Monarch is a globally-distributed in-memory time series database system in Google. Monarch runs as a multi-tenant service and is used mostly to monitor the availability, correctness, performance, load, and other aspects of billion-user-scale applications and systems at Google. Every second, the system ingests terabytes of time series data into memory and serves millions of queries. Monarch has a regionalized architecture for reliability and scalability, and global query and configuration planes that integrate the regions into a unified system. On top of its distributed architecture, Monarch has flexible configuration, an expressive relational data model, and powerful queries. This paper describes the structure of the system and the novel mechanisms that achieve a reliable and flexible unified system on a regionalized distributed architecture. We also share important lessons learned from a decade's experience of developing and running Monarch as a service in Google.

Wireless networks present unusual challenges for mobile file system clients, since they are characterised by unpredictable connectivity and widely-varying bandwidth. The traditional approach to adapting network communication to these conditions is to write back file updates asynchronously when bandwidth is low. Unfortunately, this can lead to underutilisation of bandwidth and inconsistencies between clients. We describe a new mobile file system, MAFS, that supports graceful degradation of file system performance as bandwidth is reduced, as well as rapid propagation of essential file updates. MAFS is able to achieve 10-20% improvements in execution time for real-life file system traces featuring read-write contention.

Applications on mobile computers must adapt to high variability in wireless network performance. Extending the semantics of transport protocols to offer more control over communication to the user allows applications to adapt their behavior to bandwidth variability. We examine adding bandwidth notifications, priorities and timeliness guarantees to a network API as a method for achieving greater application control over bursty traffic. Experiments demonstrate that the extended API allows applications to adjust to bandwidth variations effectively. We also compare three different implementations of the API: two which run on top of TCP, and one new protocol, ATP, which performs comparably to the TCP extensions, but has better performance for some workloads, including a workload simulating remote file system traffic.

In this paper, we describe PortOS, an educational operating system designed to complement undergraduate and graduate level classes on operating systems. PortOS is a complete user-level operating system project, with phases covering concurrency, synchronization, networking and file systems. It focuses particularly on ad hoc and peer-to-peer distributed computing on mobile devices. This paper discusses alternative approaches to operating system projects, and presents our particular design point along with pedagogical justifications.