I’ve a single Storj (Windows GUI) node with 40TB. I’ve dedicated 10TB to Storj with the idea to increase in the future.
Dashoboard indicates that node is almost full but the disk occupation is over 11,5TB.
Is this related to the fact that I’ve used 16KB cluster size in NTFS to format a 40TB disk?
A disk so big cannot be formatted with default 4KB or 8KB cluster size in accordance to following table.
Cluster size. Largest volume and file
4 KB (default size). 16 TB
8 KB. 32 TB
16 KB 64 TB
Storj stores many <16 KB files. For example one 1KB file is stored in a 16KB block resulting in wasted space (1 KB of data and 15 wasted).
Now the questions:
Is better a big node or maybe 3 node with 4KB cluster size and max capacity of 16TB?
No, Windows is in a VM on a ESXi server host.
The server has a RAID-5 with 4x16TB.
I’ve created a virtual disk of 40TB for Storj data but due to dimension I had to use 16k block when formatting the disk in Windows.
It’s nice that I can have a such big node but it waste too much space for tiny files.
I know that RAID is non ideal for Storj but I’m experimenting! =)
Do you think it would be ideal to have 3 VMs/nodes on the same server with 16TB each?
In that case I think it would be better to have 1 disks for VM in passthrough without using a single big RAID.
Ideal is to do not use VMs at all. And not Windows. Also 1 disk/1 node and not all at once, but when the previous almost full or at least vetted.
However, you are using what you have now.
If you going to reconfigure, please do not use VM, the performance will be bad especially on NTFS and especially VMWare.
Use docker instead if the host is Linux. If the host is Windows then docker is not a best option, since it will be in the Linux VM too, but at least you may run several nodes there without much of issues or you may use Win GUI Storj Node Toolbox to run several Windows services instead.
You’re absolutely right.
Thanks for all the advice!
In a future not so far I will reconfigure the server to install Proxmox and test it.
Obviously single machine and single node will be always better.
This depends on available resources. If it’s not a Raspberry Pi 2/3, you likely may run several nodes (see Step 1. Understand Prerequisites - Storj Docs), but each node should use an own disk/pool, it’s very bad for performance (and also against ToS) to run multiple nodes on the one disk/pool (the pool (or likely a volume) can be considered as a one disk too, because it’s presented as a one disk to the system).
