Honestly Patrick thank you for the reminder about the optics/DACs half the time i forget to consider the cost of these in the home setting, because at work, we often have 'stock'
Again perfect timing. I'm just quoting $200k of Cisco crap. This is so refreshing. Nobody does reviews like this. So hard to compare real specs of mid level switches. No vendor shows the real guts of the gear. Cisco and it's ridiculous licensing is absolutely crazy. Maybe it's possible to find real specs somewhere but definitely not in marketing fluff of big guys. Does it look like I'm big fan of Cisco ? I'm not.
Also be aware that it is now not supported for OS10.5 onwards, so even Dell have given up on supporting it. There is no multicast routing available, no IEEE1588 support in software (although technically it is supported in hardware). We've used one for a few years, and provided you can work with the limitations it is fine and reliable, but there are some significant caveats
One thing I remember about Cisco saying was if your device was out of warranty by the time it failed (if you just bought a device with no smart net) then you were on your own and got to buy another one. I don't know why that stuck with me.
I have a Moxa Nport 5610-16 serial console, old but hey it still works and it sits on a dedicated management network. Well apart from the fact the battery for the RTC is dead. The problem is the utter morons who designed thing decided that a Panasonic BR1225/HCN instead of a CR2032 and a holder would be a good idea. Yeah so you have to remove the PCB and desolder the battery to change it. Even better getting them in the UK is next to impossible. I have had one on back order now for over two months now. I won't care but there is plenty of space for a CR2032 coin cell. Frankly the person who made that design choice needs a jolly good thrashing with a clue stick and then sacking.
My experience with this switch series is getting an evaluation unit, turning it on and immediately finding multiple python stack traces, parser errors and other critical bugs. Maybe it's better with other OSes though :)
the demo depot probably didnt reflash it. It is relatively easy to get into ONiE and reflash with either DellOS10 (Force10 software) or any other ONiE compatible OS like SONiC
@@jeremygmail personally we are running SONiC on everything, but we are using only Edgecore and Mellanox/Nvidia, so I have no idea how Dell works with SONiC vs other OSes, but I found SONiC to be super reliable for us on all hardware.
@@jannikmeissner interesting. I do have a few switches in stock. Maybe I can run SONiC on them to see if I can manage them better than DellOS10. I prefer the ansible way to management and dont mind getting my hands dirty. Thank you for the feedback!
Thanks, I was buying a lot of the Arista switches they dropped to this $1200 range at one point for their 100g however now they are up in the 5k+ range again.
Marketing got a bit crazy with the s5200 series. Some are 25gig, and one of them is an all 100gig. The n3200 series is like that too. They range from 1gig to 10gig depending on the exact model.
The Atom C2000 failure should be DIY fixable by some minor soldering work (you have to bodge in an extra resistor or something). The question is if the correct pin you have to solder to is exposed on the PCB or not. It is exposed on some Synology NAS boards that suffer from this problem, so they are easily fixable.
Very interesting (and yes, I too got bit by the AVR54 fiasco, via an ixSystems FreeNAS Mini). Definitely interesting in everything 100 GbE, especially switches for the small-ish home lab.
Show the chip!! STH are the only ones who ever revealed (after 4 years in the market!) that this switch is a Marvell/Cavium Xpliant (XP70). It's just mentioned confidently in the video and article, but why are there no photos of it? Just of a heatsink. Where is the info from? It's the only public source on the planet, so we the people demand to know! :)
I'm still trying to figure out a cost effective way to bridge between my 100 Gbps Infiniband network and deploying a 100GbE networking layer as well so that non-IB traffic can run on the 100GbE whilst IB traffic, of course, runs on the 100 Gbps IB layer.
Hi @ServeTheHome, another interesting video. One thing I think you should have mentioned: You keep talking about the "front" and the "back" of the switch, but that front and back are defined by the airflow configuration of the switch, and in my experience, most of the switches we sell are PSU -> IO airflow, so what you keep calling the front would actually be the back and vice-versa with that configuration. There will definitely be switches out there that are IO->PSU, but for a datacentre setup, this isn't the most optimal configuration. #IWork4Dell
I don't fully agree, many manufacturers use the terms RTF/FTR (rear to front, front to rear) implying that the PSU side is the rear, and the port side is the front. The airflow direction depends on how the switch is used, if used as a TOR (top-of-rack) switch, it will be mounted with the port facing the back of the rack, and you'll use RTF airflow. If used in a networking rack, the ports typically points towards the front of the rack, and thus use FTR airflow. If you look at high-end DC switches, they are more often FTR, because they are used as cores, and not TOR.
We standardize on ports being the front and fans/ PSU being the rear. The first photos we show on the STH main site always have this orientation. We have to have a standard nomenclature and methodology across brands otherwise it gets too confusing.
@@HomelabExtreme RTF/FTR maybe correct for other manufacturers but this is a Dell switch being reviewed, so instead of using front & back we refer to IO -> PSU or PSU -> IO. So in the review they talk about no console port etc. on the "front", but on a PSU -> IO switch this would actually be the front :-)
@@ServeTheHomeVideo That makes sense, but when you mention that there are no console ports on the "front", in what I would normally consider the normal config, being PSU -> IO, those ports would actually be on the "front". Anyway, as always I enjoy your reviews, definitely not criticising the review, but just thought a mention of airflow might be appropriate.
@@strategicalit so from an industry perspective we tend to have standardized front and rear. The switch may be mounted and have airflow for port to PSU or PSU to port, but that gets into how a switch is deployed, not how we discuss the switch static on a bench. If you have PSU to port airflow, usually you are mounting it with the port face on the cold aisle and the serial/USB/management ports are somewhere 0.5m or so in the middle of a rack. Most interactions will still occur with the port side, and that is the side vendors generally paint/ spend time on so we call that the front to standardize.
Nothing to do with this video but I read somewhere that you're looking at an Ampere Altra box. You should hook up with Jeff Geerling (RUclips channel of same name) who does all sorts of interesting things with Raspberry Pi boxes (much the same ARM as the Altra) and would no doubt wibble incredibly over the chance to do interesting things with the Altra.
A quick look on ebay I see these switches both with the Dell OS installed and without, is the dell OS something you have to pay for if it's not included? are there free options from third parties?
Just done some searching and read the text article that goes with this and it looks like the dell network OS requires a license, but the switch is also compatible with openswitch which appears to be free.
@@petermichaelgreen I really WISH that this is not just what they state on their (dead) website. Installed 3.2 and 2.3.1. neither of them "work" with regards to finding the ports and map them to /dev. Got a heavy, expensive brick ... someone needs one ? EDIT: Use Version OPX 3.0 and dont (apt) update ...
Yes, but my sense is that it will not be a mainstream server processor in the short term. Even in a switch like this where an Arm CPU today may be a good choice, having x86 and broad compatibility is usually preferable (hyper-scale folks excluded since they can do their own thing)
Probably in a highly custom SoC, not vanilla RISCV. Not to mention that although the ISA is open source, the actual chip design isn’t (SciFive’s layout of the logic gates is closed-source) or not very good (the open-source Berkeley layout is practically an incomplete microcontroller, cannot run Linux at all)
Intel Atom C2000 does this mean these Dell switches is a ticking time bomb? The C2000 bug with clock degradation that murdered network appliances and Intel refuse to fix it, just buy new was solution.... I've experienced 2 Clavister firewalls per month dying, and my private nas also died. Clavister said that Intel refused to deal with it and RMA required a working version to be sent in and then be confirmed with the bug, when it died it could be confirmed hence no way to get money back, Intel is so stupid.
They have a done an okay job with the automation framework so I dont even have to login to a DellOS10 switch anymore to make changes.... Debugging is a different problem...
Dell Power edge r750 server unboxing
ruclips.net/video/282V_upA0DQ/видео.html
Dell s4128T-ON switch unboxing
ruclips.net/video/CBwvwfVs6Ws/видео.html
Me: "Ah yes, I definitely need 100GBe networking for my plex server at home."
I upgraded my server to 40GBe. Worth it.
@@xtremeownagedotcom must be fast
Honestly Patrick thank you for the reminder about the optics/DACs half the time i forget to consider the cost of these in the home setting, because at work, we often have 'stock'
Again perfect timing. I'm just quoting $200k of Cisco crap. This is so refreshing. Nobody does reviews like this. So hard to compare real specs of mid level switches. No vendor shows the real guts of the gear. Cisco and it's ridiculous licensing is absolutely crazy. Maybe it's possible to find real specs somewhere but definitely not in marketing fluff of big guys. Does it look like I'm big fan of Cisco ? I'm not.
Yes please do as many teardowns as possible
Also be aware that it is now not supported for OS10.5 onwards, so even Dell have given up on supporting it. There is no multicast routing available, no IEEE1588 support in software (although technically it is supported in hardware). We've used one for a few years, and provided you can work with the limitations it is fine and reliable, but there are some significant caveats
One thing I remember about Cisco saying was if your device was out of warranty by the time it failed (if you just bought a device with no smart net) then you were on your own and got to buy another one. I don't know why that stuck with me.
After swapping out a CV axel, replacing a CMOS battery under the SP board isn’t a big deal
Fair point.
I kinda wonder if the placement is deliberate to dissuade anyone from trying to hotswap it.
I have a Moxa Nport 5610-16 serial console, old but hey it still works and it sits on a dedicated management network. Well apart from the fact the battery for the RTC is dead. The problem is the utter morons who designed thing decided that a Panasonic BR1225/HCN instead of a CR2032 and a holder would be a good idea. Yeah so you have to remove the PCB and desolder the battery to change it. Even better getting them in the UK is next to impossible. I have had one on back order now for over two months now. I won't care but there is plenty of space for a CR2032 coin cell. Frankly the person who made that design choice needs a jolly good thrashing with a clue stick and then sacking.
My experience with this switch series is getting an evaluation unit, turning it on and immediately finding multiple python stack traces, parser errors and other critical bugs. Maybe it's better with other OSes though :)
the demo depot probably didnt reflash it. It is relatively easy to get into ONiE and reflash with either DellOS10 (Force10 software) or any other ONiE compatible OS like SONiC
@@jeremygmail Oh no, this was a brand new unit. Completely fresh everything. Their OS was just that buggy.
@@jeremygmail personally we are running SONiC on everything, but we are using only Edgecore and Mellanox/Nvidia, so I have no idea how Dell works with SONiC vs other OSes, but I found SONiC to be super reliable for us on all hardware.
As mentioned prior Dell switches are shit... Nothing but complaints in Reddit Networking threads. They are cheap for a reason.
@@jannikmeissner interesting. I do have a few switches in stock. Maybe I can run SONiC on them to see if I can manage them better than DellOS10. I prefer the ansible way to management and dont mind getting my hands dirty. Thank you for the feedback!
Thanks, I was buying a lot of the Arista switches they dropped to this $1200 range at one point for their 100g however now they are up in the 5k+ range again.
MMMMMMMM Hardware. Yes a look at the S52 would be appreciated.
Marketing got a bit crazy with the s5200 series. Some are 25gig, and one of them is an all 100gig. The n3200 series is like that too. They range from 1gig to 10gig depending on the exact model.
Thank you for the membership John! Hopefully you saw the S5200 series we did.
@@ServeTheHomeVideo Yes I did, thanks!
The Atom C2000 failure should be DIY fixable by some minor soldering work (you have to bodge in an extra resistor or something). The question is if the correct pin you have to solder to is exposed on the PCB or not. It is exposed on some Synology NAS boards that suffer from this problem, so they are easily fixable.
My old HP is chugging along quite well :)
Very interesting (and yes, I too got bit by the AVR54 fiasco, via an ixSystems FreeNAS Mini).
Definitely interesting in everything 100 GbE, especially switches for the small-ish home lab.
I am considering getting the S5200, would love to see your video and thoughts on it!
I just love your videos.. so clear and conscice. You are awesome Patrick!
Wow, thank you!
Show the chip!! STH are the only ones who ever revealed (after 4 years in the market!) that this switch is a Marvell/Cavium Xpliant (XP70). It's just mentioned confidently in the video and article, but why are there no photos of it? Just of a heatsink. Where is the info from? It's the only public source on the planet, so we the people demand to know! :)
thanks for making these video's :)
Thanks for watching and subscribing. Glad you like them!
I absolutely love your videos, please keep making them ❤️
Regarding causing too much noise in the office, why not do a liquid immersion experiment with this switch?
I'm still trying to figure out a cost effective way to bridge between my 100 Gbps Infiniband network and deploying a 100GbE networking layer as well so that non-IB traffic can run on the 100GbE whilst IB traffic, of course, runs on the 100 Gbps IB layer.
Hi @ServeTheHome, another interesting video. One thing I think you should have mentioned: You keep talking about the "front" and the "back" of the switch, but that front and back are defined by the airflow configuration of the switch, and in my experience, most of the switches we sell are PSU -> IO airflow, so what you keep calling the front would actually be the back and vice-versa with that configuration. There will definitely be switches out there that are IO->PSU, but for a datacentre setup, this isn't the most optimal configuration. #IWork4Dell
I don't fully agree, many manufacturers use the terms RTF/FTR (rear to front, front to rear) implying that the PSU side is the rear, and the port side is the front.
The airflow direction depends on how the switch is used, if used as a TOR (top-of-rack) switch, it will be mounted with the port facing the back of the rack, and you'll use RTF airflow.
If used in a networking rack, the ports typically points towards the front of the rack, and thus use FTR airflow.
If you look at high-end DC switches, they are more often FTR, because they are used as cores, and not TOR.
We standardize on ports being the front and fans/ PSU being the rear. The first photos we show on the STH main site always have this orientation. We have to have a standard nomenclature and methodology across brands otherwise it gets too confusing.
@@HomelabExtreme RTF/FTR maybe correct for other manufacturers but this is a Dell switch being reviewed, so instead of using front & back we refer to IO -> PSU or PSU -> IO. So in the review they talk about no console port etc. on the "front", but on a PSU -> IO switch this would actually be the front :-)
@@ServeTheHomeVideo That makes sense, but when you mention that there are no console ports on the "front", in what I would normally consider the normal config, being PSU -> IO, those ports would actually be on the "front". Anyway, as always I enjoy your reviews, definitely not criticising the review, but just thought a mention of airflow might be appropriate.
@@strategicalit so from an industry perspective we tend to have standardized front and rear. The switch may be mounted and have airflow for port to PSU or PSU to port, but that gets into how a switch is deployed, not how we discuss the switch static on a bench. If you have PSU to port airflow, usually you are mounting it with the port face on the cold aisle and the serial/USB/management ports are somewhere 0.5m or so in the middle of a rack. Most interactions will still occur with the port side, and that is the side vendors generally paint/ spend time on so we call that the front to standardize.
Nothing to do with this video but I read somewhere that you're looking at an Ampere Altra box. You should hook up with Jeff Geerling (RUclips channel of same name) who does all sorts of interesting things with Raspberry Pi boxes (much the same ARM as the Altra) and would no doubt wibble incredibly over the chance to do interesting things with the Altra.
Yes yes. Stop dangling the candy and just give us the s52
I want to see a vid on the S52 switch
Just one question: does it do RoCEv2 and the related pfc magic?
Wasn't that avr bug fixed by putting a resistor on the lpc bus?
Let’s open the S52!!!
This a tor switch so it's all going to be DAC cables with very few optics involved.
Have you considered reviewing something from Ruckus?
Yes. Working on a new space for WiFi testing
Ooooh, this AVR gives me flashbacks... also, yes, nud... I mean teardowns are good, more are better.
How much buffer memory is on the switch? Compare that to something like an Arista R3 switch and you can see why it is very low cost.
Service Tag 🏷 on the back of the server?
=
Vehicle licence plate under the gas ⛽️ tank
A quick look on ebay I see these switches both with the Dell OS installed and without, is the dell OS something you have to pay for if it's not included? are there free options from third parties?
Just done some searching and read the text article that goes with this and it looks like the dell network OS requires a license, but the switch is also compatible with openswitch which appears to be free.
@@petermichaelgreen I really WISH that this is not just what they state on their (dead) website. Installed 3.2 and 2.3.1. neither of them "work" with regards to finding the ports and map them to /dev. Got a heavy, expensive brick ... someone needs one ? EDIT: Use Version OPX 3.0 and dont (apt) update ...
I have a question. Does this switch do stacking?
So why is it cheaper? Cause you buy last generation second hand?
Actually the S5048F-ON, the generation before built around Brocade is even less expensive. It makes this one somewhat of an anomaly in that regard.
So ...is there any way to make this quiet?
liquid cooling
Put it in a room far away from where your computer is and run long cables :) That is about it.
Questions: How much do 100g fiber links cost and is it illegal to run a pirate ISP out if your rental duplex.
Leave it powered off
Have you got both PSU's plugged in? They used to run loud with just 1 PSU
Would you be able to get RouterOs running on this? I tried looking, I know you can run it on x86 machines in general.
Duh yeah! Of course we want to look inside of it!
TEARDOWN TEARDOWN TEARDOWN!
noise? is that for family lab?
Are RISC-V going to be used in servers.
Yes, but my sense is that it will not be a mainstream server processor in the short term. Even in a switch like this where an Arm CPU today may be a good choice, having x86 and broad compatibility is usually preferable (hyper-scale folks excluded since they can do their own thing)
Probably in a highly custom SoC, not vanilla RISCV. Not to mention that although the ISA is open source, the actual chip design isn’t (SciFive’s layout of the logic gates is closed-source) or not very good (the open-source Berkeley layout is practically an incomplete microcontroller, cannot run Linux at all)
Intel Atom C2000 does this mean these Dell switches is a ticking time bomb? The C2000 bug with clock degradation that murdered network appliances and Intel refuse to fix it, just buy new was solution....
I've experienced 2 Clavister firewalls per month dying, and my private nas also died. Clavister said that Intel refused to deal with it and RMA required a working version to be sent in and then be confirmed with the bug, when it died it could be confirmed hence no way to get money back, Intel is so stupid.
mSATA good luck 🤞 finding them
Hit Reddit networking on Dell switching. Not impressive.
Dell OS 10 makes me sick 🤮 🤮
They have a done an okay job with the automation framework so I dont even have to login to a DellOS10 switch anymore to make changes....
Debugging is a different problem...
@@jeremygmail Yeah debugging and the commands was what I was referring to
1k weeewww thats alot of money. Also in my recommended lol.
mb