Tim96

My bad, I should have posted this in general support. It's not hardware, I'm pretty sure this is just shfs overhead....Should have read one more forum post before making this and I would have figured out it out haha. This problem and solutions are outlined in this thread.

1st time posting, not sure if this is the proper section of the forum to post in, but I believe this is a hardware bottleneck of some kind, I'm just having trouble identifying where and what it is. Any help would be greatly appreciated! To make it short and simple, I have a 2.5gbps home network with an unmanaged switch. My unraid server along with the PCs (all Windows) have 2.5gbps NiCs in them. I do often reach the speed limits of around 300MB/s with large single file reads from the unraid server to my PC (NVME SSD to NVME SDD). The problems begin going from PC to the server, even with the same file transfer, I only hit 150-200MB/s (exact same SSD to SSD transfer just the other way around). Now despite being a little annoying/weird, I can live with it. The bigger problem I want to get sorted is transferring many smaller files (think hundreds of 10MB files) from the PC to the server. Again, this is NVME SSD to NVME SSD. Now don't get me wrong, I know that even fast nvme drives struggle writing small files once it runs out of cache. However, I end up with speeds anywhere from 10KB/s to a maximum of 5MB/s writing to the server. I know the drives I'm using are capable of much higher speeds than that, as the same file transfer is literally 10x faster going from the server to the PC. The array/parity (I don't think) are relevant to any of this, I'm simply transferring to an nvme ssd in a pool, no array activity is happening during this time. So to me the problem seems to be the write speed of the server drive is being bottlenecked by something, but I'm not sure what the reason for that actually is. So essentially my questions are: Is my 2.5Gbps network card in the PC not able to saturate the network completely? Is this some kind of SMB problem/bottleneck..? (the server and PC both have 8 and 12 core CPUs with 32 and 64GB of ram respectively) Is my switch not able to saturate it properly? (seems counterintuitive) Is btrfs holding back the write speeds of the nvme drive in my unraid server? Would I be better switching to something else? Is it something else that I'm completely missing here? Please let me know if there's any important info I didn't mention, any help would be massively appreciated ❤️