Fire-Flyer File System (3FS)

by wenyuanyu • 1 year ago

361 101 1 year ago

Can someone convince me this isn't NIH syndrome? Why would you use this instead of SeaweedFS, Ceph, or MinIO?

It’s not. When you are a high frequency trader and you’ve mastered RDMA, everything around you looks slow. You are thinking in terms of 20 nanoseconds intervals, while everyone around still thinks that serving a query under a millisecond is fast.

rfoo6• 1 year ago

Huh? What kind of RDMA has a completion latency of 20 nanoseconds? It's more like 5 microseconds.

I agree that a lot of "modern" storage stack is way too slow though, tried to find a replication-first object storage for crazy-fast random read in small number of objects last year and found none.

ein0p2• 1 year ago

Seems like Ceph is considerably lower in throughput: https://ceph.io/en/news/blog/2024/ceph-a-journey-to-1tibps/ A serious concern when saving hundreds of terabytes of weights and optimizer states every now and again, or loading large precomputed prefix KV-caches. Minio seems to be slower still. IDK about SeaweedFS - they don't mention performance in their selling points at all.

do_not_redeem2• 1 year ago

It's quite funny that I got two opposite answers right away: you say it's to improve throughput, and sibling says it's to improve latency, and as we know throughput and latency trade off against each other. I'm inclined to agree it's more likely they're prioritizing throughput, since their readme charts throughput but not latency. But OTOH, the project looks like it requires RDMA. I wonder if the authors have written about their motivations and the tradeoffs they made, so we don't have to speculate.

EDIT: Their blog post answered all my questions and more. https://www.high-flyer.cn/blog/3fs/

ibotty3• 1 year ago

Look at the hardware first:

The hardware in the Ceph test is only capable of max 1.7TiB/s traffic (optimally without any overhead whatsoever).

I also assume that the batch size (block size) is different enough that this alone would make a big difference.

mgerdts• 1 year ago

The final aggregate read throughput reached approximately 6.6 TiB/s with background traffic from training jobs.

The Ceph team has been working on Crimson for years to get past performance bottlenecks inherent to the HDD-based design. I’m having troubles finding any ceph benchmark results that show any close to 100 GB/s.

pat2man• 1 year ago

Seems easy to find: https://ceph.io/en/news/blog/2024/ceph-a-journey-to-1tibps/

nivertech• 1 year ago

I'd argue that they don't need a filesystem or an object storage, they need a purpose-built data serving layer optimized for their usecase.

cttet• 1 year ago

If NIH syndrome boosts morale of the team, it should be helpful on overall team progress though.

jpgvm3• 1 year ago

None of those are close to fast enough.

The only competitors in the parallel FS space that are useful for this are Lustre and Weka.

Otherwise if you don't need a single namespace a bunch of fat AF NFSv4 servers w/NFS over RDMA will also get you to 6TiB/s.

The "surefire" way though is still Lustre, it's the big daddy of distributed parallel filesystems still but it's an absolute beast to setup and operate.

doomleika• 1 year ago

Software tech in China is a different landscape. It's really common to reinvent the wheel in China. Almost every big name (Bytedance, Meituan, etc) have their own of everything with both office political and in house need reasons.

The thing is those stuff are so prevalent those in house tech have reach the point they are competitive. This doubles for quant firm like DeepSeek.

whalesalad2• 1 year ago

Sometimes you must succumb to NIH. How do you think all those tools you mentioned got produced?

brcmthrowaway• 1 year ago

What does Anthropic use?

pepsi-not-coke• 1 year ago

I love it. AWS EFS costs too much. The open source solutions are clunky. I'm hoping DS applied their ingenuity to this one, too. Can't wait to trial it.

tetron• 1 year ago

Was curious how they get such performance with a FUSE based design. It seems that they sort of cheat, FUSE is used to manage metadata but to get high performance you have to link in the C++ client library and do all your reads and writes through that. So it isn't general purpose, you have to modify your application to take advantage of it. Still, that's a clever trick, and makes me wonder if there's a LD_PRELOAD strategy that could generalize.

grohan• 1 year ago

They appear to have Python bindings which seems reasonable from an API / usability perspective? https://github.com/deepseek-ai/smallpond

In terms of fast FUSE - also my first question, appears to be`io_uring` + FUSE :)

https://github.com/deepseek-ai/3FS/blob/main/src/lib/api/Usr...

amelius• 1 year ago

Why is FUSE that much slower than providing your own read/write functions? I get that it has to go through the kernel, but the operations are on entire blocks and network should be the bottleneck by far (and disk/main memory should be a bottleneck if the data is local).

vlovich1232• 1 year ago

You have to bounce through the kernel back out to use space. The number of syscalls is quite high. In many cases this is mitigated somewhat by the page cache making reads cheaper, but that’s explicitly an anti design here.

I believe there’s work to minimize this using io_uring so that you can talk to the fuse driver without the kernel being in the middle, but that work isn’t ready last time I checked.

For what it’s worth at Palm we had a similar problem because our applications were stored compressed but exposed through fuse uncompressed, instead of O_DIRECT I just did an fadvise to dump the cache after a read. Not as high throughput but the least risky change to get the same effect.

ammo1662• 1 year ago

For those who are interested, the design was originally published here:

(Chinese) https://www.high-flyer.cn/blog/3fs/

This file system has been developed and utilized by them for several years .

Compared to the traditional file systems, it is more focused on model training that contains a lot of random reads. Read cache and prefetching are useless in this case. Therefore, they designed the file system without those features to improve the performance.

I google translated some key parts here:

3FS is a special file system because it is almost only used in the scenario of batch reading sample data in the computing node during AI training, and accelerates model training through high-speed computing and storage interaction. This is a large-scale random reading task, and the read data will not be used again in a short time, so we cannot use the most important tool "read cache" to optimize file reading, and even advance reading is useless. Therefore, the implementation of 3FS is also quite different from other file systems.

Specifically, as shown in the figure above, 3FS uses the Linux-based AIO and io_uring interfaces to complete sample reading, because in the 3FS scenario, File Cache has no effect at all, but will consume system memory in a way that is difficult for users to control, affecting the operation of subsequent tasks, so we turned off File Cache and only used Direct I/O mode to read data. But it should be noted that when reading in this way, the buffer pointer, offset and length all need to be aligned. If the user is allowed to do this alignment, additional memory copies will be generated, so we have done the alignment inside the file system, which not only optimizes performance but also facilitates users.

rvba• 1 year ago

Why is that a random read? Also is it truely random, or from seed? But if prng then they could cache right?

kilburn• 1 year ago

Random is prng. They still cannot cache though because they do many reading "passes" through the same data.

If build a cache that gets hits on the first pass, then it won't work for the second and later passes.

vlovich1232• 1 year ago

I hope they chose a multiple of 4096 for the alignment to minimize flash read amplification. QLC drives even use 16kib pages.

dekhn5• 1 year ago

How critical is random reading of training data when assembling batches?

Put another way: in my experience, supporting fast random reads is a challenging problem, while supporting high sequential reads is fairly straightforward. When is random access to a training set absolutely necessary for training a model?

c4wrd• 1 year ago

Imagine you're studying for a test where you are given an image and need to answer the correct class. To prepare, you're given a deck of flashcards with an image on the front and the class on the back.

(Random) You shuffle the deck every time you go through it. You're forced to learn the images and their classifications without relying on any specific sequence, as the data has no signal from sequence order.

(Fixed order) Every time you go through the deck, the images appear in the exact same order. Over time you may start to unconsciously memorize the sequence of flashcards, rather than the actual classification of each image.

When it comes to actually training a model, if the batches are sampled sequentially from a dataset, it risks learning from correlations caused by the sequencing of the data, resulting in poor generalization. In contrast, when you sample the batches randomly, the model is biased and encouraged to learn features from the data itself rather than from any signals that arise from artifacts of the ordering.

arjvik• 1 year ago

On an SSD, random and sequential reads have nearly the exact same performance. Even on large arrays of spinning rust this is essentially true.

budududuroiu• 1 year ago

Does anyone know if there’s a benefit to porting this to an orchestrator like K8s, maybe overkill for training but the KVCache might be useful when having multiple replicas for inference?

jeffbee2• 1 year ago

Interesting that their GraySort result is CPU bound while they are using 3x more CPUs than the record holder from ten years ago.

sitkack2• 1 year ago

How can you determine that it CPU bound from the attached charts?

jeffbee2• 1 year ago

Because it hits a read peak in the first wave and never hits it again.

bee_rider5• 1 year ago

They sure are productive.

What are we going to see tomorrow? DeepSeek OS or something?

logicallee2• 1 year ago

They sure are productive.

I have a theory as to why...

tuyguntn• 1 year ago

enlighten us

digdugdirk• 1 year ago

996 work culture?

vitaflo• 1 year ago

To be fair they’ve been working on this since 2019 for HFT. So it’s not like they just whipped this up.

bee_rider5• 1 year ago

The amount of brainpower wasted on HFT games that will never see the light of day is kind of a bummer. Congrats to China.

thohj4234234324• 1 year ago

This is very humbling.

OpenAI et. al kind of have also been very deep down the systems rabbit hole (eg. Triton), but I can't think of anyone else (outside of Google/Facebook) who pay this amount to attention to things.

Great work; hope Deepseek does even more awesome things going forward.

richardw• 1 year ago

I’ve assumed that it’s partly because the company has done a lot of HFT, which is very focused on performance. But I’m not an expert in either.

WiSaGaN• 1 year ago

Indeed, the blog mentioned in the other comment showed part of 3FS code was completed at least since 2019, when this was still a project of the quant funds. In HFT, you tend to dogfood a lot of the things to achieve low latency, high performance, sometimes just because HFT system just need to do one specific thing, and those off the shelf stuff usually cater for a lot wider scenarios where HFT doesn't really care about. Here you see similar case which they focus specifically on loading large amount of data during training, and implement that to the extreme.

jauntywundrkind• 1 year ago

Man, 6.6TB/s across 180 nodes is 300Gbps/node, or 37.5GBps.

That's with 14 unnamed SSD per node. I wonder how this would scale to higher end SSD,dealing from PCIe 4 to PCIe 5 or PCIe 6... Particularly whether one could scale down!

WithinReason• 1 year ago

Why is this even necessary? Can you just shard your training set to the training nodes ahead of time instead?

wenyuanyu• 1 year ago

No, besides accessing training data, there is also logging and checkpointing... When you run k8s over it, and there are multiple training jobs... isolated local storage is a nightmare...

yalogin• 1 year ago

It’s not clear to me where and how the current popular systems fall short. Do they talk about I anywhere?

Also, what specifically is the data access patterns for training and inference that are different from traditional use cases?

jpgvm3• 1 year ago

Well current popular systems are pretty much limited to Lustre and the new kid Weka, mostly Lustre though tbh.

You can try to use "standard" options like MinIO/Ceph(RADOS)/SeaweedFS but you will very quickly learn those systems aren't remotely fast enough for these usecases.

AI training is what this is used for, not inference (which has absolutely no need for any filesystem at all). What makes the workload somewhat special is that it's entirely random read and not cacheable at all as most reads are one and done.

Would Lustre be perfectly fine at 6TiB/s? Yes. Is it a huge pain in the ass to operate and make remotely highly available? Also yes. If this thing is capable of the throughput but easier to operate and generally more modern and less baroque it's probably an improvement. TLDR is Lustre is fast but that is literally it's only redeeming quality. I have lost far too many hours of my life to the Lustre gods.

cyanf• 1 year ago

They’re using the FS for caching the KV caches of past requests. It’s why they’re able to charge so little on prompt cache hit.

rfoo6• 1 year ago

What makes the workload somewhat special is

I'll add that latency also doesn't matter that much. You are doing batched data loading for batch n+1 on CPU when GPUs are churning batch n-1 and copying batch n from host memory at the same time.

So as long as your "load next batch" doesn't run for like >1s it would be fine. But one single "load next batch" on one worker means thousands (if not more) random read.

pella• 1 year ago

related research paper (english - html ) - https://arxiv.org/html/2408.14158v2

arXiv:2408.14158v2 [cs.DC] 31 Aug 2024

"Fire-Flyer AI-HPC: A Cost-Effective Software-Hardware Co-Design for Deep Learning"

Abstract:

"The rapid progress in Deep Learning (DL) and Large Language Models (LLMs) has exponentially increased demands of computational power and bandwidth. This, combined with the high costs of faster computing chips and interconnects, has significantly inflated High Performance Computing (HPC) construction costs. To address these challenges, we introduce the Fire-Flyer AI-HPC architecture, a synergistic hardware-software co-design framework and its best practices. For DL training, we deployed the Fire-Flyer 2 with 10,000 PCIe A100 GPUs, achieved performance approximating the DGX-A100 while reducing costs by half and energy consumption by 40%. We specifically engineered HFReduce to accelerate allreduce communication and implemented numerous measures to keep our Computation-Storage Integrated Network congestion-free. Through our software stack, including HaiScale, 3FS, and HAI-Platform, we achieved substantial scalability by overlapping computation and communication. Our system-oriented experience from DL training provides valuable insights to drive future advancements in AI-HPC."

codingwagie• 1 year ago

I think the difference between deepseek and OpenAI/Anthropic is one of the difference between practitioners and academics. Ofcourse there is world class talent at OpenAI. But there are also alot of "I went to Harvard and want to work in AI", and those types of people just simply dont have the technical exposure to even think of building something like this.

mustpax• 1 year ago

Someone should write a blog post about the prestige/effectiveness negative feedback loop. This is also the Achilles heel of top tier SV VCs including YC.

dvaun• 1 year ago

Can you expand on this?

djtango• 1 year ago

You mean this one? https://news.ycombinator.com/item?id=9125816

bugglebeetle• 1 year ago

The problem isn’t the prestige it’s that prestigious institutions in America don’t produce high-quality talent. They’re instead mostly corrupt credentialing mills for the rich and well-connected. From what I understand, DeepSeek also only hires from the best universities in China, but “best” actually means something relative to how difficult entrance to those organizations is to achieve and their coursework.

tway2232• 1 year ago

I would say most if not every large company in China has their own AI infra stack, partially because tech talent is relatively more abundant and partially some of the tech leads have been exposed to western tech via open source and work experience so they have a good success rate (which makes it a more common practice). Anecdotally, specifically Google, FB ex-employees from oversea offices, MSFT and Intel ex-employees from their China offices could be the key elements for this trend in the past two decades (Google left China around 2010).

The infra work is usually technically tedious so I think it may become some lost art in the west just like those manufacturing jobs.

smallmancontrov• 1 year ago

As opposed to the US, where every large company has its own AI infra stack, often extending down to the silicon and up to large open source projects?

What's going on here, why are people forgetting what's around them? Does familiarity breed contempt? Are attention spans so shot that failure to participate in this week's news cycle is enough for "out of sight, out of mind"? Or is HN full of Chinese bots now?

robotnikman• 1 year ago

Makes me wonder where is the best place to learn how to put together and operate something like this then? Certainly there should be resources out there somewhere to teach yourself?

cma• 1 year ago

Weren't the flash attention authors not just from academia but in academia at the time?

sureglymop• 1 year ago

I think it would be a bit irrational to claim that so broadly. I've met some incredibly talented people in academia and I've also met people that made me question how they even pass.

My hypothesis is that there is not such a big difference at all. All three of the companies you mentioned are world class competitors in this. DeepSeek were the last to have a "hit" but that isn't an indication that they'll be the next of the three (or other yet unknown entities) to have the next hit. We try to predict what happens next now but perhaps we should rather focus on who or what we want to succeed. For me it's quite clear: it should be open source or I'm long term not that interested.

whalesalad2• 1 year ago

The throughput on those charts is pretty wild - multiple terabytes per second.

hintymad3• 1 year ago

A distributed file system is honed as one of the trickiest software to write, and we are usually advised not to write a file system from scratch (even on top of FUSE), let alone a highly optimized one. When a silicon value company is having the 100th meeting to align god-knows-what, a team of fewer than 60 already came up with a production-grade highly efficient parallel file system.

Have we in the valley companies lost touch?

htrp2• 1 year ago

team of fewer than 10

the highflyer team are pretty well resourced.... think they have more than 10 people

hintymad3• 1 year ago

Thanks! Updated to 60 per their author list in their paper.

ycui19862• 1 year ago

yes

rvz• 1 year ago

Once again, DeepSeek continues with another home run.

Can't wait to see what they release next. DeepSeek should be studied carefully.

Fire-Flyer File System (3FS)

Related Stories