Combining SNO with Satellite Node Operator

Hello everyone!

Recently joined the StorJ network and I have been reading up on topics other than being a good node operator and I stumbled on the topic of nodes also running satellites. Little is known about it, at least I could not find a lot of information on the topic. I understand that is SoonTM but I was hoping if someone has some more information. Is there a Beta coming that I could sign up for? What would be the hardware and connection requirements(no doubt those would be high). Any information would be greatly appreciated.


Hello @HisEvilness,
Welcome to the forum!

You need to have a great hardware to offer a SLA to Customers and you should pay to storagenode operators, receive payments from customers, search for customers, incitive SNO to whitelist your satellite.

At the moment only Tardigrade satellites are whitelisted by default.

We do not have any documentation regarding joining the Tardigrade or run a Community satellite.
But it is planned in the future when the network will be more mature.

Hello and thank you for the warm welcome.

Could you give me a ballpark idea on what the specs would be? 4 x Tesla with a 64 core CPU and 256GB of RAM? And yes i understand to become whitelisted is a process. Currently undergoing the 1 month vetting process so i can understand i am not on satellite levels yet.


Not sure what you’re planning to use those 4 Tesla’s for, unless you mean a fleet of the kinds of Tesla’s you can use to drive to your several server locations for the HA setup you would need with real time replication and automated fail over. Because if you’re planning to provide a reliable satellite, that’s the kind of setup you should be thinking about. Either that or an extensive cloud infrastructure.

Hello @BrightSilence

All options are on the table I am exploring and looking to gain more information and knowledge on the subject matter. But to formulate some sort of plan I would need specs for the server, what demands Tardigrade would have. And what is expected of a satellite operator as a whole who would loan out hardware to Tardigrade.


Alright, if that didn’t scare you off yet, I probably am not the person to give you more details than that on specific specs. But I can point you to a good place to start doing your own investigation into this.

That describes how to set up your own test network. Which would help you get an idea of the kind of scale you’re talking about.

I’ll leave comments on actual specs to the experts. But you should know the Tardigrade implementation isn’t hosted on a single machine but rather as a modular set of cloud services.

1 Like

Awesome thanks for the info this is really helpful.

If community satellites offered different pricing this might be a good thing and a bad thing.

We will see. It’s open source and there should be an alternative from the Community.

Taking my time to read the material and see if this is something interesting. Would be interesting to get SkyNet going on some good hardware.

Tesla’s not only a car company, but also nvidia’s top tier multi user "gpu"processing cards, if it isn’t their flagship, but most likely not any more… its fast and it’s expensive…
so i will strongly assume he is talking about :smiley:

they can be used for a wide range of tasks like most gpu’s only difference is that these are build and has drivers that will allow multiple vm’s or such tasks to run in parallel, thus enabling a server to host graphic accelerated vm’s exceeding atleast 4 vm’s pr card and with normal use cases and users on the vm’s maybe even much higher numbers…
i believe they are like 2 or 4 times faster or more ram / bandwidth than nvidia 2080

cool stuff… been wanting to get something similar, but yeah not really that useful for storj…
tho in his defence running satellite tasks could very well be accelerated greatly through tensor cores due to them running i think 16 bit instruction sets making it possible to run many more threads parallel processing tasks on less silicon, just like the old sun servers also ran 16bit instruction set.
instead of the regular 32 - 64 due to the simplicity and thus better management of super high user numbers…

:smiley: so yeah his statement of his gear actually fits what i would expect a satellite to benefit from, if the software supports it ofc

I’m well aware. I was just making a joke because I don’t think there is much about a satellite that would benefit from GPU acceleration.

ha ha
well then i think you would be wrong. but technically it’s not because of the “GPU” acceleration but because of the 16bit instruction sets of the tensor cores that is a game changer for parallel processing.
ofc 16bit will not do anything fancy, just a lot and fast…

Yes, the idea behind that is to have GPU acceleration to handle files etc. Basically the hashing power of a GPU just as you would with other mining operations. In this case, getting the files from and to a client with encryption/decryption. And yes @SGC those are the GPU variants, not the carmaker.

Which doesn’t happen on the satellite.

Yes GPU’s can be good at very specific highly parallelizable workloads. The kinds of stuff the satellite does doesn’t seem to be very easily done like that.

I could see a case for a GPU accelerated Database though- figure not only does Sat node A need to make sure it can keep track of ingress and egress requests, but also needs to make sure that Sat node B through Z are in sync with it as well. Highly parallel transaction handling with row based locking seems like a nice fit overall.

1 Like

With the exception of repair, the satellite really doesn’t do much compute intensive stuff. Almost everything else is likely to be network and storage IO constrained. You’d still need a good CPU to deal with the many connections, but I don’t think it’s the kind of stuff that would benefit from GPU compute.

1 Like

Logically thinking, why does there have to be a satellite? StorJ have talked about wanting to make this more decentralized in the future so moving into that direction it does makes sense that SNO’s provide that decentralized service. Certain components that are handled by Satellite would have to be split and parts made decentralized so that SNO’s could be utilized. I see this happening in stages rather than overnight. Part of this decentralized improvement design SNO’s should be compensated for the additional work.


If the SNO’s were to start to handle some of the Sat things (piece delegation, repair setup, SNO management) would not things like client managed, direct to SNO, piece lookup and also distributed metadata (which I believe is already done) need to be done almost immediately? There’s also things like trustless authentication of pieces that would need to be done- basically who audits who if there isn’t a Sat to do it, and how do you establish who’s the authority on it without going back to a client local source file?

1 Like

The satellite is the only one that knows where all the pieces of a file are. Also checks on them ocasionally (AUDIT) to make sure there are enough pieces to restore the file. Also if pieces are lost it will calculate REPAIR pieces.

1 Like