I don't know how to run multiple drives with Spacemesh. I currently have three drives, one drive on its each computer. I am running the Spacemesh GUI program in Windows.
Could you point me in the right direction? I see you have a ton of videos.
1. Spacemesh in Windows will only allow you to run one drive, is that correct?
2. Do I need to get off of Windows in order to do a multiple drive system?
3. Are the videos of you detailing how to run Linux and Proxmox what it takes to run a setup like you have listed above? I am not familiar with Linux so that would be a little more complicated for me.
I am so confused as where to begin.
I am looking to run just a few hard drives I have that are 6tb and 8tb each. Like 4 total drives.
Filling up the hard drives with the plots in a timely manner seems like the trickiest part. If you've got 66TiB to fill, even with a 4090 it will take you 4 months. You'll miss out on a lot of epochs. What are your thoughts on that?
It's a struggle for most miners, so we are all working against that limitation. My plan is lots of small nodes, which is why I don't recommend RAID. For example, on one host I have 10x2TB HDD, and as I finish each HDD I get it registered and earning rewards. That way each epoch I have more and more space joining.
Yes, makes sense. So one approach could be to optimize for the size of node that you're able to generate in the span of each epoch (with a 4090 that would be like 6TB?) But then again, small nodes do carry advantages: they're easier to move around and allocate to what hardware you have as you see fit. Do you see any advantage really to having larger overall nodes compared to smaller ones? It's not like you get any sort of bonus for single large plots. (Just thinking out loud here, feel free to share your thoughts. And thank you for all the knowledge and tools you've put out there so far, I for one really and truly appreciate it!)
The main problem with lots of small nodes is it takes a lot of RAM. For each node you need to have ~5GiB. So for 10x nodes you need 50 GiB. Not a terrible limitation, but something to be aware of. Also with a lot of nodes you need to run public/private nodes, as the amount of connections and bandwidth usage can cause issues. Basically you designate 2 nodes to be public, and they get external peers. Then the rest are private and only talk to the public nodes.
This is going to become less of a problem once go-sm v1.3.0 comes out because it will allow for a dedicated machine to run the nodes, then you just run a post-service on the host with all the postdata. Initially this will be 1:1 where you still have the same limitations. But eventually it will be 1:N where you can have just a single node for 100s of postdata.
What is the best tool for benchmarking how many nodes I can run on my smeshing machine. I have a AMD 3900x with 2x12TB drives striped (raid 0). Epoch 9 I had 4 nodes (2 TiB each) that qualified for 5 layers. I finished plotting 2 more 2 TIB nodes and only 1 node of the 4 qualified for layers. (the 2 new ones won't be eligible until epoch 11).
I'm just looking for a way to benchmark my smh rigs to know what the limit of my setup is. The CrystalMark on my raid 1 drive has .42, so according to the this page I should be able to run 12TB of PoET data. I wish someone would do a youtube video on how to benchmark node hardware so we can maximize PoET data.
It usually not super trivial to benchmark, especially if you are running more than one node. Typically when you finish your postdata, it will generate a proof. If you time that you can get an idea of how long it will take in real life.
Hi Alex, could please tell if there is kind of trick in order to run the HBA.
I've purchased the exact same HBA and MB and already spend days trying to run this HBA, but with no luck.
I will be really thankful if you could give me some directions.
Thank you for this article.
I don't know how to run multiple drives with Spacemesh. I currently have three drives, one drive on its each computer. I am running the Spacemesh GUI program in Windows.
Could you point me in the right direction? I see you have a ton of videos.
1. Spacemesh in Windows will only allow you to run one drive, is that correct?
2. Do I need to get off of Windows in order to do a multiple drive system?
3. Are the videos of you detailing how to run Linux and Proxmox what it takes to run a setup like you have listed above? I am not familiar with Linux so that would be a little more complicated for me.
I am so confused as where to begin.
I am looking to run just a few hard drives I have that are 6tb and 8tb each. Like 4 total drives.
Filling up the hard drives with the plots in a timely manner seems like the trickiest part. If you've got 66TiB to fill, even with a 4090 it will take you 4 months. You'll miss out on a lot of epochs. What are your thoughts on that?
It's a struggle for most miners, so we are all working against that limitation. My plan is lots of small nodes, which is why I don't recommend RAID. For example, on one host I have 10x2TB HDD, and as I finish each HDD I get it registered and earning rewards. That way each epoch I have more and more space joining.
Yes, makes sense. So one approach could be to optimize for the size of node that you're able to generate in the span of each epoch (with a 4090 that would be like 6TB?) But then again, small nodes do carry advantages: they're easier to move around and allocate to what hardware you have as you see fit. Do you see any advantage really to having larger overall nodes compared to smaller ones? It's not like you get any sort of bonus for single large plots. (Just thinking out loud here, feel free to share your thoughts. And thank you for all the knowledge and tools you've put out there so far, I for one really and truly appreciate it!)
The main problem with lots of small nodes is it takes a lot of RAM. For each node you need to have ~5GiB. So for 10x nodes you need 50 GiB. Not a terrible limitation, but something to be aware of. Also with a lot of nodes you need to run public/private nodes, as the amount of connections and bandwidth usage can cause issues. Basically you designate 2 nodes to be public, and they get external peers. Then the rest are private and only talk to the public nodes.
This is going to become less of a problem once go-sm v1.3.0 comes out because it will allow for a dedicated machine to run the nodes, then you just run a post-service on the host with all the postdata. Initially this will be 1:1 where you still have the same limitations. But eventually it will be 1:N where you can have just a single node for 100s of postdata.
What is the best tool for benchmarking how many nodes I can run on my smeshing machine. I have a AMD 3900x with 2x12TB drives striped (raid 0). Epoch 9 I had 4 nodes (2 TiB each) that qualified for 5 layers. I finished plotting 2 more 2 TIB nodes and only 1 node of the 4 qualified for layers. (the 2 new ones won't be eligible until epoch 11).
You can use https://plan.smesh.online/ it's made by a trusted community member.
I'm just looking for a way to benchmark my smh rigs to know what the limit of my setup is. The CrystalMark on my raid 1 drive has .42, so according to the this page I should be able to run 12TB of PoET data. I wish someone would do a youtube video on how to benchmark node hardware so we can maximize PoET data.
It usually not super trivial to benchmark, especially if you are running more than one node. Typically when you finish your postdata, it will generate a proof. If you time that you can get an idea of how long it will take in real life.
On that page it talks about running " postcli -genproof", but what other params do you need to do to do a test on an existing PoST data set?
You could run -verify on it as well if you want. But not really needed.