Zenoh – Zero Overhead Pub/Sub, Store/Query and Compute | Hacker News Reader

Query and Compute

zenoh.io

68 pointsshandor18 comments

> What is zenoh?

This page describes various "why?" motivations of Zenoh, but at the end of it I was none the wiser about what it is.

Something to do with pub/sub. It mentions Arduino so I guess it's targeted at IoT applications? It mentions MQTT so I guess it's a competitor?

(...to be fair the front page http://zenoh.io/ has more relevant info)

Having read http://zenoh.io/docs/overview/#unified-abstractions and being pretty familiar with the IoT space, I also don't yet understand what it's for or why you'd use it instead of X, Y, and/or Z. (I'd like to, though!)

I also found it hard to discern the architecture

There are storage backends, there is a router component...

the examples at http://zenoh.io/docs/getting-started/first-app/ show curl-ing to the router(?), and it's presumably the router that talks to storage backend... so Zenoh is the router, and that runs on the remote device?

https://github.com/eclipse-zenoh/zenoh-python/tree/master/ex...

Alternative to the REST API it appears there is also some sort of peer network, and it mentions Zenoh peers and routers as separate entities?

The docs would benefit from some diagrams and definition of terms

> zenoh /zeno/ unifies data in motion, data in-use, data at rest and computations. It carefully blends traditional pub/sub with geo-distributed storages, queries and computations, while retaining a level of time and space efficiency that is well beyond any of the mainstream stacks.

I wonder what goes into the minds of developers when they use grandiose, unbacked descriptions like this. It is just missing the word "modern" :-)

Wouldn't it be better to have some numbers to compare to those others "mainstream stacks"? It would help to show not only how "fast" zenoh is but also actually what is it (still not sure ...), by comparing to other established players.

I see many comments here unsure about what Zenoh is. I'm probably more clueless than the average HN reader, but here is what I found out regarding the company invested into it.

Most of the Cargo.toml files contain:

  # Copyright (c) 2017, 2020 ADLINK Technology Inc.

And the CONTRIBUTORS.md file lists all ADLINK employees: https://github.com/eclipse-zenoh/zenoh/blob/9737c6b59900144a...

The extraordinary claims on the Zenoh site are matched by the wide claims made on the ADLINK Technology, Inc. site. Poke around and you will find they have a big global presence: https://www.adlinktech.com/en/index

It must be quite exciting to work all the time on buzzwordy things like edge computing, AI, & 5G, in industries like aerospace, defense, & manufacturing. It would seem that the things Zenoh strives to do are at least in-line with the types of stuff the company is involved in, however grandiose it all comes off as.

This sounded very intriguing, but I have to agree with the other comments that it is not very clear what the actual focus is.

It is at the same time confusing and frankly quite unbelievable that zenoh seems to target everything from couple-hundred-kb systems to scaling well in the cloud, and still being best at everything. I mean ok, that's rather awesome if even half of it is achieved, but I would definitely like someone more knowledgeable about the project to expand on their trade-offs etc.

The example makes it seem like another key value store. Any example for the Pubsub / compute piece would be good to see.

I am working with ADLINK under the 5GDIVE project. We are currently integrating it for our use cases, as we plan it to use as our DASS inside the architecture. For more context: https://5g-dive.eu/ We are the only partner using Java instead of Python so we unfortunately haven't yet tested it, as there are only bindings available for python. But they plan to add it soon in Java. Pretty competent company as of my opinion while dealing with them, especially with the creators of Zenoh.

The copy on the landing page irritates my product side. Lack of clarity, super focused on features without relatable context and no direct mention of what it does for me as a user.

I wish a tool like this would publish a usage guide that suggests some of the optimal architectures for using it. I read over the series of initial docs and now I understand what the basic abstractions are, but I’m still at a loss for what important problems I should use these abstractions to solve. Maybe because it’s not obvious, this product isn’t for me?

Very cool tech. The 8000 default port though, while cute, sounds like it will clash commonly with test setups.

what should the port be

Ideally, something that does not clash with extremely common default ports.

ah yes. something that does not clash with extremely common default ports.

Dynamically allocated using portmap(8).

looks interesting. is there any place to read about its design goals or how it is being used?

It is so painful reading through all of their media. They never once describe the system. They use all of the hated buzzwords to describe the design goals, what's problems it solves, but never actually talk about the system.

theres actually a few pretty clever things in here—- the uri scheme stuff, the library loading stuff, and, especially(!), the timestamping / logical clock bit stuff come to mind. neat ideas and project.

Reply on news.ycombinator.com