstract":"The increased use of deep neural networks has stimulated the growing demand for cloud-based model serving platforms. Serverless computing offers a simplified solution: users deploy models as serverless functions and let the platform handle provisioning and scaling. However, serverless functions have constrained resources in CPU and memory, making them inefficient or infeasible to serve large neural networks-which have become increasingly popular. In this paper, we present Gillis, a serverless-based model serving system that automatically partitions a large model across multiple serverless functions for faster inference and reduced memory footprint per function. Gillis employs two novel model partitioning algorithms that respectively achieve latency-optimal serving and cost-optimal serving with SLO compliance. We have implemented Gillis on three serverless platforms-AWS Lambda, Google Cloud Functions, and KNIX-with MXNet as the serving backend. Experimental evaluations against popular models show that Gillis supports serving very large neural networks, reduces the inference latency substantially, and meets various SLOs with a low serving cost.","PeriodicalId":222376,"journal":{"name":"2021 IEEE 41st International Conference on Distributed Computing Systems (ICDCS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130690876","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-01DOI: 10.1109/ICDCS51616.2021.00109
Heena Nagda, Rakesh Nagda, Nik Sultana, B. T. Loo
Modern programmable network hardware enables in-network computing-pushing increasingly-complex logic into the network to improve the performance, flexibility and reliability of network services. But the current network programming paradigm is constrained to programming a single network device at a time. The lack of support for in-network programs that use several and heterogeneous network hardware simultaneously constrains the scale and behaviour of in-network programs. Dataplane Disaggregation is a new paradigm that addresses this problem. It distributes computations across programmable network hardware including switches and smart NICs. This paradigm transforms a monolithic in-network program into a distributed system executing on possibly heterogeneous resources. The goal of this demo is to make an accessible presentation of Dataplane Disaggregation to the wider distributed systems community. This is intended to stimulate discussion on effective ways to program distributed and heterogeneous systems. Our demo is based on the Flightplan system prototype. Flightplan is open-source and comes with detailed documentation and support scripts, yet it requires some effort to set up and run. This impedes its study by others. Our demo runs completely in the browser and does not burden viewers with any installation effort at all. The technical contribution of this demo consists of a customised visualisation of Flightplan experiments. Moreover, the demo is well-suited to virtual events—as is being planned for ICDCS'21—since it can be run independently and asynchronously by viewers of the demo. This is especially helpful for viewers with slow or intermittent Internet connections. We make the demo's source code freely available online for use by others, including researchers who want to build similar demos.
{"title":"Demo: Disaggregated Dataplanes","authors":"Heena Nagda, Rakesh Nagda, Nik Sultana, B. T. Loo","doi":"10.1109/ICDCS51616.2021.00109","DOIUrl":"https://doi.org/10.1109/ICDCS51616.2021.00109","url":null,"abstract":"Modern programmable network hardware enables in-network computing-pushing increasingly-complex logic into the network to improve the performance, flexibility and reliability of network services. But the current network programming paradigm is constrained to programming a single network device at a time. The lack of support for in-network programs that use several and heterogeneous network hardware simultaneously constrains the scale and behaviour of in-network programs. Dataplane Disaggregation is a new paradigm that addresses this problem. It distributes computations across programmable network hardware including switches and smart NICs. This paradigm transforms a monolithic in-network program into a distributed system executing on possibly heterogeneous resources. The goal of this demo is to make an accessible presentation of Dataplane Disaggregation to the wider distributed systems community. This is intended to stimulate discussion on effective ways to program distributed and heterogeneous systems. Our demo is based on the Flightplan system prototype. Flightplan is open-source and comes with detailed documentation and support scripts, yet it requires some effort to set up and run. This impedes its study by others. Our demo runs completely in the browser and does not burden viewers with any installation effort at all. The technical contribution of this demo consists of a customised visualisation of Flightplan experiments. Moreover, the demo is well-suited to virtual events—as is being planned for ICDCS'21—since it can be run independently and asynchronously by viewers of the demo. This is especially helpful for viewers with slow or intermittent Internet connections. We make the demo's source code freely available online for use by others, including researchers who want to build similar demos.","PeriodicalId":222376,"journal":{"name":"2021 IEEE 41st International Conference on Distributed Computing Systems (ICDCS)","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133523831","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：