All RAM utilized

Hello dear node operators,

I always seen very little usage of the RAM. Recently i moved some nodes to VM (Proxmox+ubuntu) and now i see that 4 nodes, allocate ±30GB of RAM (started with 2GB). I believe if i would give more, it could utilize more. Is there any changes in the code, or in the software? Will it benefit for the speed of nodes/data transfers? Should i buy even more of ram?

p.s. now i’ve looked to the other node, also ubuntu, but not in VM. There is 2GB of RAM, and it all used. Before it was using like ±300Mb :slight_smile: or so

Thank you.

I believe most of ram is used by Os as cache/buffer not storj node
“ free -m “ command give you real used ram

2 Likes

You can also use the command

docker stats
1 Like

So the answers are the following:


In the morning i see less used ram :slight_smile:
This VM is only for storj. No other processes running.

So my question is, will Storj performance drop if i reduce RAM for this VM to let’s say 2GB? Other question will storj performace increase if i put more ram on the another server where ubuntu runs only on 2gb ram

Under Linux increasing RAM usually helps disks to work faster.
However, 1GB of RAM usage by each node looks too much for me. Usually it starting to use more RAM, if the disks subsystem is slow, for example - RAID any kind, SMR or network filesystems like SMB/CIFS, NFS, etc. Or filesystems like exFAT or BTRFS, or NTFS under Linux.
By the way, the network filesystems are not supported and the node will stop working sooner or later. If you forced to use a network connection to your storage, you need to use iSCSI instead of network filesystems.

All drives are internal (SATA 6GB/s), with ext4 file format. Single node per hdd.

Linux storage nodes with slower random read/write consume more RAM when the traffic increases like these days. And, the read/write speed through VM is slower than the native one; the older VMs had only 1/5 - 1/10 performance compared to the native. I am not familiar with Proxmox, but you would be able to find many tips on the web to improve the file I/O.

The most impruvement I see when using more RAM is FileWalker running faster. With 1GB RAM, it took days, with 18GB it takes under 1h.
The other benefit: you can set filestore write buffer to 4MiB, without worrying you run out of memory, and reduce fragmentation on disk.
As earnings, ingress and egress, I didn’t saw any differences in almost an year, between nodes.
Same specs: ext4, linux, docker. The 1GB machine has FW disabled.
Maybe the 1GB system will die sooner, because the HDDs are more accesed. In short, more memory is better, if you can afford it.

2 Likes