For the viability of future “mission-critical” applications such as remote surgery and connected cars, the customers must trust the network connection and operators must adhere to the Service Level Agreement (SLA). The key to enabling trust between the customer and the operator is the transparency and accountability of the SLA. That is, the operators ensure the transparent and appropriate execution of SLA clauses (e.g., Quality of Service (QoS) and compensations if the SLA is violated). In this work, we argue that today’s network is highly volatile. Therefore, it is convenient for the operators to provide service guarantees for short-term rather than traditional long-term methods. Consequently, we advocate short-term and dynamic service contracts considering spatial and temporal characteristics rather than typical long-term contracts. We propose a Distributed Ledger Technology (DLT)-focused end-to-end transparent, accountable and automated resource provisioning system architecture in which are installed as smart contracts. In our architecture, resources can be requested and allocated dynamically and automatically with Quality-of-Service (QoS) monitoring. Our architecture is scalable through a side-channel based QoS monitoring protocol that guarantees data integrity and minimises the Permissioned Distributed Ledgers (PDL) updates. To measure the viability of our proposal, we first evaluate resource provisioning in the context of network slicing. Then we assess the DLT performance for smart contract execution, in both permissioned and permission-less settings. In the end, we evaluate the monitoring tools and compare and contrast sketches and bloomfilters, and justify our choice of bloomfilters.

Connected Cars (CCs) and vehicle-to-everything (V2X) use cases require stringent reliability for safety and non-safety uses. With increasing network softwarisation, it has become easier to use multiple, redundant connectivity options instead of relying on a single network connectivity. But where should these redundant connections be managed? Is it at a network provider’s core network - i.e., supply side managed (SSM) - or at the CC - i.e., demand side managed (DSM)? In our work, we investigate the use of SSM and DSM for CCs on four separate days and across 800 kilometers of major / minor roads in South East England. For Day 1, we captured performance indicators, and determined hypothetical multi-operator configurations for four U.K. providers and a global Universal SIM. For Day 2, 3 & 4, we built and deployed a test-bed to actually implement network switching and understand performance (incl. for TCP & UDP) either on the road or in a stationary location. Based on our results, we make three contributions. First, we show that DSM can deliver superior performance for CCs more than any individual network (up to 28 percentage points in a hypothetical scenario), or SSM which had up to 4.8x longer page load times. Second, unlike other smartphone-only studies, our system-level study demonstrates that improvements of at least 12% can be obtained in a practical DSM field implementation for a CC. Third, we confirm that the advantage of DSM in a field implementation is higher for UDP traffic (23% better latency) compared to TCP (13%).

The viability of new mission-critical networked applications such as connected cars or remote surgery is heavily dependent on the availability of truly customized network services at a Quality of Service (QoS) level that both the network operator and the customer can agree on. This is difficult to achieve in today’s mainly "best effort" Internet. Even if a level of service were to be agreed upon between a consumer and an operator, it is important for both parties to be able to scalably and impartially monitor the quality of service delivered in order to enforce the service level agreement (SLA). Building upon a recently proposed architecture for automated negotiation of SLAs using smart contracts, we develop a low overhead solution for monitoring these SLAs and arranging automated payments based on the smart contracts. Our solution uses cryptographically secure bloom filters to create succinct summaries of the data exchanged over fine-grained epochs. We then use a state channel-based design for both parties to quickly and scalably agree and sign off on the data that was delivered in each epoch, making it possible to monitor and enforce at run time the agreed upon QoS levels.

As adoption of connected cars (CCs) grows, the expectation is that 5G will better support safety-critical vehicle-to-everything (V2X) use cases. Operationally, most relationships between cellular network providers and car manufacturers or users are exclusive, providing a single network connectivity, with at best an occasional option of a back-up plan if the single network is unavailable. We question if this setup can provide QoS assurance for V2X use cases. Accordingly, in this paper, we investigate the role of redundancy in providing QoS assurance for cellular connectivity for CCs. Using our bespoke Android measurement app, we did a drive-through test on 380 kilometers of major and minor roads in South East England. We measured round trip times, jitter, page load times, packet loss, network type, uplink speed and downlink speeds on the four UK networks for 14 UK-centric websites every five minutes. In addition, we did the same measurement using a much more expensive universal SIM card provider that promises to fall back on any of the four UK networks to assure reliability. By comparing actual performance on the best performing network versus the universal SIM, and then projected performance of a two/three/four multi-operator setup, we make three major contributions. First, the use of redundant multi-connectivity, especially if managed by the demand-side, can deliver superior performance (up to 28 percentage points in some cases). Second, despite costing 95x more per GB of data, the universal SIM performed worse than the best performing network except for uplink speed, highlighting how the choice of parameter to monitor can influence operational decisions. Third, any assessment of CC connectivity reliability based on availability is sub-optimal as it can hide significant under-performance.

Datacenter networks are commonly structured with hierarchical topologies in which multipaths are provided to create redundant paths between end-to-end servers. Without a dedicated flow control mechanism, different sizes of traffic flows are statically allocated to links without sensing the current link utilization, which can result in transient network congestion that impacts on the latency experienced by users of the hosted applications. Existing works aim to realize load balancing by efficiently scheduling flows based on their sizes and link utilization. However, many applications' performance is bounded with specific QoS targets, so load balancing may not directly address their QoS requirements. In this paper, we present a latency-aware flow scheduling system that schedules flows based on a tenant application's QoS target and empirical estimations of the effective bandwidth required to meet these QoS targets. The approach seeks to ensure that, following a network-aware placement of VMs hosting application components, the latency associated with communications between these VMs is controlled even as the inter-VM traffic patterns change, or as the background traffic associated with other applications changes. We describe a prototype implementation of the approach that extends our MAPLE network-aware VM placement system. Our experiments in an emulated datacenter indicate that our scheduling approach leads to fewer QoS violations compared to ECMP.

A huge amount of data is everyday produced, even in many critical business sectors, in a distributed, collaborative way with the support of spreadsheet applications. The manual fusion of such data often yields errors: copy-paste operations, loss of alignment and coherency due to multiple spreadsheet