So i finally upgraded my home network by accident. I never thought the seller on eBay would sell this switch for so cheap! so now I'm running 10 Gigabit and multi-Gig!
Totally agree! Plus the next evolutionary step in Wi-Fi access points is 2.5GbE. I wonder if a PoE injector that's Gigabit will allow more than a single gig through if the switch, access point and cabling can handle it because both speeds use the same amount of wires so it's not like a 10/100 Poe injector that doesn't have all 8 wires in it.
So, unfortunately, it's not your cheap Intel NVMe disk or your RAID 5 array that's causing the speed issue. If you try Atto Disk Benchmark against both drives you'll find pretty average performance with small block transfers and better performance with large block transfers. I've had 10Gb Ethernet in my home lab for years and just did some tweaks that improved some transfers, but others, particularly from/to Windows disks remain variable, minute to minute. There is something I am honestly missing, or something missing from my implementation. I am using an Arista Networks 10Gb switch and testing against NVMe, 10-drive NAS with SSD (Samsung) caching, and other endpoints. What I've tweaked recently, which helped throughput, at least in Atto, was have only Tx flow control enabled, increase my Receive buffers to the max of 4096, and increase my Transmit buffers to their max of 8184. (I have an Aquantia AQtion 10Gbit NIC on the motherboard of this workstation. Most of the other 10G NICs are Intel. Everything is copper. There may be other parameters that are worth playing around with on the NIC configuration. What seems to be happening is some sort of overflow, of packets or something else, that forces the transmission to slow down until what has already been sent/received is processed. Any help debugging this would be much appreciated. It would be nice to get 10Gb speeds on something other than a synthetic benchmark.
I plan on tuning it in a bit but even when I copy disk to itself with nothing over the network it does slow down. I'll try some synthetic benchmarks and set the file size to like 20-30GB or something huge and let run overnight next time I login. I'll also tune the NIC's and release another video. I do get the same performance with jumbo frames set to 9k on the machines and switch. Stay tuned for my update and I'll show what I had to do.
@Red Sanders I totally appreciate this criticism and would appreciate if you would like to let me know how I can do it better and I might even re-release this video with the edits that you suggest. I also have chapters on most of my videos to skip around. Please let me know how this could be edited better for your enjoyment, thanks!👍
I eventually went NVMe to RAM drive for benchmarks. I had to pull some ram out of my old server and put it in the new one just to be able to make a ram Drive big enough without buying any extra ram 🤣🤣🤣🤣
@Farrenz great question but what all nodes would you have connected and at what speeds? At a minimum you'll need CAT6A and network cards that support speeds over gigabit and a switch that supports at least multi-gig. How many devices? What are the devices? How long are the lengths of the cable runs?
@@ShinyTechThings I was thinking of just connecting my computer, and eventually sometime a router for my family. Right now, I’m looking at a mikrotik switch, the pcie card, cat6A cables, and that little sfp+ thing for my Ethernet into my pc
@Farrenz the SFP+ to Ethernet (GBIC adapter) might only be gigabit speeds so you might want to get a SFP+ card and a DAC cable if it's a short distance from your PC to the switch. The issue is then it would connect to the router and be bottlenecked from there. If you only plan on having say 2 devices at 10Gig then you could just connect those 2 nodes together from SFP+ to SFP+ cards and a DAC cable and not need the switch. Just have as a separate network defined for the 10 gig interfaces so if your home is say 192.168.1.1-254 then define the other 2 interfaces on 10.97.0.7-8 without a gateway defined and say a 255.255.255.0 subnet. So if your machine is .7 you could ping .8 or vice versa from those 2 nodes but their 1 gig interfaces would still be on 192.168.1.x and use the route to the internet over that interface. Let me know if that helps.
are you still happy with the XS512EM? I am looking for this switch as well, but I am not sure, as I did read some bad reviews, regarding poweroutages and that you have to unplug all, wait 5 minutes and plug all in again. Your answer is much appreciated.
I only had one outage when we were out from the house and the breaker tripped from my additional A/C, fans and servers on full throttle for my live stream. Now it's colder out so I don't need to run the A/C anymore but I didn't experience any issues when I turned the breaker back on, however the power was off for probably an hour after my battery backups ran out of juice. So are you saying if I yank the power cord out and plug back in then it could have issues running again without waiting 5+ mins? If so I can give it a try and see and make a video on it for you 🤷♂️. It could also be affecting certain firmware versions as well. Let me know your thoughts on this and I'll see what I can do.
@@ShinyTechThings I am doubting a bit between this Switch and the Mikrotik CRS309 The last one needs extra hardware for connecting to RJ45, but is completely silent. And I need the Switch on my desk Yes, that about yanking out the powersupply is probably the same as an poweroutage. And when the breaker went in, the switch did not work, until all plugged out and waited 5 minutes. Thank you so far for your input. It helps in forming an opinion.
@@aarbbee I want to reiterate something real quick is that after the power outage in my house from the breaker getting tripped once I got the power back on I did not have to do anything on the switch it just worked and came right back up. What I don't know is that if this is because it was over the course of hours that the power was off or if it is because I am running the latest firmware.
Home lab "cheap" 12 port 10G managed switch at $1,000 USD. Sure seems like my cheap home lab is more on the ghetto side. At $1,000 you can make your own. Intel E3-1270v2, Asus P8P67 WS Revolutiin with (2x PCIe 2.0 x16, 2x PCIe 2 0 x8, and 3x PCIe 2.0 x1) can give you 8x 10G connections, 6x 2.5G ports, and dual 1G all for about $350.
@Voodoo Vinny I didn't pay a grand, I paid like $450ish I just didn't expect the seller to accept my offer when they had it for closer to a grand. That's true and maybe a good upcoming video since I have some spare servers why not build a switch?🤓😎 What cards would you be putting in at that server?
@@ShinyTechThings cool I will check it out. I just bought a Dell PowerEdge R710. I'm thinking on how I can put it somewhere neatly. So far I love it. Only thing is I may replace the Seagate drives with Western Digital Red NAS drives.
I have a friend who works in data storage and they refuse to use seagate due to too many RMA's, however I work at a Dell server only shop and Dell uses the cheapest drives they can buy in bulk and it's a mix. I will say that the failure rates seem to be similar across the board but realize it could be those models when released could be more reliable than the current offerings from those manufacturers. I will say for new expensive SAS drives the Hitachi Helium ones draw less power than the competition. I'm not sure if I can find my notes for the video I started to shoot on a Synology SA3200D I built up but I was supposed on how low the power usage was compared to Synology's official power usage in their documentation. (That video will be a while before I finish it)
Eventually yes, but it's a home lab server so it's not at the top of my list. The 10Gbe and multigig was more for faster throughout to my NAS and will do a video on that so stay tuned!
@humanflycb I run an EnGenius cloud managed switch with 2.5GbE PoE, and SFP+ 10 gig ports but I also got a 10 gig Ubiquiti switch as well to expand my 10 gig ports as I have a handful of servers in my rack, plus my desktop, M1 Max MBP, and 2 10 gig NAS's. My favorite switch though is the other Netgear I reviewed that I got to borrow from a friend and it has 5GbE PoE ports and all multigig. If I ever find one for cheap I'll switch to that over the EnGenius one although the cloud controlled switch will alert me if my house goes offline which has happened before. Otherwise Ubiquiti UniFi is great and easy to manage. The older 10 gig switches are very affordable used but long term firmware support may or may not be there but they did update in Q4 of 2023 so I snagged it. What does your current network look like? How fast do you want to go and what's your budget?
Thanks but since I'm a workaholic I'm pretty sure I'll always be working full time anyways but I always try to do what I can to make some good content. Any suggestions besides a faster NVMe SSD to test from? I have one in my old server but need to pull some data from it before I can pull the SSD and some other components a d then eBay it. It's a E5-16XX V3 I believe. I don't use all it's power anymore and heat is my enemy in AZ so went smaller E3-1220V6 in my R330.
Yeah you might have got the switch for cheap, but you paid way to much for fake "Cat7" cables. Even Amazon is getting in on the Cat7 scam - All they did was take Cat6A cables and shield them and repackaged them as "Cat7".
I want a fluke tester that can do 10Gbe RJ45, that would be handy. But it works for now, next to do some Wireshark analysis to see how good the cable actually is.
Umm not very cheap LOL If they were cheap I would get 2 of them so I can build a cheap 10 GB network between my home and out side office. and build a 10gb back bone with 1gb feeding computers and 10gb for all servers.
@McCuneWindandSolar MSRP it's not cheap for home but for a business that wants multigig speeds it's very affordable. I would NOT consider this enterprise grade yet as I have had to reboot it twice since I've owned it to fix a port that stopped working. I do have the latest firmware on it but IMO for it to be enterprise quality to never need a reboot to fix a issue like a enterprise Cisco switch would then it would be enterprise ready. So if a reboot would kill your business and cost thousands to reboot during the day, buy a Cisco for $5K+ and the support contract to go with it for updates, but for home or a SMB that a 3 minute outage won't be too painful it's a decent switch. Again I bought for less than half MSRP on eBay but for home I would not pay full price. For a small office IF you NEED the speed it's not a bad option and is very inexpensive for the feature set costing less than the competition.
Yeah, I made an offer for half their asking price for the switch and they accepted it 😮. Then I was like shoot! Now I need to Craigslist some stuff and float the $ but it all worked out in the end except my cheap Intel drive but I have a few other tests to do with different variables and make a video on that.
@@ShinyTechThings You have a short high speed burst which soon drops to zero and stays there for a long time. This is messing up your average transfer rate resulting in average speed below 1 Gbps. It's much better to have 1 Gbps switch which constantly does about 111 MB/s with no drops at all.
@@discoHR I understand what you are saying now. The reason for this is because the cash on the source or destination Drive has been exceeded. This is why when I run a ram drive and copy tons of data over to it that you get the full speed all the way through it it has nothing to do with the network it is the storage subsystem that is just not fast enough to handle that much data for that long. and this is a normal problem I see it all the time in the data centers that I work in. People complain that they should be able to copy things faster and they are only running on 4-8 SATA drives in RAID10. The RAID controller cache gets filled up and then the speeds drop because it cannot write that fast to the disks on the destination. Even with nvme drives you will still hit bottlenecks with sustained large transfers usually when you hit between 12 GB and 24 GB depending on the drives being used. basically what's happening is everything went across so fast and then it has to come to a screeching halt because the queue for writing is too long and it has to finish writing everything from the cache to the drive. Once the queue has been written to the disk and the cache is flushed out then you can have more data coming across to be written.
@@ShinyTechThings I understand why you think the storage is the problem but I'm not convinced it is a problem in this specific case. A single NVMe SSD drive can handle more than 10 Gbps (1250 MB/s) SUSTAINED write speed, actually almost twice that. There will be no drops at all no matter how much data you write (unless TRIM is disabled but that's another problem). RAID controller cache is small, nowhere near 12 or 24 GB. I'd check NVMe SSD locally with dd command to see if it's really SSD that's bottleneck. For instance: dd if=/dev/zero of=deleteme bs=1M count=32768 This will write 32 GB of zeros to file named "deleteme" and report average write speed. Make sure to use count larger than total RAM on the local machine to avoid caching entire file in RAM which would report false (higher) speed if you have write back enabled on your RAID/SSD. You can check read speed in a similar way: dd if=deleteme of=/dev/null bs=1M I'm a Mac/Linux guy so I have dd command on both platforms by default but IIRC Microsoft's Unix Subsystem for Windows provides the dd command for Windows. If you get sustained write speed below 1250 MB/s then yes, SSD is the bottleneck. If it's at 1250 MB/s or above then it's the switch.
@@discoHR it is and I verified the cheap Intel NVMe I had was in fact the bottleneck. Now I got a higher end Intel DC P3605 that I can abuse 👍. I'm going to be doing some other videos on different server caching software and I'll also do another large file transfer test as well to see how well it handles the writes.
@Cash Dollar that was a much older video and I appreciate the feedback. Please let me know what you think of the latest release. I don't use music through the full clip so more like used through fast forwarded sections of the video but I'd still like to hear back from you on if there's anything else I could improve. Thanks in advance! ua-cam.com/video/2r2opLCQKlo/v-deo.html
For some reason it was slightly cheaper plus it does support higher frequencies and so far 10Gbe without issues over short runs where attenuation can be a concern.
@Matt Chaput I really do try to watch how I talk because of repetition. Please let me know how I'm doing on my newer videos as this is one of my older videos.
@@khanoff13 Correct, but also different versions of iperf will yield different results on different OS's. For Windows to Windows file transfers exceeding NVMe write cache the RAM drive shows the bottleneck of even a NVMe SSD. Is there a specific scenario that you want me to test?
Did you expect a URL right to Netgear's site or like Amazon or something? I'll clarify, but because the thumbnail is presented in Reddit that identifies as a video link but I'll do my best to clarify so please let me know. Thanks!
So i finally upgraded my home network by accident. I never thought the seller on eBay would sell this switch for so cheap! so now I'm running 10 Gigabit and multi-Gig!
Best versatile switch for the price!
bought myself Netgear MS510TX more versatile i think esp with now usb 2.5gb and 5gb adaptors
Totally agree! Plus the next evolutionary step in Wi-Fi access points is 2.5GbE. I wonder if a PoE injector that's Gigabit will allow more than a single gig through if the switch, access point and cabling can handle it because both speeds use the same amount of wires so it's not like a 10/100 Poe injector that doesn't have all 8 wires in it.
@@ShinyTechThings yep the cisco meraki stuff are using 2.5gb and 5gb nic which are very useful as with so many devices rely on wireless.
Have you seen any 2.5GbE PoE injectors yet?
So, unfortunately, it's not your cheap Intel NVMe disk or your RAID 5 array that's causing the speed issue. If you try Atto Disk Benchmark against both drives you'll find pretty average performance with small block transfers and better performance with large block transfers.
I've had 10Gb Ethernet in my home lab for years and just did some tweaks that improved some transfers, but others, particularly from/to Windows disks remain variable, minute to minute. There is something I am honestly missing, or something missing from my implementation.
I am using an Arista Networks 10Gb switch and testing against NVMe, 10-drive NAS with SSD (Samsung) caching, and other endpoints.
What I've tweaked recently, which helped throughput, at least in Atto, was have only Tx flow control enabled, increase my Receive buffers to the max of 4096, and increase my Transmit buffers to their max of 8184. (I have an Aquantia AQtion 10Gbit NIC on the motherboard of this workstation. Most of the other 10G NICs are Intel. Everything is copper.
There may be other parameters that are worth playing around with on the NIC configuration. What seems to be happening is some sort of overflow, of packets or something else, that forces the transmission to slow down until what has already been sent/received is processed.
Any help debugging this would be much appreciated. It would be nice to get 10Gb speeds on something other than a synthetic benchmark.
I plan on tuning it in a bit but even when I copy disk to itself with nothing over the network it does slow down. I'll try some synthetic benchmarks and set the file size to like 20-30GB or something huge and let run overnight next time I login. I'll also tune the NIC's and release another video. I do get the same performance with jumbo frames set to 9k on the machines and switch. Stay tuned for my update and I'll show what I had to do.
After getting through the prescribed "What's up guys", unpacking, etc., I'm almost half the way through your video. Forget it.
@Red Sanders I totally appreciate this criticism and would appreciate if you would like to let me know how I can do it better and I might even re-release this video with the edits that you suggest. I also have chapters on most of my videos to skip around. Please let me know how this could be edited better for your enjoyment, thanks!👍
Copying from NVMe to RAID5 is garbage, but you're limited to the write speed on that Raid5. I'm impressed with the initial copy you did.
I eventually went NVMe to RAM drive for benchmarks. I had to pull some ram out of my old server and put it in the new one just to be able to make a ram Drive big enough without buying any extra ram 🤣🤣🤣🤣
I’m so confused, so I want to build a 10 gig network like this, what is EVERYTHING that I need? Can I do it with just a switch and cat6A/cat7 cables?
@Farrenz great question but what all nodes would you have connected and at what speeds? At a minimum you'll need CAT6A and network cards that support speeds over gigabit and a switch that supports at least multi-gig. How many devices? What are the devices? How long are the lengths of the cable runs?
@@ShinyTechThings I was thinking of just connecting my computer, and eventually sometime a router for my family. Right now, I’m looking at a mikrotik switch, the pcie card, cat6A cables, and that little sfp+ thing for my Ethernet into my pc
@Farrenz the SFP+ to Ethernet (GBIC adapter) might only be gigabit speeds so you might want to get a SFP+ card and a DAC cable if it's a short distance from your PC to the switch. The issue is then it would connect to the router and be bottlenecked from there. If you only plan on having say 2 devices at 10Gig then you could just connect those 2 nodes together from SFP+ to SFP+ cards and a DAC cable and not need the switch. Just have as a separate network defined for the 10 gig interfaces so if your home is say 192.168.1.1-254 then define the other 2 interfaces on 10.97.0.7-8 without a gateway defined and say a 255.255.255.0 subnet. So if your machine is .7 you could ping .8 or vice versa from those 2 nodes but their 1 gig interfaces would still be on 192.168.1.x and use the route to the internet over that interface. Let me know if that helps.
Nice! How about running some iPerf tests to see network throughput?
It's on my list to do a video just on that.
are you still happy with the XS512EM?
I am looking for this switch as well, but I am not sure, as I did read some bad reviews, regarding poweroutages and that you have to unplug all, wait 5 minutes and plug all in again.
Your answer is much appreciated.
I only had one outage when we were out from the house and the breaker tripped from my additional A/C, fans and servers on full throttle for my live stream. Now it's colder out so I don't need to run the A/C anymore but I didn't experience any issues when I turned the breaker back on, however the power was off for probably an hour after my battery backups ran out of juice. So are you saying if I yank the power cord out and plug back in then it could have issues running again without waiting 5+ mins? If so I can give it a try and see and make a video on it for you 🤷♂️. It could also be affecting certain firmware versions as well. Let me know your thoughts on this and I'll see what I can do.
@@ShinyTechThings I am doubting a bit between this Switch and the Mikrotik CRS309
The last one needs extra hardware for connecting to RJ45, but is completely silent.
And I need the Switch on my desk
Yes, that about yanking out the powersupply is probably the same as an poweroutage.
And when the breaker went in, the switch did not work, until all plugged out and waited 5 minutes.
Thank you so far for your input. It helps in forming an opinion.
@@aarbbee I want to reiterate something real quick is that after the power outage in my house from the breaker getting tripped once I got the power back on I did not have to do anything on the switch it just worked and came right back up. What I don't know is that if this is because it was over the course of hours that the power was off or if it is because I am running the latest firmware.
Home lab "cheap" 12 port 10G managed switch at $1,000 USD. Sure seems like my cheap home lab is more on the ghetto side. At $1,000 you can make your own. Intel E3-1270v2, Asus P8P67 WS Revolutiin with (2x PCIe 2.0 x16, 2x PCIe 2 0 x8, and 3x PCIe 2.0 x1) can give you 8x 10G connections, 6x 2.5G ports, and dual 1G all for about $350.
@Voodoo Vinny I didn't pay a grand, I paid like $450ish I just didn't expect the seller to accept my offer when they had it for closer to a grand. That's true and maybe a good upcoming video since I have some spare servers why not build a switch?🤓😎 What cards would you be putting in at that server?
Need some cable management. Get some velcro and tighten that up.
I've got something in the works so stay tuned as I'll try to get that out in Feb. 2021
@@ShinyTechThings I'm just messing with you. I just fixed my messy setup recently.
@Johnny Greene I actually have a purpose for how I did some of my cabling that I will reveal in a future video.
@@ShinyTechThings cool I will check it out. I just bought a Dell PowerEdge R710. I'm thinking on how I can put it somewhere neatly. So far I love it. Only thing is I may replace the Seagate drives with Western Digital Red NAS drives.
I have a friend who works in data storage and they refuse to use seagate due to too many RMA's, however I work at a Dell server only shop and Dell uses the cheapest drives they can buy in bulk and it's a mix. I will say that the failure rates seem to be similar across the board but realize it could be those models when released could be more reliable than the current offerings from those manufacturers. I will say for new expensive SAS drives the Hitachi Helium ones draw less power than the competition. I'm not sure if I can find my notes for the video I started to shoot on a Synology SA3200D I built up but I was supposed on how low the power usage was compared to Synology's official power usage in their documentation. (That video will be a while before I finish it)
Will you get a better NVMe drive to test with?
Eventually yes, but it's a home lab server so it's not at the top of my list. The 10Gbe and multigig was more for faster throughout to my NAS and will do a video on that so stay tuned!
Any reason why you migrated away from Netgear and this switch and toward Ubiquiti?
@humanflycb I run an EnGenius cloud managed switch with 2.5GbE PoE, and SFP+ 10 gig ports but I also got a 10 gig Ubiquiti switch as well to expand my 10 gig ports as I have a handful of servers in my rack, plus my desktop, M1 Max MBP, and 2 10 gig NAS's. My favorite switch though is the other Netgear I reviewed that I got to borrow from a friend and it has 5GbE PoE ports and all multigig. If I ever find one for cheap I'll switch to that over the EnGenius one although the cloud controlled switch will alert me if my house goes offline which has happened before. Otherwise Ubiquiti UniFi is great and easy to manage. The older 10 gig switches are very affordable used but long term firmware support may or may not be there but they did update in Q4 of 2023 so I snagged it. What does your current network look like? How fast do you want to go and what's your budget?
Future full time content creator right here!
Thanks but since I'm a workaholic I'm pretty sure I'll always be working full time anyways but I always try to do what I can to make some good content. Any suggestions besides a faster NVMe SSD to test from? I have one in my old server but need to pull some data from it before I can pull the SSD and some other components a d then eBay it. It's a E5-16XX V3 I believe. I don't use all it's power anymore and heat is my enemy in AZ so went smaller E3-1220V6 in my R330.
Yeah you might have got the switch for cheap, but you paid way to much for fake "Cat7" cables. Even Amazon is getting in on the Cat7 scam - All they did was take Cat6A cables and shield them and repackaged them as "Cat7".
I want a fluke tester that can do 10Gbe RJ45, that would be handy. But it works for now, next to do some Wireshark analysis to see how good the cable actually is.
Hilarious, nuts for your rack!
You bet, I got plenty!
Umm not very cheap LOL If they were cheap I would get 2 of them so I can build a cheap 10 GB network between my home and out side office. and build a 10gb back bone with 1gb feeding computers and 10gb for all servers.
@McCuneWindandSolar MSRP it's not cheap for home but for a business that wants multigig speeds it's very affordable. I would NOT consider this enterprise grade yet as I have had to reboot it twice since I've owned it to fix a port that stopped working. I do have the latest firmware on it but IMO for it to be enterprise quality to never need a reboot to fix a issue like a enterprise Cisco switch would then it would be enterprise ready. So if a reboot would kill your business and cost thousands to reboot during the day, buy a Cisco for $5K+ and the support contract to go with it for updates, but for home or a SMB that a 3 minute outage won't be too painful it's a decent switch. Again I bought for less than half MSRP on eBay but for home I would not pay full price. For a small office IF you NEED the speed it's not a bad option and is very inexpensive for the feature set costing less than the competition.
*"Accidentally"*
Yeah, I made an offer for half their asking price for the switch and they accepted it 😮. Then I was like shoot! Now I need to Craigslist some stuff and float the $ but it all worked out in the end except my cheap Intel drive but I have a few other tests to do with different variables and make a video on that.
Smells like a faulty 10 Gbps switch. 1 Gbps switch outperforms this one in the long run.
How so?
@@ShinyTechThings You have a short high speed burst which soon drops to zero and stays there for a long time. This is messing up your average transfer rate resulting in average speed below 1 Gbps. It's much better to have 1 Gbps switch which constantly does about 111 MB/s with no drops at all.
@@discoHR I understand what you are saying now. The reason for this is because the cash on the source or destination Drive has been exceeded. This is why when I run a ram drive and copy tons of data over to it that you get the full speed all the way through it it has nothing to do with the network it is the storage subsystem that is just not fast enough to handle that much data for that long. and this is a normal problem I see it all the time in the data centers that I work in. People complain that they should be able to copy things faster and they are only running on 4-8 SATA drives in RAID10. The RAID controller cache gets filled up and then the speeds drop because it cannot write that fast to the disks on the destination. Even with nvme drives you will still hit bottlenecks with sustained large transfers usually when you hit between 12 GB and 24 GB depending on the drives being used. basically what's happening is everything went across so fast and then it has to come to a screeching halt because the queue for writing is too long and it has to finish writing everything from the cache to the drive. Once the queue has been written to the disk and the cache is flushed out then you can have more data coming across to be written.
@@ShinyTechThings I understand why you think the storage is the problem but I'm not convinced it is a problem in this specific case. A single NVMe SSD drive can handle more than 10 Gbps (1250 MB/s) SUSTAINED write speed, actually almost twice that. There will be no drops at all no matter how much data you write (unless TRIM is disabled but that's another problem). RAID controller cache is small, nowhere near 12 or 24 GB.
I'd check NVMe SSD locally with dd command to see if it's really SSD that's bottleneck. For instance:
dd if=/dev/zero of=deleteme bs=1M count=32768
This will write 32 GB of zeros to file named "deleteme" and report average write speed. Make sure to use count larger than total RAM on the local machine to avoid caching entire file in RAM which would report false (higher) speed if you have write back enabled on your RAID/SSD.
You can check read speed in a similar way:
dd if=deleteme of=/dev/null bs=1M
I'm a Mac/Linux guy so I have dd command on both platforms by default but IIRC Microsoft's Unix Subsystem for Windows provides the dd command for Windows.
If you get sustained write speed below 1250 MB/s then yes, SSD is the bottleneck. If it's at 1250 MB/s or above then it's the switch.
@@discoHR it is and I verified the cheap Intel NVMe I had was in fact the bottleneck. Now I got a higher end Intel DC P3605 that I can abuse 👍. I'm going to be doing some other videos on different server caching software and I'll also do another large file transfer test as well to see how well it handles the writes.
Hey a new ending but not outtakes?
Correct, not any good ups when I made that video.
I'll just "go ahead" and leave now... 😎👍
What questions do you have on this?
How many times can someone say “go ahead and” in one video.
@Jeff Gibson I didn't notice but will try to be better about that moving forward. What would you like to see more videos of?
Please dump the cheesy music.
@Cash Dollar that was a much older video and I appreciate the feedback. Please let me know what you think of the latest release. I don't use music through the full clip so more like used through fast forwarded sections of the video but I'd still like to hear back from you on if there's anything else I could improve. Thanks in advance!
ua-cam.com/video/2r2opLCQKlo/v-deo.html
That nvme is not worth it
@Ricardo Trujillo I have a Intel DC P3605 now which performs much better now. What NVMe drive(s) do you have?
Cat7. Should have just bought cat6
For some reason it was slightly cheaper plus it does support higher frequencies and so far 10Gbe without issues over short runs where attenuation can be a concern.
This guy is a . . . Geek. 🤓
What gave it away?🤓🤣
not that there is anything wrong with that haha, im one too.
Please try to stop saying "go ahead and..." in every. single. sentence.
@Matt Chaput I really do try to watch how I talk because of repetition. Please let me know how I'm doing on my newer videos as this is one of my older videos.
Iperf3
What about it? Could you please elaborate?
@@ShinyTechThings if you want to test network throughput, you should overcome influences of other factors like filesystems, IOPS limitations etc.
@@khanoff13 Correct, but also different versions of iperf will yield different results on different OS's. For Windows to Windows file transfers exceeding NVMe write cache the RAM drive shows the bottleneck of even a NVMe SSD. Is there a specific scenario that you want me to test?
Crazy take, don't bait and switch people to get your videos on Reddit. You posted a link that should've linked to a switch. Not a youtube video.
Did you expect a URL right to Netgear's site or like Amazon or something? I'll clarify, but because the thumbnail is presented in Reddit that identifies as a video link but I'll do my best to clarify so please let me know. Thanks!