eBPF Reading Group Paper 000

# eBPF Reading Group Paper 000 ### Paper 000 Ref: https://arxiv.org/abs/2102.12922 PDF: https://arxiv.org/pdf/2102.12922.pdf ### BPF for storage: an exokernel-inspired approach Yu Jian Wu1, Hongyi Wang1, Yuhong Zhong1, Asaf Cidon1, Ryan Stutsman2, Amy Tai3, and Junfeng Yang1 Published: 2021 1-Columbia University 2-University of Utah 3-VMware Research ## Group Discussion ### Time June 23rd - Time 09:00 Pacific (16:00 UTC) ### Place Invite Only Video Google Meet ### Facilitator Jef Spaleta (eBPF Slack: @jspaleta) ### How to Join Invite for this first meeting is closed. Stay tuned for information on the next meeting. ### Intended Audience English speaking eBPF enthusiasts and industry practicioners of varying levels of technical ability. **Note:** The English speaking intent is just a practical limitation to help facilite discussion in this first time attempting this and is not meant to be exclusionary. If you are native speaker of another language and want to facilite a reading group discussion for the next paper, that would be great! I think ideally there should be a small group for every native language of interest, but this will require native language speaker facilitators to act as host for peer reading groups that run in parallel to the English speaking group. Please contact @jspaleta on the Cilium & eBPF Slack if you want to stand up a new language group. ### Format The discussion will be an informal facilitated remote video round table requiring cameras to be turned on. The facilitor will prepared to prompt initial discussion and to ensure discussion wraps up in a timely manner. This discussion will not be recorded. The intent of the discussion is to be comfortable enough to ask questions about the selected research paper. It's the job of the facilitator and scribe for this session to capture key group take-aways (both understandings and difficulties with the material) as discussion notes on this page. My hope is that the discussion notes can be used to inform future eBPF community content. ### Code of Conduct The reading group discussions will be using the [eBPF Summit Code of Conduct ](https://docs.google.com/document/d/1pmplQdjghadioJVHaGS5X6KrecbYLWVA2VgVz0BzG9E)as an interim CoC, until a specific CoC is needed. The facilitator for each reading group is responsible for enforcing the CoC. ### How to prepare for the meeting All participants are expected to be active in the discussion so come prepared to talk about what you found interesting or what you didn't understand in the paper. At a minimum, make sure you read the paper before coming to the discussion, but you might want to consider reading some of the cited references as well, especially in any section that is hard to understand. Here are some questions you might want to think about while reading the paper and be prepared to discuss in the meeting: * Is this paper currently industry relevant? * After this paper, has there been any relevent upstream kernel discussion related to it? * Have there been any follow up papers citing this work from different authors? Were any of them rebuttable papers? * Are there any open source projects implementing ideas presented in this paper? For those participating with technical ability: * Would you be able to prepare a quick live technical demo of something in this paper for the group to discuss? ### Meeting Structure * Introductions (5 mins) * Intro to the reading group (2 minutes) * This first meeting is an experiment. * Should this be a discussion group or symposium? * Trying a discussion group format and seeing how it goes * Google Meet transcripton is on to help with compiling summary notes after the meeting. * General overview discussion of paper (15 minutes) * Why did I chose this paper? * Is the abstract an accurate summary or did they bury the lede? * Real World Applicability * Does the paper catch all the important pain points that an eBPF implementation have to address? * Is there anything in here you didn't understand or found particular clever and want to call it out? * Quick poll: are the ideas implementable? * Methodoloy discussion (15 minutes) * Is the benchmarking methodology sound? * Any questions about the benchmarking? * Quick poll: Would you be able to reproduce the benchmarks from just the paper's description and cited resources? * Progress since this paper (15 minutes) * Follow-up paper introducing XRP implementation * There's a hint in the follow up paper about combining XRP and XDP into a network attached storage implementation * Anyone try out the XRP github project resources? * Anything in process to be upstreamed into the kernel? * Wrap up/Additional Discussion (remaining time)