That port on the back looks like a Cisco RPS power port. Some Catalyst series switches will have this. A Cisco RPS will be a 1u device providing resillient power to switches using the connectors. There is a data reporting aspect to this as well on the pins. I dont know the CLI of CBS switches, but you might see an entry for RPS status if you enter "show environment all".
The internal card edge could well be for manufacturing: When the board has been assembled, it has no software yet. We add connectors to similar boards, where the connector carries power as well as programming signals. This allows manufacturing to plug one connector into the assembled board, and the board gets powered, programmed, and tested using this connector before it gets put into the case. Since the connector is only used once, cheap is good, and a card edge is cheap.
The 24 port was easily the most interesting because of all the exposed packages (lack of heatsinks), and I suspect part of that is that it's newer, the ICs are based on newer processes that requires less power and therefore less heat is being produced per IC; the way it's structured is fascinating, the exposed chips would be little more than port multipliers, bringing in a 8-10Gbps bus connection, likely as high as 16-20 gpbs to upstream (for bi-directional data transfer) up to a controller IC that directly connects to each switch chip, and provides the 4x10G SFP+ links; obviously the 10G chip with the three downstream connections to the 1GbE ports will be the fastest chip on the board to handle all the throughput. The interesting thing is that it connects to the FPGA from there. Knowing this switch line (which used to be the SG300X or similar), there's something called SDR, or Software-Defined Routing, and what SDR does is that it can actually handle L3 traffic being switched by the unit based on conditions; I suspect that the FPGA is to make the SDR features work faster to keep up with the 10G + speeds of the switch. there's a 512mbyte RAM chip on the 10G chip, which is likely the CAM table (aka mac-address table).... So what happens is, if a packet comes in on a VLAN that transits VLAN boundaries (L3 switching) it's forwarded up to the FPGA, which will do the forwarding, if the FPGA isn't programmed to handle the traffic (maybe there isn't a rule programmed for the flow), it will forward it to the CPU, which is the ARM CPU there; that ARM CPU runs all GUI/CLI, config, and CPU-based routing and switching, as well as monitoring and whatnot (and likely has it's own RAM in a Package on package design). The Flash chip hold the at-rest config, firmware, etc. for all chips. It's an incredibly interesting design and it's likely that other switches follow the same concepts, but run at relatively faster speeds (like the 12XT will move quite a lot faster than the 24T-4X), and require active cooling to keep up with the faster chips and design. Usually the micro-b USB is simply an internal USB to Serial adapter which links up with the serial interface on the RJ45 port for console, specifically for programming and would generally be wired directly into the CPU (the ARM chip), and of course a USB host port for the external USB-A port, which would normally be used for flashing - eg, in the event the firmware on flash is corrupt or missing, the base programming on the board (UEFI or similar) would be able to check the chip for a firmware file and it would be able to follow a procedure to copy that to the internal flash storage and recover the switch. I suspect the edge connector is for initial programming, with enough to power on the unit and have a data connection to the ARM CPU and/or direct to the flash storage, to do initial programming and testing while it's on the assembly line. I'm mostly surmising from experience with how systems are designed from a long history of dealing with computers, and a long history of dealing with switches, as well as knowledge of the inner-workings of how some similar items do what they do. Bluntly, you'd get more performance putting in an FPGA to do L2/L3 switching when you have programmable interfaces, such as with SDR, than you would relying on a CPU, but you would still need a CPU in there to do the general work of GUI/CLI, and programming the FPGA, as well as system monitoring; but the CPU can be incredibly low power, compared to everything else since it wouldn't need to do more than that 99.999% of the time, with all the traffic being handled by the IC's (ASICs) and/or the FPGA. Since something like SDR can be changed on the fly, the FPGA would be a logical choice compared to simply throwing in a very powerful CPU to try to keep up with switching the related traffic; and overall it would provide a faster and better flow of data at a nominally lower cost than either putting in an ASIC (which cannot be changed/programmed) or by upgrading the CPU enough to handle the flows, which may require something a lot more power hungry and it would also require rather extensive software switching on the CPU - which would also be CPU bound, so if the CPU is busy handling CLI/GUI requests, it may bog down, and vice-versa, the GUI/CLI may be bogged down by incredibly high switching activity. Provides a better and more consistent experience overall. Like I said, I expect most of the (especiallly CBS350) switches are designed similarly. Very fascinating, and thank you.
I guess it's nice to know you're paying for FPGAs and plastic covers on PSUs rather than just the decent CLI! It's interesting finding out how platforms for these span across different product ranges rather than one platform being used in a single range.
Wow! Thanks for show. Most products do not need more 19" form factor, 10" enough. Yeah switches these days looks very clean and unstagged. 😉🤗 1st comment after orginal author wrote. 😂
@@camerongray1515 Yeah, in history the admins in companies build DIY boxes for small boxes but these days the most companies haven't admins nor house managers with offices. Possible do you can build some self and resell the blue print? This makes you independent from YT, Twitch, and other big companies. A horizontal small switch with keystone place at one other side sounds great. Same what I can't understand why not all network related stuff only in front - and power management at back.
I suppose historically a lot of that DIY construction was done due to a lack of off the shelf solutions. Nowadays, the cost to a company of staff time vs buying off the shelf solutions just makes the DIY route no longer financially viable. Complete off the shelf solutions also come with commercial support which many companies want. As for making something myself, it's not really something I have any experience with. CZcams isn't my full time job so I'm not reliant on it for my main source of income. I'm already busy enough as is with my day job and other general life things. As for why switches have network ports on the front and power at the rear - this simply makes more sense for most rack situations. Ports on the back would be awkward to access in almost all racks. Netgear do actually sell some rack mounted switches with rear ports as part of their AV range however these are designed specifically for AV racks where it's desireable for all cables to be behind the equipment to give a clean front.
I noticed the one switch says "Manufactured for Cisco" which isn't something I would expect an an actual made by Cisco switch to say. So I'm thinking they either contracted out the switch development, or they found a third party switch and put their own custom OS on it and rebranded it.
Yeah, it's interesting, I'd be interested to know what that "R" logo is for - some sort of OEM possibly? That said, those definitely aren't a white label product that has been simply rebranded, but I wouldn't be surprised if the manufacturing was outsourced to an external company. Although that's dead common for most manufacturers nowadays.
PS in CBS-250-24T-4X is almost identical as in old Netgear GS724T. I got mine recently recapped. I bet it is done by the same factories, even if brand would be diffrent. Apparently, world of network equipment power supplies is pretty small.
I've been looking at the CBS350-48XT-4X but it doesn't look like it plays nice with AOCs which is frustrating as the furthest run I have is 10m+ and I'm not 100% sure a DAC will work. I also didn't realise they don't have stackwise cables on them
Where are you seeing this about AOCs, I can't seem to find anything? Personally if I needed a fibre link between the machines, I'd get separate transceivers with a separate fibre cable. That way you can use different transceivers on each end if either device requires a specific transceivers vendor. That said, the switch has 10GBaseT ports so as long as the device on the other end can also support a 10GBaseT connection, you'd be able to just run a regular copper cable between them and get up to 100m with CAT6A or 55m of CAT6. They don't have Stackwise connections, instead they stack using SFP+ cables between the switches (either DACs or fibre) although they can't stack using the regular RJ45 ports, only the SFP+ ones.
It depends on the model across the 220, 250 and 350 series. The gigabit, non-PoE models are generally passively cooled but the larger ones with PoE or a large number of multi-gig ports have fans. The model comparison charts for each series detail how many fans each model has, e.g. www.cisco.com/c/en/us/products/switches/business-350-series-managed-switches/models-comparison.html
For smaller project installs I've been using the Cisco C9200CX-12P-2X2G switches. A typical application would be like a security camera deployment with VOIP and maybe wifi. (Think security guard stations at a business park connected to a corporate network) They are passively cooled PoE+ switches with 12 powered gigabit ports, 2 unpowered gigabit ports, and 2 10G SFP+ ports and can deliver up to 240W. Now they do run hot but they are completely fanless with a giant heatsink block on top. (The heatsink is basically the case.) They have a full sized SD slot for hosting applications such as an embedded wireless controller and run IOS 17 Lite XE which to me doesn't seem or visibly appear any different than normal IOS XE. The suggested retail is about $1800, but you can get them for quite a bit less if you're a corporate customer.
They look good, however as soon as you go to the higher end Cisco Catalyst switches you're needing to deal with licencing and support contracts, something that makes them unviable for many small businesses settings. They're great switches for sure, but just not practical cost wise for many environments.
Nah industrial switches would be like an IE 3000 or Stratix line. These aren't even based on the normal Cisco catalyst line which is their normal business switch.
Short answer is "I don't need them" however I enjoy tinkering with these things and it also allows me to produce these CZcams videos. On the other hand, a CZcams video showing a basic wireless router and unmanaged desktop switch wouldn't exactly be interesting! Additionally, having 10GbE switching allows me to transfer large video files to and from my NAS at a significantly higher speed than I could over a basic gigabit connection.
Buy on Amazon (Affiliate):
- CBS350-12NP-4X: geni.us/YnP0
- CBS350-12XT: geni.us/6ZVqo
- CBS250-24T-4X: geni.us/RP5OYi
- CBS250-16T-2G: geni.us/ZLjI
That port on the back looks like a Cisco RPS power port. Some Catalyst series switches will have this. A Cisco RPS will be a 1u device providing resillient power to switches using the connectors. There is a data reporting aspect to this as well on the pins. I dont know the CLI of CBS switches, but you might see an entry for RPS status if you enter "show environment all".
My 3750 v2 has this port too, i was gonna say its a Cisco RPS power port. Quirky how its there tho :>
Exactly what I was trying to say.
The internal card edge could well be for manufacturing: When the board has been assembled, it has no software yet. We add connectors to similar boards, where the connector carries power as well as programming signals. This allows manufacturing to plug one connector into the assembled board, and the board gets powered, programmed, and tested using this connector before it gets put into the case. Since the connector is only used once, cheap is good, and a card edge is cheap.
There are similar looking edge connectors that I've seen on the small 8 to 12 port small form factor Catalyst switches.
The 24 port was easily the most interesting because of all the exposed packages (lack of heatsinks), and I suspect part of that is that it's newer, the ICs are based on newer processes that requires less power and therefore less heat is being produced per IC; the way it's structured is fascinating, the exposed chips would be little more than port multipliers, bringing in a 8-10Gbps bus connection, likely as high as 16-20 gpbs to upstream (for bi-directional data transfer) up to a controller IC that directly connects to each switch chip, and provides the 4x10G SFP+ links; obviously the 10G chip with the three downstream connections to the 1GbE ports will be the fastest chip on the board to handle all the throughput. The interesting thing is that it connects to the FPGA from there. Knowing this switch line (which used to be the SG300X or similar), there's something called SDR, or Software-Defined Routing, and what SDR does is that it can actually handle L3 traffic being switched by the unit based on conditions; I suspect that the FPGA is to make the SDR features work faster to keep up with the 10G + speeds of the switch. there's a 512mbyte RAM chip on the 10G chip, which is likely the CAM table (aka mac-address table).... So what happens is, if a packet comes in on a VLAN that transits VLAN boundaries (L3 switching) it's forwarded up to the FPGA, which will do the forwarding, if the FPGA isn't programmed to handle the traffic (maybe there isn't a rule programmed for the flow), it will forward it to the CPU, which is the ARM CPU there; that ARM CPU runs all GUI/CLI, config, and CPU-based routing and switching, as well as monitoring and whatnot (and likely has it's own RAM in a Package on package design). The Flash chip hold the at-rest config, firmware, etc. for all chips. It's an incredibly interesting design and it's likely that other switches follow the same concepts, but run at relatively faster speeds (like the 12XT will move quite a lot faster than the 24T-4X), and require active cooling to keep up with the faster chips and design.
Usually the micro-b USB is simply an internal USB to Serial adapter which links up with the serial interface on the RJ45 port for console, specifically for programming and would generally be wired directly into the CPU (the ARM chip), and of course a USB host port for the external USB-A port, which would normally be used for flashing - eg, in the event the firmware on flash is corrupt or missing, the base programming on the board (UEFI or similar) would be able to check the chip for a firmware file and it would be able to follow a procedure to copy that to the internal flash storage and recover the switch. I suspect the edge connector is for initial programming, with enough to power on the unit and have a data connection to the ARM CPU and/or direct to the flash storage, to do initial programming and testing while it's on the assembly line.
I'm mostly surmising from experience with how systems are designed from a long history of dealing with computers, and a long history of dealing with switches, as well as knowledge of the inner-workings of how some similar items do what they do. Bluntly, you'd get more performance putting in an FPGA to do L2/L3 switching when you have programmable interfaces, such as with SDR, than you would relying on a CPU, but you would still need a CPU in there to do the general work of GUI/CLI, and programming the FPGA, as well as system monitoring; but the CPU can be incredibly low power, compared to everything else since it wouldn't need to do more than that 99.999% of the time, with all the traffic being handled by the IC's (ASICs) and/or the FPGA. Since something like SDR can be changed on the fly, the FPGA would be a logical choice compared to simply throwing in a very powerful CPU to try to keep up with switching the related traffic; and overall it would provide a faster and better flow of data at a nominally lower cost than either putting in an ASIC (which cannot be changed/programmed) or by upgrading the CPU enough to handle the flows, which may require something a lot more power hungry and it would also require rather extensive software switching on the CPU - which would also be CPU bound, so if the CPU is busy handling CLI/GUI requests, it may bog down, and vice-versa, the GUI/CLI may be bogged down by incredibly high switching activity.
Provides a better and more consistent experience overall. Like I said, I expect most of the (especiallly CBS350) switches are designed similarly. Very fascinating, and thank you.
I guess it's nice to know you're paying for FPGAs and plastic covers on PSUs rather than just the decent CLI!
It's interesting finding out how platforms for these span across different product ranges rather than one platform being used in a single range.
Wow! Thanks for show. Most products do not need more 19" form factor, 10" enough. Yeah switches these days looks very clean and unstagged. 😉🤗 1st comment after orginal author wrote. 😂
Yeah, I quite like what some vendors are doing now with half-width switches where you can install two in 1U.
@@camerongray1515 Yeah, in history the admins in companies build DIY boxes for small boxes but these days the most companies haven't admins nor house managers with offices. Possible do you can build some self and resell the blue print? This makes you independent from YT, Twitch, and other big companies.
A horizontal small switch with keystone place at one other side sounds great. Same what I can't understand why not all network related stuff only in front - and power management at back.
I suppose historically a lot of that DIY construction was done due to a lack of off the shelf solutions. Nowadays, the cost to a company of staff time vs buying off the shelf solutions just makes the DIY route no longer financially viable. Complete off the shelf solutions also come with commercial support which many companies want.
As for making something myself, it's not really something I have any experience with. CZcams isn't my full time job so I'm not reliant on it for my main source of income. I'm already busy enough as is with my day job and other general life things.
As for why switches have network ports on the front and power at the rear - this simply makes more sense for most rack situations. Ports on the back would be awkward to access in almost all racks. Netgear do actually sell some rack mounted switches with rear ports as part of their AV range however these are designed specifically for AV racks where it's desireable for all cables to be behind the equipment to give a clean front.
I noticed the one switch says "Manufactured for Cisco" which isn't something I would expect an an actual made by Cisco switch to say. So I'm thinking they either contracted out the switch development, or they found a third party switch and put their own custom OS on it and rebranded it.
Yeah, it's interesting, I'd be interested to know what that "R" logo is for - some sort of OEM possibly? That said, those definitely aren't a white label product that has been simply rebranded, but I wouldn't be surprised if the manufacturing was outsourced to an external company. Although that's dead common for most manufacturers nowadays.
PS in CBS-250-24T-4X is almost identical as in old Netgear GS724T. I got mine recently recapped. I bet it is done by the same factories, even if brand would be diffrent. Apparently, world of network equipment power supplies is pretty small.
Lovely stuff. Like the references to "rambling" :)
Bit of a shame they’ve covered up the RPS port, I guess it’s to stop people that would need RPS from buying a lower end switch. But it’s right there!
Excellent video !!
I've been looking at the CBS350-48XT-4X but it doesn't look like it plays nice with AOCs which is frustrating as the furthest run I have is 10m+ and I'm not 100% sure a DAC will work.
I also didn't realise they don't have stackwise cables on them
Where are you seeing this about AOCs, I can't seem to find anything? Personally if I needed a fibre link between the machines, I'd get separate transceivers with a separate fibre cable. That way you can use different transceivers on each end if either device requires a specific transceivers vendor. That said, the switch has 10GBaseT ports so as long as the device on the other end can also support a 10GBaseT connection, you'd be able to just run a regular copper cable between them and get up to 100m with CAT6A or 55m of CAT6. They don't have Stackwise connections, instead they stack using SFP+ cables between the switches (either DACs or fibre) although they can't stack using the regular RJ45 ports, only the SFP+ ones.
Hey! The 250s ARE passively cooled. Neat. :D
It depends on the model across the 220, 250 and 350 series. The gigabit, non-PoE models are generally passively cooled but the larger ones with PoE or a large number of multi-gig ports have fans. The model comparison charts for each series detail how many fans each model has, e.g. www.cisco.com/c/en/us/products/switches/business-350-series-managed-switches/models-comparison.html
For smaller project installs I've been using the Cisco C9200CX-12P-2X2G switches. A typical application would be like a security camera deployment with VOIP and maybe wifi. (Think security guard stations at a business park connected to a corporate network)
They are passively cooled PoE+ switches with 12 powered gigabit ports, 2 unpowered gigabit ports, and 2 10G SFP+ ports and can deliver up to 240W. Now they do run hot but they are completely fanless with a giant heatsink block on top. (The heatsink is basically the case.) They have a full sized SD slot for hosting applications such as an embedded wireless controller and run IOS 17 Lite XE which to me doesn't seem or visibly appear any different than normal IOS XE.
The suggested retail is about $1800, but you can get them for quite a bit less if you're a corporate customer.
They look good, however as soon as you go to the higher end Cisco Catalyst switches you're needing to deal with licencing and support contracts, something that makes them unviable for many small businesses settings. They're great switches for sure, but just not practical cost wise for many environments.
old cisco devices used to tie stones inside to make them heavy
7:15 it looks like a Cisco RPS connector.
why the hell do you need industrial switches for a home network lmao there is no possible reason for that
because that's the way he likes to do things.
Nah industrial switches would be like an IE 3000 or Stratix line. These aren't even based on the normal Cisco catalyst line which is their normal business switch.
@@stonent shut up nerd
Short answer is "I don't need them" however I enjoy tinkering with these things and it also allows me to produce these CZcams videos. On the other hand, a CZcams video showing a basic wireless router and unmanaged desktop switch wouldn't exactly be interesting! Additionally, having 10GbE switching allows me to transfer large video files to and from my NAS at a significantly higher speed than I could over a basic gigabit connection.