Optimal node scale

Hello all!

There is unused ~300+TB storage and two of IBM BladeCenter E with 14 HS 21/22 blades.

Also Iam owning AS+PI and have multiple 10GE ISP-grade interconnects ~20+ Gbit of unused (wasted) upload (node-to-client) bandwidth. Also I can plan QoS to obtain at last 5Gbit of download (client-to-node) bandwidth with no additional cost.

Question: what is optimal node size ? 10TB? 100TB? 300TB?

Each node will utilize dedicated blade, separate IP address, 2x1GE IP up-link and 2x2GB FC storage connection.

Welcome to the forum @ntil!


Looks like 24 TB is max LUN size for each node.

There is actually no maximum, but you shouldn’t expect the kind of space you’re talking about to fill up any time soon. The idea behind Storj is to be distributed across independent nodes with relatively low individual system requirements. Your kind of hardware seems to be overkill for Storj.

I’ve been running a 10TB node since the beginning and at the moment only 5TB is being used. So that should give you some indication on what kind of data usage to expect. Keep in mind that it is early days though, usage may go up.

If you can get all nodes at size 10TB to spin up on a separate public IP, that might work. But for this size of setup I recommend contacting support at support.storj.io for advise.

and one more question - what outgoing monthly data for this node ?

Old hardware will not be beneficial on the StorJ network as the price is so low. The best is actually to run stuff on PIs.

I’m running 32 Gbit/s FC SAN at home with 24 drives in two enclosures, ~150TB, and have multiple ISPs w BGP but that does not mean shit for the network.

It only cares about a few things:

  • That my node is close enough to the clients,
  • The speed of my response. The faster my node responds, the better
  • Reputation, no downtime please.

I’m running my stuff in my closet, and have a quite complex setup, meaning its harder for me to maintain my setup compared to a Pi. My setup is now virtualised and have multiple layers, but I doubt its actually profitable based on the amount of time required to run such setup.

1 Like

Pi’s are great if you want to buy dedicated hardware for it. But I’d argue that ideally you’d use hardware that is already set up and running. This makes home servers or NAS devices ideal hardware for running a node. There is often spare room on those to begin with and you’re not running any additional hardware so the power use is only the slightly higher use because your disks would be more active.

But I kind of like that the setups out there go from the tiniest simplistic Pi setups to data center grade overkill setups. Everyone can find a setup they’re happy with. And for a lot of us this is more hobby that profit anyway. That’s not to say you can’t make money, you definitely can. But some people would enjoy spending much more on overkill hardware that’s not really necessary just because they enjoy tinkering with that stuff.