Potential Bottleneck?

tep1981 · 28 Sep 2018 at 12:19

I've configured the following system which will be used to gather data and transfer to a NAS using a 10Gbe line, the main requirement is that it will handle two cameras flinging data at 850mb/s each (connected by a single PCIe frame grabber (uses an x1 PCIe interface).

I want to make sure there is no potential bottleneck as the most important feature is sustained read/write speeds to and from the optane drives

Intel Core i9 7900x
Asus WS X299 SAGE 10/G
Noctua NH-D15
64gb Corsair Vengeance 3600mhz
Corsair AX860
Gigabyte RTX 2080
Samsung 970 Pro 1TB nvme (windows)
2x Intel Optane 905p 960gb (both drives used a scratch disks then data will be uploaded via 10Gbe to NAS)

LuckyBenski · 29 Sep 2018 at 03:17

That's 10 terabytes a day assuming it runs continuously. 2.4 hours for 1TB. Can the NAS disks write that fast? 100MB/s per camera?

SSD write speed itself should be fine. I'd query the network's capability (not simply the basic speed tier), and the NAS disks. I'd also look into the transfer capabilities of the capture card.

tep1981 · 29 Sep 2018 at 10:36

The data is being uploaded to a Storinator XL45 which I believe is full up with 10TB Ironwolf drives, we have a 40Gb backplane so network transfer speeds will not be an issue

The cameras throw data up to 850mbps but not continuously

FloppyPoppy · 29 Sep 2018 at 11:03

Is this being used for anything else? i suspect it's actually considerably overkill for what your after.

tep1981 · 29 Sep 2018 at 20:09

Nope its purely for capturing images from a microscope at high speed, the datasets created can be as large as 900gb but as soon as they are created they are uploaded to the storinator

This was put together literally a couple days before they announced the 905p 1.5TB so I may change it to two of those

FloppyPoppy · 29 Sep 2018 at 20:35

I would do a little bit of research in that case on what the software for those microscopes uses when doing the data crunching on the system side. As it's likely not a GPU intensive workload and rather, a cpu one(maybe). So you may be able to save some money there on the GPU and possibly get a faster CPU but that is only if it's necessary. Honestly it's quite an unusual use case that i highly doubt the majority of people in here would have set up or have the knowledge of what hardware is important to have powerful with that workload.

What Microscopes are they by the way? Il have a look myself to see if i can dig up any information to help myself understand what i am looking at

.