-
MPNA: A Massively-Parallel Neural Array Accelerator with Dataflow Optimization for Convolutional Neural Networks
Authors:
Muhammad Abdullah Hanif,
Rachmad Vidya Wicaksana Putra,
Muhammad Tanvir,
Rehan Hafiz,
Semeen Rehman,
Muhammad Shafique
Abstract:
The state-of-the-art accelerators for Convolutional Neural Networks (CNNs) typically focus on accelerating only the convolutional layers, but do not prioritize the fully-connected layers much. Hence, they lack a synergistic optimization of the hardware architecture and diverse dataflows for the complete CNN design, which can provide a higher potential for performance/energy efficiency. Towards thi…
▽ More
The state-of-the-art accelerators for Convolutional Neural Networks (CNNs) typically focus on accelerating only the convolutional layers, but do not prioritize the fully-connected layers much. Hence, they lack a synergistic optimization of the hardware architecture and diverse dataflows for the complete CNN design, which can provide a higher potential for performance/energy efficiency. Towards this, we propose a novel Massively-Parallel Neural Array (MPNA) accelerator that integrates two heterogeneous systolic arrays and respective highly-optimized dataflow patterns to jointly accelerate both the convolutional (CONV) and the fully-connected (FC) layers. Besides fully-exploiting the available off-chip memory bandwidth, these optimized dataflows enable high data-reuse of all the data types (i.e., weights, input and output activations), and thereby enable our MPNA to achieve high energy savings. We synthesized our MPNA architecture using the ASIC design flow for a 28nm technology, and performed functional and timing validation using multiple real-world complex CNNs. MPNA achieves 149.7GOPS/W at 280MHz and consumes 239mW. Experimental results show that our MPNA architecture provides 1.7x overall performance improvement compared to state-of-the-art accelerator, and 51% energy saving compared to the baseline architecture.
△ Less
Submitted 30 October, 2018;
originally announced October 2018.
-
A Roadmap Towards Resilient Internet of Things for Cyber-Physical Systems
Authors:
Denise Ratasich,
Faiq Khalid,
Florian Geissler,
Radu Grosu,
Muhammad Shafique,
Ezio Bartocci
Abstract:
The Internet of Things (IoT) is a ubiquitous system connecting many different devices - the things - which can be accessed from the distance. The cyber-physical systems (CPS) monitor and control the things from the distance. As a result, the concepts of dependability and security get deeply intertwined. The increasing level of dynamicity, heterogeneity, and complexity adds to the system's vulnerab…
▽ More
The Internet of Things (IoT) is a ubiquitous system connecting many different devices - the things - which can be accessed from the distance. The cyber-physical systems (CPS) monitor and control the things from the distance. As a result, the concepts of dependability and security get deeply intertwined. The increasing level of dynamicity, heterogeneity, and complexity adds to the system's vulnerability, and challenges its ability to react to faults. This paper summarizes state-of-the-art of existing work on anomaly detection, fault-tolerance and self-healing, and adds a number of other methods applicable to achieve resilience in an IoT. We particularly focus on non-intrusive methods ensuring data integrity in the network. Furthermore, this paper presents the main challenges in building a resilient IoT for CPS which is crucial in the era of smart CPS with enhanced connectivity (an excellent example of such a system is connected autonomous vehicles). It further summarizes our solutions, work-in-progress and future work to this topic to enable "Trustworthy IoT for CPS". Finally, this framework is illustrated on a selected use case: A smart sensor infrastructure in the transport domain.
△ Less
Submitted 6 November, 2018; v1 submitted 16 October, 2018;
originally announced October 2018.
-
Eliminating Tight Coupling using Subscriptions Subgrou** in Structured Overlays
Authors:
Muhammad Shafique
Abstract:
Advertisements and subscriptions are tightly coupled to generate publication routing paths in content--based publish/subscribe systems. Tight coupling requires instantaneous updates in routing tables to generate alternative paths which prevents offering scalable and robust dynamic routing in cyclic overlays when link congestion is detected. We propose, OctopiA, first distributed publish/subscribe…
▽ More
Advertisements and subscriptions are tightly coupled to generate publication routing paths in content--based publish/subscribe systems. Tight coupling requires instantaneous updates in routing tables to generate alternative paths which prevents offering scalable and robust dynamic routing in cyclic overlays when link congestion is detected. We propose, OctopiA, first distributed publish/subscribe system for content--based inter--cluster dynamic routing using purpose--built structured cyclic overlays. OctopiA uses a novel concept of subscription subgrou**, which divides subscriptions into disjoint sets called subscription subgroups. The purpose--built structured cyclic overlay is divided into identical clusters where subscriptions in each subgroup are broadcast to an exclusive cluster. Our advertisement and subscription forwarding algorithms use subscription subgrou** to eliminate tight coupling to offer inter--cluster dynamic routing without requiring updates in routing tables. Experiments on a cluster testbed with real world data show that OctopiA reduces the number of saved advertisements in routing tables by 93%, subscription broadcast delay by 33%, static and dynamic publication delivery delays by 25% and 54%, respectively.
△ Less
Submitted 8 July, 2017; v1 submitted 26 November, 2016;
originally announced November 2016.
-
Adaptive Content-based Routing using Subscription Subgrou** in Structured Overlays
Authors:
Muhammad Shafique
Abstract:
Cyclic or general overlays may provide multiple paths between publishers and subscribers. However, an advertisement tree and a matching subscription activates only one path for notifications routing in publish/subscribe systems. This poses serious challenges in handling network conditions like congestion, and link or broker failures. Further, content-based dynamic routing of notifications requires…
▽ More
Cyclic or general overlays may provide multiple paths between publishers and subscribers. However, an advertisement tree and a matching subscription activates only one path for notifications routing in publish/subscribe systems. This poses serious challenges in handling network conditions like congestion, and link or broker failures. Further, content-based dynamic routing of notifications requires instantaneous updates in routing paths, which is not a scalable option. This paper introduces a clustering approach with a bit-vector technique for inter-cluster dynamic routing of notifications in a structured cyclic topology that provides multiple paths between publishers and interested subscribers. The advertisement forwarding process exploits the structured nature of the overlay topology to generate advertisement trees of length 1 without generating duplicate messages in the advertisement forwarding process. Issued subscriptions are divided into multiple disjoint subgropus, where each subscription is broadcast to a cluster, which is a limited part of the structured cyclic overlay network. We implemented novel static and intra-cluster dynamic routing algorithms in the proposed overlay topology for our advertisement-based publish/subscribe system, called OctopiA. We also performed a pragmatic comparison of our two algorithms with the state-of-the-art. Experiments on a cluster testbed show that our approach generates fewer inter-broker messages, and is scalable.
△ Less
Submitted 22 April, 2016;
originally announced April 2016.
-
Content-based Dynaic Routing in Structured Overlays Networks
Authors:
Muhammad Shafique
Abstract:
Acyclic overlays used for broker-based publish/subscribe systems provide unique paths for content-based routing from a publisher to interested subscribers. Cyclic overlays may provide multiple paths, however, the subscription broadcast process generates one content-based routing path per subscription. This poses serious challenges in offering dynamic routing of notifications when congestion is det…
▽ More
Acyclic overlays used for broker-based publish/subscribe systems provide unique paths for content-based routing from a publisher to interested subscribers. Cyclic overlays may provide multiple paths, however, the subscription broadcast process generates one content-based routing path per subscription. This poses serious challenges in offering dynamic routing of notifications when congestion is detected because instantaneous updates in routing tables are required to generate alternative routing paths. This paper introduces the first subscription-based publish/subscribe system, OctopiS, which offers inter-cluster dynamic routing when congestion in the output queues is detected. OctopiS is based on a formally defined Structured Cyclic Overlay Topology (SCOT). SCOT is divided into homogeneous clusters where each cluster has equal number of brokers and connects to other clusters through multiple inter-cluster overlay links. These links are used to provide parallel routing paths between publishers and subscribers connected to brokers in different clusters. While aiming at deployment at data center networks, OctopiS generates subscription-trees of shortest lengths used by Static Notification Routing (SNR) algorithm. Dynamic Notification Routing (DNR) algorithm uses a bit-vector mechanism to exploit the structuredness of a clustered SCOT to offer inter-cluster dynamic routing without making updates in routing tables and minimizing load on overwhelmed brokers and congested links. Experiments on a cluster testbed with real world data show that OctopiS is scalable and reduces the number of inter-broker messages in subscription delivery by 89%, subscription delay by 77%, end-to-end notification delay in static and dynamic routing by 47% and 58% respectively, and the lengths of output queues of brokers in dynamic routing paths by 59%.
△ Less
Submitted 2 March, 2017; v1 submitted 20 December, 2015;
originally announced December 2015.
