Performing data-intensive analytics is an essential part of modern Earth science. As such, research in atmospheric physics and meteorology frequently requires the processing of very large observational and/or modeled datasets. Typically, these datase
ts (a) have high dimensionality, i.e. contain various measurements per spatiotemporal point, (b) are extremely large, containing observations over a long time period. Additionally, (c) the analytical tasks being performed on these datasets are structurally complex. Over the years, the binary format NetCDF has been established as a de-facto standard in distributing and exchanging such multi-dimensional datasets in the Earth science community -- along with tools and APIs to visualize, process, and generate them. Unfortunately, these access methods typically lack either (1) an easy-to-use but rich query interface or (2) an automatic optimization pipeline tailored towards the specialities of these datasets. As such, researchers from the field of Earth sciences (which are typically not computer scientists) unnecessarily struggle in efficiently working with these datasets on a daily basis. Consequently, in this work, we aim at resolving the aforementioned issues. Instead of proposing yet another specialized tool and interface to work with atmospheric datasets, we integrate sophisticated NetCDF processing capabilities into the established SparkSQL dataflow engine -- resulting in our system Northlight. In contrast to comparable systems, Northlight introduces a set of fully automatic optimizations specifically tailored towards NetCDF processing. We experimentally show that Northlight scales gracefully with the selectivity of the analysis tasks and outperforms the comparable state-of-the-art pipeline by up to a factor of 6x.
We present distributed algorithms for training dynamic Graph Neural Networks (GNN) on large scale graphs spanning multi-node, multi-GPU systems. To the best of our knowledge, this is the first scaling study on dynamic GNN. We devise mechanisms for re
ducing the GPU memory usage and identify two execution time bottlenecks: CPU-GPU data transfer; and communication volume. Exploiting properties of dynamic graphs, we design a graph difference-based strategy to significantly reduce the transfer time. We develop a simple, but effective data distribution technique under which the communication volume remains fixed and linear in the input size, for any number of GPUs. Our experiments using billion-size graphs on a system of 128 GPUs shows that: (i) the distribution scheme achieves up to 30x speedup on 128 GPUs; (ii) the graph-difference technique reduces the transfer time by a factor of up to 4.1x and the overall execution time by up to 40%
In distributed applications, Brewers CAP theorem tells us that when networks become partitioned, there is a tradeoff between consistency and availability. Consistency is agreement on the values of shared variables across a system, and availability is
the ability to respond to reads and writes accessing those shared variables. We quantify these concepts, giving numerical values to inconsistency and unavailability. Recognizing that network partitioning is not an all-or-nothing proposition, we replace the P in CAP with L, a numerical measure of apparent latency, and derive the CAL theorem, an algebraic relation between inconsistency, unavailability, and apparent latency. This relation shows that if latency becomes unbounded (e.g., the network becomes partitioned), then one of inconsistency and unavailability must also become unbounded, and hence the CAP theorem is a special case of the CAL theorem. We describe two distributed coordination mechanisms, which we have implemented as an extension of the Lingua Franca coordination language, that support arbitrary tradeoffs between consistency and availability as apparent latency varies. With centralized coordination, inconsistency remains bounded by a chosen numerical value at the cost that unavailability becomes unbounded under network partitioning. With decentralized coordination, unavailability remains bounded by a chosen numerical quantity at the cost that inconsistency becomes unbounded under network partitioning. Our centralized coordination mechanism is an extension of techniques that have historically been used for distributed simulation, an application where consistency is paramount. Our decentralized coordination mechanism is an extension of techniques that have been used in distributed databases when availability is paramount.
Resource disaggregation has gained huge popularity in recent years. Existing works demonstrate how to disaggregate compute, memory, and storage resources. We, for the first time, demonstrate how to disaggregate network resources by proposing a new di
stributed hardware framework called SuperNIC. Each SuperNIC connects a small set of endpoints and consolidates network functionalities for these endpoints. We prototyped SuperNIC with FPGA and demonstrate its performance and cost benefits with real network functions and customized disaggregated applications.
The salient features of blockchain, such as decentralisation and transparency, have allowed the development of Decentralised Trust and Reputation Management Systems (DTRMS), which mainly aim to quantitatively assess the trustworthiness of the network
participants and help to protect the network from adversaries. In the literature, proposals of DTRMS have been applied to various Cyber-physical Systems (CPS) applications, including supply chains, smart cities and distributed energy trading. In this chapter, we outline the building blocks of a generic DTRMS and discuss how it can benefit from blockchain. To highlight the significance of DTRMS, we present the state-of-the-art of DTRMS in various field of CPS applications. In addition, we also outline challenges and future directions in developing DTRMS for CPS.
While many researchers adopt a sharding approach to design scaling blockchains, few works have studied the transaction placement problem incurred by sharding protocols. The widely-used hashing placement algorithm renders an overwhelming portion of tr
ansactions as cross-shard. In this paper, we analyze the high cost of cross-shard transactions and reveal that most Bitcoin transactions have simple dependencies and can become single-shard under a placement algorithm taking transaction dependencies into account. In addition, we perform a case study of OptChain, which is the state-of-the-art transaction placement algorithm for sharded blockchains, and find a defect of it. A fix is proposed, and our evaluation results demonstrate that the fix helps OptChain improve the system throughput by 4x.
Dynamic languages like Erlang, Clojure, JavaScript, and E adopted data-race freedom by design. To enforce data-race freedom, these languages either deep copy objects during actor (thread) communication or proxy back to their owning thread. We present
Dala, a simple programming model that ensures data-race freedom while supporting efficient inter-thread communication. Dala is a dynamic, concurrent, capability-based language that relies on three core capabilities: immutable values can be shared freely; isolated mutable objects can be transferred between threads but not aliased; local objects can be aliased within their owning thread but not dereferenced by other threads. Objects with capabilities can co-exist with unsafe objects, that are unchecked and may suffer data races, without compromising the safety of safe objects. We present a formal model of Dala, prove data race-freedom and state and prove a dynamic gradual guarantee. These theorems guarantee data race-freedom when using safe capabilities and show that the addition of capabilities is semantics preserving modulo permission and cast errors.
To meet the extreme compute demands for deep learning across commercial and scientific applications, dataflow accelerators are becoming increasingly popular. While these domain-specific accelerators are not fully programmable like CPUs and GPUs, they
retain varying levels of flexibility with respect to data orchestration, i.e., dataflow and tiling optimizations to enhance efficiency. There are several challenges when designing new algorithms and mapping approaches to execute the algorithms for a target problem on new hardware. Previous works have addressed these challenges individually. To address this challenge as a whole, in this work, we present a HW-SW co-design ecosystem for spatial accelerators called Union within the popular MLIR compiler infrastructure. Our framework allows exploring different algorithms and their mappings on several accelerator cost models. Union also includes a plug-and-play library of accelerator cost models and mappers which can easily be extended. The algorithms and accelerator cost models are connected via a novel mapping abstraction that captures the map space of spatial accelerators which can be systematically pruned based on constraints from the hardware, workload, and mapper. We demonstrate the value of Union for the community with several case studies which examine offloading different tensor operations(CONV/GEMM/Tensor Contraction) on diverse accelerator architectures using different mapping schemes.
Decentralized control, low-complexity, flexible and efficient communications are the requirements of an architecture that aims to scale blockchains beyond the current state. Such properties are attainable by reducing ledger size and providing paralle
l operations in the blockchain. Sharding is one of the approaches that lower the burden of the nodes and enhance performance. However, the current solutions lack the features for resolving concurrency during cross-shard communications. With multiple participants belonging to different shards, handling concurrent operations is essential for optimal sharding. This issue becomes prominent due to the lack of architectural support and requires additional consensus for cross-shard communications. Inspired by hybrid Proof-of-Work/Proof-of-Stake (PoW/PoS), like Ethereum, hybrid consensus and 2-hop blockchain, we propose Reinshard, a new blockchain that inherits the properties of hybrid consensus for optimal sharding. Reinshard uses PoW and PoS chain-pairs with PoS sub-chains for all the valid chain-pairs where the hybrid consensus is attained through Verifiable Delay Function (VDF). Our architecture provides a secure method of arranging nodes in shards and resolves concurrency conflicts using the delay factor of VDF. The applicability of Reinshard is demonstrated through security and experimental evaluations. A practical concurrency problem is considered to show the efficacy of Reinshard in providing optimal sharding.
A benchmark study of modern distributed databases is an important source of information to select the right technology for managing data in the cloud-edge paradigms. To make the right decision, it is required to conduct an extensive experimental stud
y on a variety of hardware infrastructures. While most of the state-of-the-art studies have investigated only response time and scalability of distributed databases, focusing on other various metrics (e.g., energy, bandwidth, and storage consumption) is essential to fully understand the resources consumption of the distributed databases. Also, existing studies have explored the response time and scalability of these databases either in private or public cloud. Hence, there is a paucity of investigation into the evaluation of these databases deployed in a hybrid cloud, which is the seamless integration of public and private cloud. To address these research gaps, in this paper, we investigate energy, bandwidth and storage consumption of the most used and common distributed databases. For this purpose, we have evaluated four open-source databases (Cassandra, Mongo, Redis and MySQL) on the hybrid cloud spanning over local OpenStack and Microsoft Azure, and a variety of edge computing nodes including Raspberry Pi, a cluster of Raspberry Pi, and low and high power servers. Our extensive experimental results reveal several helpful insights for the deployment selection of modern distributed databases in edge-cloud environments.
