When bottleneck buffers are large, loss-based congestion control keeps them full, causing bufferbloat. When bottleneck buffers are small, loss-based congestion control misinterprets loss as a signal of congestion, leading to low throughput. Fixing these problems requires an alternative to loss-based congestion control. Finding this alternative requires an understanding of where and how network congestion originates.

Modern server software is demanding to develop and operate: it must be available at all times and in all locations; it must reply within milliseconds to user requests; it must respond quickly to capacity demands; it must process a lot of data and even more traffic; it must adapt quickly to changing product needs; and in many cases it must accommodate a large engineering organization, its many engineers the proverbial cooks in a big, messy kitchen.

Designing software for modern multicore processors poses a dilemma. Traditional software designs, in which threads manipulate shared data, have limited scalability because synchronization of updates to shared data serializes threads and limits parallelism. Alternative distributed software designs, in which threads do not share mutable data, eliminate synchronization and offer better scalability. But distributed designs make it challenging to implement features that shared data structures naturally provide, such as dynamic load balancing and strong consistency guarantees, and are simply not a good fit for every program. Often, however, the performance of shared mutable data structures is limited by the synchronization methods in use today, whether lock-based or lock-free. To help readers make informed design decisions, this article describes advanced (and practical) synchronization methods that can push the performance of designs using shared mutable data to levels that are acceptable to many applications.

Microservices are an approach to building distributed systems in which services are exposed only through hardened APIs; the services themselves have a high degree of internal cohesion around a specific and well-bounded context or area of responsibility, ...

An everyday problem in our industry is understanding how software is consuming resources, particularly CPUs. What exactly is consuming how much, and how did this change since the last software version? These questions can be answered using software ...

Though widespread interest in software containers is a relatively recent phenomenon, at Google we have been managing Linux containers at scale for more than ten years and built three different container-management systems in that time. Each system was heavily influenced by its predecessors, even though they were developed for different reasons. This article describes the lessons we’ve learned from developing and operating them.

Applying statistical techniques to operations data

Once I was managing a software team and we were working on several initiatives. Projects were assigned based on who was available, their skillsets, and their development goals. This resulted in two developers, Mary and Melissa, being assigned to the same project.

In the long run, the most memorable event of 1989 will probably be that Tim Berners-Lee hacked up the HTTP protocol and named the result the "World Wide Web." Tim’s HTTP protocol ran on 10Mbit/s, Ethernet, and coax cables, and his computer was a NeXT ...

What happened to software engineering? What happened to the promise of rigorous, disciplined, professional practices for software development, like those observed in other engineering disciplines? What has been adopted under the rubric of "software engineering" is a set of practices largely adapted from other engineering disciplines: project management, design and blueprinting, process control, and so forth. The basic analogy was to treat software as a manufactured product, with all the real "engineering" going on upstream of that - in requirements analysis, design, modeling, etc.