Dong Du (@ddnirvana) profile picture

Assistant Professor at IPADS Lab, Shanghai Jiao Tong University. Working on OS, architecture, and cloud (serverless computing).

Joined April 2017
New SIGOPS Blog by Mingyu Li from Chinese Academy of Sciences (CAS), "A discussion on how to build a better ChinaSys workshop" sigops.org/2024/personal-vie…
3
8
1,224
Please always reach out to @ddnirvana or me if you'd like to writing on @ACMSIGOPS Blog, sigops.org/blog/ We publish blogposts very broadly as long as they offer insights, foster discussions, articulate viewpoints, etc., basically anything that could benefit computer system research, development, and operation.
There's actually plenty of incentive to write those blog posts. It just takes away from writing the N+1th paper, and so e people would rather do that (and also makes it easier to blame industry for not reading the papers).
2
1
27
2,843
#ILLINOIS CS professor Tianjin Xu and his brilliant team of graduate students developed a push button to aid in testing cloud systems! ▶️bit.ly/3x6Gw6C
1
4
2
39
6,362
Loved every bit of the epic articles written by @akatsarakis. Very much agree with the 10 Tips (and the wisdom!) for any PhD journey and love the poem in the last chapter. Apparently, Antonis had so much fun doing the research. I miss my time as a PhD student :) – at Urbana, IL
Part 2 of "Thinking outside the box: My PhD Odyssey From Single-Server Architecture to Distributed Datastores" is published, where @akatsarakis writes about Zeus [EuroSys'21] and Themis [EuroSys'23] and 10 Tips for PhD, with a rhymed summary Read here: sigops.org/2024/thinking-out…
1
2
20
4,207
[SIGOPS Blog] "Thinking outside the box: My PhD Odyssey From Single-Server Architecture to Distributed Datastores (Part 1)" by Antonis Katsarakis (@akatsarakis) who received the Honorable Mention of the 2023 Roger Needham PhD Award. Link to the article: sigops.org/2024/thinking-out…
1
11
2
33
9,507
Incredibly insightful article by @marius from @fb_engineering on Global Capacity Management for @Meta 's Massive Infra (millions of servers). Many great lessons and practices on balancing tradeoffs, taming complexity, and accommodating dynamics. Remind me of the old, good days. – at Urbana, IL
New SIGOPS Blog: The journey of real-life industry work behind an #OSDI paper — global capacity management for millions of servers by Marius Eriksen (@marius) from @Meta. sigops.org/2024/the-journey-…
4
26
4,137
It's been an incredibly impressive experience. Wishing all the best to our amazing system community and thanks for all the supports!!!
New SIGOPS Blog by the #SOSP 2023 AE Chairs, Dong Du (@ddnirvana) from SJTU and Jiayi Meng from UT Arlington: "Lessons Learned from Chairing the Artifact Evaluation at SOSP 2023" sigops.org/2024/lessons-lear…
1
45
Very interesting workshop!
Looking forward to receiving your paper submissions!
72
The ChinaSys (SIGOPS Chinese Chapter) has had their 25th Workshop a month ago. See a newsletter about the Keynotes, Rising Star Awards, and Dissertation Awards at the workshop, sigops.org/2024/the-25th-chi…
2
1
3
4,838
Very insightful and interesting talk! It is always important (and challenging!), as a university faculty, to rethink better way to help students understand the system software and architecture.
51
🌟PowerInfer boosts LLM serving speeds by up to 11x on consumer-grade GPUs! Inspired by our Deja Vu paper (ICML'23 arxiv.org/abs/2310.17157), it serves ReLUified LLMs, keeping heavy hitter (hot) neurons in GPU and offloading sparsely fired ones on CPUs. Proud of my undergraduate/master alma mater SJTU (Shanghai Jiao Tong University) for this work! github.com/SJTU-IPADS/PowerI…
3
22
3
120
14,739
PowerInfer can massively speed up inference on consumer GPUs. Almost reaching A100 levels. It outperforms llama.cpp by up to 11.69x while retaining model accuracy. PowerInfer reached an average token generation rate of 13.20 tokens/s, with a peak of 29.08 tokens/s, across various LLMs (including OPT-175B) on a single NVIDIA RTX 4090 GPU, only 18% lower than that achieved by a top-tier server-grade A100 GPU. You can use these models with PowerInfer today: - Falcon-40B - Llama2 family
15
77
6
348
41,002
"Perspect — Exploiting essential characteristics of performance issues for automatic performance diagnosis" sigops.org/2023/perspect-exp… New SIGOPS Blog article by Xiang (Jenny) Ren (@xiang_jenny_ren) from University of Toronto @eceuoft @UofT.
3
1
10
5,377
TEE systems tend to utilize third table (besides PT and NPT) to manage permissions, which brings new trade-off between scalability and efficiency. This work gives one solution: Accelerating Extra Dimensional Page Walks for Confidential Computing | MICRO. dl.acm.org/doi/10.1145/36134…
1
44
Simplicity is largely overlooked as a key success metric of computer system designs and is often misinterpreted as being easy. In the world of MLSys, @ThaleiaDimitra rephrases the KISS principle to be "keep it simple (and!) smart." Thanks @ddnirvana for editing the blog post. – at Urbana, IL
Thaleia Dimitra Doudali (@ThaleiaDimitra) writes about the new KISS principle of machine learning for computer systems in a new SIGOPS Blog article: "K.I.S.S.: Keep It Simple, Smart!" sigops.org/2023/k-i-s-s-keep…
4
18
2,365
Replying to @DmitriiUstiugov
Thanks, Dmitrii, for sharing this work! Open-source platforms also help in exploring and integrating new ideas (from both academia and indus).
1
1
83
Kubernetes exposes a robust API that lets you control every aspect of the cluster. Most of the time, it's hidden behind kubectl, but no one stops you from using it directly. Let's dive into it.
12
71
6
333
62,324
What happens when you create a Pod in Kubernetes? A surprisingly simple task reveals a complicated workflow that touches several components in the cluster. Let's dive into it.
14
153
9
574
73,503
Firecracker: Lightweight Virtualization - Opportunities and Challenges youtu.be/ADOfX2LiEns via @YouTube