Hyperscalers dominate the landscape of large network deployments, yet they rarely share data or insights about the challenges they face. In light of this supremacy, what problems can we find to solve in this space? We take an unconventional approach to find relevant research directions, starting from public plans to build a $100 billion datacenter for machine learning applications [53]. Leveraging the language models scaling laws, we discover what workloads such a datacenter might carry and explore the challenges one may encounter in doing so, with a focus on networking research. We conclude that building the datacenter and training such models is technically possible, but this requires novel wide-area transports for inter-DC communication, a multipath transport and novel datacenter topologies for intra-datacenter communication, high speed scale-up networks and transports, outlining a rich research agenda for the networking community.

Clouds offer an opaque I/O API to their customers: details of the underlying resources (network topology, disk drives) or their current load are kept hidden. Tenants can profile the I/O performance in their VMs and optimise accordingly, but the side effect is increased load. Certain cloud providers try to discourage profiling by enforcing strict I/O isolation, at the cost of reduced utilisation in the average case. In this paper we challenge this status quo and propose CloudTalk, an API that allows tenants to communicate with the cloud provider and receive hints used to optimise their workloads.

Modern datacenter networks provide very high capacity via redundant Clos topologies and low switch latency, but transport protocols rarely deliver matching performance. We present NDP, a novel data-center transport architecture that achieves near-optimal completion times for short transfers and high flow throughput in a wide range of scenarios, including incast. NDP switch buffers are very shallow and when they fill the switches trim packets to headers and priority forward the headers. This gives receivers a full view of instantaneous demand from all senders, and is the basis for our novel, high-performance, multipath-aware transport protocol that can deal gracefully with massive incast events and prioritize traffic from different senders on RTT timescales. We implemented NDP in Linux hosts with DPDK, in a software switch, in a NetFPGA-based hardware switch, and in P4. We evaluate NDP's performance in our implementations and in large-scale simulations, simultaneously demonstrating support for very low-latency and high throughput.

In order to improve users ability to interact with devices in an intelligent house, we propose a platform for intelligent device composition, based on Service Oriented Architecture. In such an environment, devices can be controlled and monitored using Service invocations, or they can interoperate in order to fulfil complex tasks, using Service composition. As part of our platform, we define a language for dynamic service composition, called MetaBPEL. It extends the WS-BPEL 2.0 language with semantic information, and acts as an abstract workflow definition mechanism. Unlike BPEL and other conventional composition languages, MetaBPEL does not bind to actual service implementations, but merely describes capabilities that services must have, in order to participate in a workflow. This feature allows both workflow migration between different environments as well as service replacement without modifying the composition scheme. In the article, we describe the MetaBPEL structure and the associated creation, generalization and instantiation mechanisms.