Systems research prizes efficiency and performance. For the last decade, I’ve worked on ultra-low-latency systems: the goal of Demikernel and its related nanosecond-scale systems (e.g., Cornflakes, Capybara) is to do more with fewer cycles. AI coding has brought the same level of efficiency to building systems, letting fewer programmers write more code. However, it has also decoupled the speed of…
The Capybara system for microsecond load-balancing with programmable switches has been accepted to SIGCOMM! Congrats to Inho Choi and the entire Capybara team.
Updates on workshop publications this year! Congratulations to Katie Lim on her HotOS paper on operating systems for FPGA accelerated NICs and Inho Choi on his APSys paper on ML-tuned datapath operating systems architecture.
For the last 6 months, the UW Allen School of CSE has been investigating the toxic and sexist environment in the systems and networking labs. Nine doctoral and postdoctoral women were in the lab or joined during the investigation; out of these, two left academia and two are now at other schools. The others remain at UW but avoid the lab by working from MSR or other labs. All have lost up to a year…
Our work, led by Katie Lim , on a modular architecture for NIC-attached accelerators was presented in MICRO 2024 !
I recently reached a few high points in my career that coincided, not coincidentally, with some of the worst harassment of my life. It made me reflect on how my career has been defined as much in terms of misogyny as technical excellence (I’ve garnered quite a CV in both), and how I have struggled to explain to others my experience. This blog post builds on previous work, like Kathryn McKinley’s…
Recently, I spent some time reflecting on my 6 years at Microsoft Research. I have to thank my new management chain – Shan Lu, Ed Nightingale, Desney Tan – for triggering this analysis of how I view my job. My previous management chain certainly spoke about their values and MSR culture, but I had not been at MSR for long enough to think about or understand what they were saying. Certainly, my…
Our work, led by Deepti Raghavan , on zero-copy serialization for Demikernel was accepted to SOSP 2023 !
The research community has long predicted the death of Moore’s law and attendant growth in datacenter hardware speeds. In a few years, datacenter networks will grow an order of magnitude from 40Gb to 400Gb. Systems researchers, including myself, have been preparing for this new world for almost a decade [1, 2]. Demikernel is the first fully-featured nanosecond-scale OS: 100ns to get a packet from…
I’m extremely happy to announce that I will have three papers on kernel-bypass systems in SOSP 2021! Our work on the Demikernel datapath OS , the Persephone scheduler for microsecond RPC and the PRISM RDMA interface for distributed storage.
I spent the bulk of last week doing two types of service that I don’t necessarily enjoy but feel is important: reviewing SOSP papers and dealing with harassment. As an increasingly senior woman in the systems community, I get called on regularly to do both. When I review for SOSP, I lend my expertise to the community by critiquing papers aiming to solve a variety of problems in systems. I’m an…
After reviewing between 50-100 papers per year for the last few years, I’ve noticed some common problems in the papers that I review. While spotting these issues makes it faster for me to review papers, I thought I would summarize them here. Note that these are generally aimed at systems papers but likely apply to other areas as well. Generally speaking, these are the bare minimum requirements for…
Check out our paper on accelerating serialization for microsecond applications at HotOS 2021
Check out our new paper on Persistent State Machines for NVRam to appear at OSDI 2020
Excited to see Marvin, our new memory manager for Android, at USENIX ATC . Also, check out our HotCloud paper on Secure RDMA for Datacenter Storage Systems.
Excited to see our paper on Meerkat, the world’s first multi-core scalable replication protocol, presented at Eurosys 2020 !
The Demikernel/Catnip networking stack is now available on Github . Catnip is a TCP/IP stack for DPDK written from scratch in Rust. We will be presenting its design and implementation at DPDK Summit North America .
The Demikernel is now available open-source on GitHub . This is an initial release, more features are in the works, including SPDK support and a Rust-based TCP stack.
After chatting with people at HotOS, I was inspired to write this blog post summarizing my HotOS paper. At a high-level, the HotOS paper advocates for a new high-level I/O abstraction for kernel-bypass. In the past, kernel-bypass has been associated with giving applications low-level access to devices. We argue that modern kernel-bypass applications should use a high-level abstraction instead,…
I will be presenting Amino, Huawei’s open-source version of Sapphire, at KubeCon Europe and KubeCon China . Read the OSDI 2014 paper .
I’m happy to announce that our work on co-designing network congestion control and OS scheduling has been accepted to HotCloud 2019 . Read the camera-ready .
I’m happy to announce that our paper on the future of datacenter operating systems has been accepted to HotOS 2019 . Read the camera-ready .
I’m honored to have received the ACM SIGOPS Dennis M. Ritchie Dissertation Award for my thesis on mobile/cloud operating systems.
I’m happy to announce that my thesis won the UW CSE William Chan Memorial Dissertation Award .
Our short paper on a unified model for coherence, consistency and isolation has been accepted to PaPoC 2018 .
I have been working on a theoretical paper on a unified theory of consistency, coherence and isolation, so it seemed like a good time to update my blog post on the difference .
I am very excited that my first program committees will be VEE , USENIX ATC and OSDI 2018.
I have finally finished my thesis! It is now available online .
My “Research for Practice” article on distributed transactions is in the recent issue of the CACM .
In the end-of-year rush, I completely missed that the ACM Queue published my invited article on distributed transactions in their “Research for Practice” series. My article summarizes three recent papers – Spanner, TAPIR, and Callas – and discusses why they are important innovations in distributed transactions.
GeekWire selected me as their Geek of the Week, including a very nice write-up on me and my research .
The morning paper blog by Adrian Colyer recently gave a nice summary of Diamond .
I recently presented Diamond at OSDI 2016 in Savannah. If you are interested, you can read the paper or take a look at the code .
I’m excited to be attending the Rising Stars workshop at CMU in October. From the webpage, Rising Stars “brings together the world’s brightest women Ph.D. students, postdocs, and engineers/scientists, for two days of scientific interactions and career-oriented discussions aimed at navigating the early stages of careers in academia.”
The Diamond source code is now available on github and gitlab . Take a look back in a few weeks for a draft of our paper to appear in OSDI 2016.
Our paper on the Diamond automated data management platform will appear in OSDI 2016 and our work on the Disciplined Inconsistency type system will appear at SoCC 2016 .
I just attended the NCWIT Conference on Women and IT to receive honorable mention for the NCWIT Collegiate Award . It was great to see what people in positions of more power than me are doing to increase diversity in computer science.
Crosscut has a great article highlighting CSE’s efforts to bring more women into computer science and keep them there.
After much clean-up, the TAPIR codebase is available! Check it out on github . The repo also includes implementations of Viewstamped Replication and conventional two-phase commit with both optimistic concurrency control and strict two-phase locking.
TL;DR A consistency model only makes sense in the context of a particular system model. Each research area assumes a different system model, so they all ended up with different consistency models that can’t be compared. Doh! For my generals exam, I did a literature review of models for operation ordering from architecture, systems and database communities. I read a cross-section of papers that…
TAPIR has been in the news recently! GeekWire featured us as runners-up for the Madrona Prize. Madrona awards this prize annually at the UW CSE Industrial Affiliates Meeting to to the research projects with the most start-up potential. In other news, the morning paper blog summarizes our work for Twitter readers.
Our paper on building consistent transactions with inconsistent replication will appear in the 25th ACM Symposium on Operating Systems Principles in Monterey, CA. Learn more about SOSP 2015 .
I recently co-chaired the Symposium on Potentially Computer Science (PoCSci ‘15). As many people know, PoCSci is the premier venue for fake computer science research, and there was a lot of exciting work presented this year, especially from our keynote speaker . More on the event at the CSE News .
I recently organized a mini-conference at CSE in the spirit of the annual Grace Hopper Conference as a celebration of women in the department. Ed has some of the background on how the event came to be on the CSE News .
Google has just named me one of the 2015 Anita Borg Scholars. More info about the Google Anita Borg Memorial Scholarship .
Having just finished yet another iteration of the TAPIR paper , I thought I would try to distill what I have learned from the project. It has been a much harder and more interesting project than I anticipated but the photos of baby tapirs helped me get through it. 1. Replication protocols should be co-designed with the systems that use them, not applications. Applications do not commonly use…
UW CSE has created a series of short videos showcasing research in the department and, apparently, I am CSE! Learn more from Ed at the CSE News . Watch the video of me discussing Agate, and Adriana and I attempting to use a mobile device .
I hadn’t thought that I would blog, but since Jekyll makes it so easy, I’m feeling like I need to give it a whirl. No promises about how often this will happen, of course, because the virtual world is littered with dead blogs of the well-intentioned. I thought I would try to clear up, if only for myself, the terminology surrounding consistency and isolation, or, more specifically, operation and…
Microsoft Research has just named me a 2015 Ph.D. Fellow . Thanks to Microsoft for their support of Ph.D. students! As usual, Ed has more to say on the CSE news .
On Friday, the department held a party celebrating graduate counselor, Lindsay Michimoto. Lindsay has been with the department for 15 years, and in that time has processed 14,245 applications to the Ph.D. program (including mine twice!) and advised 602 Ph.D. students. She is definitely one of the very unique features of CSE and is probably responsible for me going to grad school at all. A few PhD…