MikroTik

This is just a crazy theoretical concept to spark a discussion. What would you guys think about a switch that is aligned like this, with ports facing upwards, taking up a 2U space. The device would be able to slide out, to acess the ports. Much more ports than on a regular 2U switch.

Image is pure concept art, nothing real.

Cabling nightmare ?

Higher port density is awesome.

Just practically though I image a lot of problems replacing flyleads or cables in the middle of the switch (like row 6 column 12 for example).

Otherwise I like it. Do it. With poe please

Do you imagine this FAT-FAT-FAT cable boundle which you try to move to slide out/in a switch?
How to organize them to ease slide device out and not to have big gnarl when slided in?
How to protect cables against braking down RJ45 connectors?
How many empty Us above device needed to organize cables?

160 cables moved at once? No gym needed

Maybe "stupid" expander which increases number of ports without need to configure next device?

Interesting concept, but I think a total cabling nightmare. Maybe a slightly different variation, have ports on the front and back, even if it's 2u that's fine. Maybe make an expander module that can be mounted on the front or back of the rack, via a fiber optic cable and connected power back to the switch. Would make it easy to have a top of rack back and front switch ports.


Sent from my SM-A520W using Tapatalk

The concept is very creative, but in a "real world" deployment it would be a nightmare to cable.

I would prefer the traditional 96port 2RU option.

I agree with the others. 48 port switches / patch panels already have very thick cable bundles, this would be a nightmare to manage cable-wise.

It would definitely need built in cable management arm at the rear that helps slide and guide the cable bundle in and out. Even then I see the high density and effort as a very niche product personally.

I...Maybe make an expander module that can be mounted on the front or back of the rack, via a fiber optic cable and connected power back to the switch. Would make it easy to have a top of rack back and front switch ports....

Just PoE powering. No need for next PSU.

I prefer this design personally:

Interesting idea. But would it be denser? It is, already, 2U in height. We will need at least another 2U - to keep it "organized" and not break the cables. So, it's a 4U switch already. I'd say 5 or 6U - it's hard to pass all this cables.

Wouldn't be easier to just make a 4U switch?

Nice idea, but no thanks. You would looks at least 1 - 2u anyway as you need space above and or below the switch - so no real point. A normal 48 port 1u or like NZ_Monkey said that Juniper is very nice. There is a much greater risk one could damage or unseat the cabling.

No thanks, for cable managment this is nightmare.

I think it would be great, but you have to provide a specular, horizontal patch panel to put over...

I'm not hating on it but wouldn't you just be better off with a design like that nz_monkey posted or lift a design like Netgears M4300-96X that can fit 96 Ports in a 2U Design ?

Interesting idea, but I totally agree about the cable management nightmare.
Something like that could possibly work mounted in the ceiling of a rack cabinet.

If you keep the idea rack mounted, I'd still keep the ports on the front and/or back, and maybe put a high-speed interconnect on the top and bottom so they could be stacked without a backplane or using available ports.

Another idea is to combine a switch and a punchdown panel and made it rack and wall mountable.
However, that would be an absolute pain to replace if it ever went bad.

BTW: I realize my ideas aren't great, but maybe it'll help someone else come up with an even better idea.

How are you going to keep dust and other foreign debris out of the jacks? If they are open to the top gravity will cause open ports to fill up and make them unusable.

Like others have said, cabling would be a nightmare.

Jerry

Nice concept, but it would be unusable in real world applications...

I get it why you would think of this...Mikrotik's switches and routers fill only a small portion of what is considered usable rack space when it comes to size.
As others have pointed out, I don't think this design would be feasible in a standard rack.
You could build your own cabinets, but that would be very difficult to integrate in existing dc.

Why not try to use a vertical blade model on a unified chassis?
2 rows of vertical blades back-to-back should offer high density and enough access to ports. A custom wire manager should be simple enough to design and integrate.

You'd have a central system and just add interfaces to the backplane.

How about 1U with 48x 1G ETH ports facing both ways with 100Gbps QSFP stacking intreface ?
This way you could get 192 ports in 2U space. To even more save sapace make it with external, rack mountable, redundant power supply.

HOLY CRAP - how about SFP version???

Imagine using trunk cables for it like those:
https://www.fs.com/products/33887.html


This wouldn't be nightmare to cable manage at all, Uniboot fiber patchcirds are incredibly thin, though it'd require to put SFP cages under like 45 deg angle to fit 2U with cables.

Also realistically it HAS TO use actual rails, not just ears to make it usable in conventional rack. If it could be moved out on rail for recabling it'd be awesome.

Also remember about vertical racks like this one:

Is port-density an actual problem today? Especielly since we're getting affordable 10/25/40GE.

I would like to see an DIN-mounted switch of some kind that fits nice in outdoor/industrial enclousures. Maybe with 6-12 ports and some combos.

Upwards facing ports always are collecting dust.
And don't tell me everyone is putting dust-caps on unused ports...

Look nice but will soon raise up problems in usage.
The best way to use it is in the ceiling of the rack, port facing down to avoid dust and ease usage.
Better to make front and back ports.

Sent from my TECNO Camon CX using Tapatalk

Better to make front and back ports.

And airflow from left to right. Hmm...

Front and back ports are actually devious idea since some types of hardware (networking gear) have front facing ports while some types of hardware (servers) have back facing ports. Getting rid of cables routing from back to front of rack would be wonderful. It'd be like cable management+switch device in one.

I dont think it is wise and personally I wouldn't buy it even if I had the need to switch so many ports.

The concept shows 224 eth ports in 2U space
Normal 2*1U switch has 2*48=96 eth ports.
Unusual, yet conventional 2U switch could have bit more: If you remove all vertical spacing, you could actually fit 3 rows of eth ports in 1U therefore 6 rows in 2U. That might give us 3*48=144 eth ports which is already pretty high density. And all that without any sliding, vertical ports or cable bending issues...

Personally, I would be much happier if we finally get CRS354-48P-4S+2Q+ which was presented almost year ago. I actually had to buy switch from different company because we needed to clean up rack and there is not a single 48 port switch in mikrotik's range. :'( largest is 24 port and that is not enough nowadays. Instead of giant leaps, let's increase the density in usual way

edit: rear ports (even if its just 4 or 8 of them) would be actually really cool!

Personally, I would be much happier if we finally get CRS354-48P-4S+2Q+ which was presented almost year ago. I actually had to buy switch from different company because we needed to clean up rack and there is not a single 48 port switch in mikrotik's range. :'( largest is 24 port and that is not enough nowadays. Instead of giant leaps, let's increase the density in usual way

Yeah I'm waiting CRS326-24S+2Q+ because current CRS317 also has pretty low port density. Not gonna even ask for 48 SFP+ ports in single switch bc I may get old before it happens...

Although higher port density would be nice, the design doesn't seem feasible to me. In my opinion you should better focus on new RB or CCR/CRS models which can be dual-mounted in 1U. I definitely see a market for such devices in cases where redundancy is required but budget and space are limited (for example on customer premises, branch offices, etc.).

A RB4011 with smaller width and two SFP+ instead of one would be perfect:
- One fiber (ISP A / ISP B) to each RB as upstream
- One fiber between both RBs as example for running iBGP
- 8-12 copper ports for attaching customers switches, etc.

I agree, SFP-ports is important, however I think all ports should be on the same side.

Standard 1U 48 ports are already a mess when cable arrangement is not managed
That layout would be a pain in the.......rack....
A front-side high density would be ok for a 3 or 4 rack units, but a lot of space wasted in depht.

Standard 1U 48 ports are already a mess when cable arrangement is not managed
That layout would be a pain in the.......rack....
A front-side high density would be ok for a 3 or 4 rack units, but a lot of space wasted in depht.

how about... vertical switch >_> like only 5cm deep so that you could mount it behind normal equipment (especially shorter ones), somewhat like giant rackmount PDU.

how about... vertical switch >_> like only 5cm deep so that you could mount it behind normal equipment (especially shorter ones), somewhat like giant rackmount PDU.

You can mount 19" equipment in front and back of a rack, the only notable problem can be air-flow if two devices share same U-position and force air towards their corresponding back sides (one against the other) ... not sure how would vertical switch make any difference in this regard?

how about... vertical switch >_> like only 5cm deep so that you could mount it behind normal equipment (especially shorter ones), somewhat like giant rackmount PDU.

You can mount 19" equipment in front and back of a rack, the only notable problem can be air-flow if two devices share same U-position and force air towards their corresponding back sides (one against the other) ... not sure how would vertical switch make any difference in this regard?

Vertical switch would technicality use 0U since it could be mounted behind normal equipment just like PDUs are usually mount behind servers in the same U row. The only downside would be airflow obstruction since it'd realistically put metal wall behind equipment exhaust.

I'm not sure if I understood correctly but if you meant mounting regular equipment on both sides of the same U (so that they kinda blow exhaust air to the middle of rack) then it may sound doable with short equipment like routers or switches but long equipment like 1000mm servers usually don't leave enough space behind them to mount anything deeper than 10cm. Vertical switch could have 0cm depth if mount on the outside or around 4-5cm if mounted on the inside of rail.

It could be a bit problematic to mount it though since long equipment uses 4-point mounted rails and you can't really put one ears on top of another... But it could be mounted in different way. Or maybe bundle thinner ears which actually could be put on top of rear server rail ears.

Airflow obstruction wise it'd be equivalent of having 2U or 3U PDU behind server and you could use *all* U slots for long servers and still have switch in rack.

You're right, mounting two regular equipment pieces in same U-position is only possible for short equipment and that's what I've had in mind.
But then I'd never mount just anything behind full server chasis which could obstruct warm air exhaust ... 1-U server can easily consume 500W+ (and generate just as much heat) and I wouldn't like to dampen the air-flow cooling that owen.

You're right, mounting two regular equipment pieces in same U-position is only possible for short equipment and that's what I've had in mind.
But then I'd never mount just anything behind full server chasis which could obstruct warm air exhaust ... 1-U server can easily consume 500W+ (and generate just as much heat) and I wouldn't like to dampen the air-flow cooling that owen.

True, but from experience I can tell you that servers have insanely overkill cooling for extreme worst case scenarios. Such obstruction probably wouldn't change much (especially since you probably don't need to cover half of back door with switches, just 1 or 2 2U switches (I estimate CRS326 depth to a bit above 2U height)). I regularly replace those 15k rpm blowers with silent 5k fans that have significantly lower airflow and they keep up. It's just that in servers world 40 degree in chassis means "overheating". Whereas gaming desktop computers with noise optimized cooling usually have 50-60 deg ambient in chassis. You can compare it to CRS317 vs CRS326 - 317 enables fans above 40deg while 326 being passively cooled is totally fine with 70+ deg on CPU. It's not that 317 has lower tolerance, it's just configured this way following "better safe than sorry" rule.

My friend once witnessed complete malfunction of AC in one datacenter. AC was not functioning at all for around 8h. Temperature in room was so extreme that it was dangerous for human to spend more than few minutes in here, even though all doors were open and giant fans were brought to pump air through doors.

All servers survived without any damage. So I really doubt covering 2U on the back of rack would do anything. After all PDUs are commonly mount behind servers.

how about... vertical switch >_> like only 5cm deep so that you could mount it behind normal equipment (especially shorter ones), somewhat like giant rackmount PDU.

Not a bad idea, but if mounted IN FRONT of other equipment.
Cable management must be in front side of rack to avoid headaches when maintaning patch cord connections

Unless the unit is intended for RIGID connections like a datacenter/server farm where all patch cords are installed once and moved no more like this:

dont think this concept will work.
what i would like to see is a "port expander":
a Master Switch with all the intelligence and 1 or 2 Expand-Ports
a Expander with 24 or 48 Ports without intelligence

Just expand the port count without the need to manage another switch (nexus like)

dont think this concept will work.
what i would like to see is a "port expander":
a Master Switch with all the intelligence and 1 or 2 Expand-Ports
a Expander with 24 or 48 Ports without intelligence

Just expand the port count without the need to manage another switch (nexus like)

+1 but it'd probably need implementation of switch clustering in the first place. So I guess it'd be hard to make with current ROS capabilities.

Not a bad idea, but if mounted IN FRONT of other equipment.
Cable management must be in front side of rack to avoid headaches when maintaning patch cord connections

I though of rear mounting since we were talking about environment with long servers and servers always have rear facing network cards. For example at work we have racks where all cables are in back or rack and in the end switch ia the only device with front facing ports and it's always pain in the ass since we always struggle to trace cables to back of rack.

What I would like to have from MikroTik is an standard 1U switch with more ports: 48x 10GbE, 6x 100GbE, dual power supplies. Therefore the Maxwell Prestera EX chipset (98EX54xx) could be used.

So, this would be a CRS354-48S+6Q28+

At this point, just add the ports to the back of the switch. You'll get 96 10g ports + 6-12x 100g uplinks All in a 1U formfactor, then if you wanted to get fancy, i'm sure you could come up with a way to get 5 or even 6 port height in a combined 2U space. which would be up to 144 ports on the front plus another hundred on the back, etc...

That being said, mikrotik excels at having the weird solution no one else does(at a good price). so this may just be the solution for the guy who needs it.

I had suggested in a previous similar post to build a chassis with blades. Could be switch blades, routing blades, whatever port configuration/speed. Could even have a fan blade, just in case. Hot swappable power supplies.

One blade is old? A faster one comes out? No problem, swap it.

Just make a fast backplane so the chassis can survive the tests of time and voila!

Cheers,

Sent from my cell phone. Sorry for the errors.

I had suggested in a previous similar post to build a chassis with blades. Could be switch blades, routing blades, whatever port configuration/speed. Could even have a fan blade, just in case. Hot swappable power supplies.

One blade is old? A faster one comes out? No problem, swap it.

That would add one additional "0" to price. I saw modular hardware vs non-modular equivalent prices difference for other vendors. It's still hard for me to understand why would anyone buy modular hardware...

I had suggested in a previous similar post to build a chassis with blades. Could be switch blades, routing blades, whatever port configuration/speed. Could even have a fan blade, just in case. Hot swappable power supplies.

One blade is old? A faster one comes out? No problem, swap it.

That would add one additional "0" to price. I saw modular hardware vs non-modular equivalent prices difference for other vendors. It's still hard for me to understand why would anyone buy modular hardware...

Agreed, but it depends on clients and needs. Cisco versus more extensive MikroTik? Some might prefer the 2nd choice.

As I've said, blades could easily be replaced for a quick upgrade without having to rip the rest out.

Other idea I just had: stackable switches that get managed as a single device. Could be a good compromise.

Cheers,

Sent from my cell phone. Sorry for the errors.

Other idea I just had: stackable switches that get managed as a single device. Could be a good compromise.

From what I understand actual stacking is work-in-progress. Iirc someone mentioned after MUM that QSFP+ ports in CRS326-24S+2Q+RM are supposed to be used for more advanced stacking/clustering (not just uplinks)

note way. Ports would be filled with dust etc and become useless.

way better to have a dual stack switch in the normal orientation but with a second switch right behind the first in the same U with just enough gap and cable management.

I would also prefer stack-able switch unit before so highly concentrated array of connectors. For example stacking ability to ten units would do very good job - with not so deep case, two units can be mounted in the same 1U space in "back-to-back" setup, if someone need to fill the rack effectively.

Personally, I would be much happier if we finally get CRS354-48P-4S+2Q+ which was presented almost year ago. I actually had to buy switch from different company because we needed to clean up rack and there is not a single 48 port switch in mikrotik's range. :'( largest is 24 port and that is not enough nowadays. Instead of giant leaps, let's increase the density in usual way

They are on the site.

It’s a fun idea, but I think I’d rather see a 2u front facing high density. Oh and stacking or expander functionality in ROS!

If you are looking at density on the switch, why not do something like the old amphenol telephone connectors. Then either use a breakout patch panel and run the Amphenol cable between the switch and breakout/patch panel?

If you're looking at 10/25g ethernet, is something like an MPO fiber connector and option?

Almost year and a half from the first post in this topic... Is there any news about multiport switch, maybe something like old Cisco 6500? With modular slots?

If you really have switch chip with physical QSFP+ ports (interfaces) 384.. Then divide them on a plate of 8 QSFP+ with different Wavelength and combine into one = Switch must have a physical 48 ports all ready Multiplexing LC connectors! On other side IF you need split physical Wavelength use your product CWDM-MUX8A
I understand that the switch (inside) now requires SFP chips and optical cables, BUT no more cable nightmares! Cable nightmares should now remain on the other side

Small DIN switches.. Wide input DC power. It'd be nice to be able to mount small switches on walls in closets, cabinets, backboards, industrial situations, etc... I'm not into zip tying mikrotiks down, or photocopying them to make screw hole templates with little screws that don't work nicely and probably won't be ideally aligned. Include a 2" piece of din rail, even plastic rail, and it would be awesome. Even those HapAC2 with the small plastic plate, make a DIN plate that replaces the clear plastic piece. If it's not rackmount or DIN or have mounting ears it's hard to physically install for non-home-use.

Small DIN switches.. Wide input DC power. It'd be nice to be able to mount small switches on walls in closets, cabinets, backboards, industrial situations, etc...

+1

What about 3 rows in 1U?
https://www.youtube.com/watch?v=5kILCRsachk

Small DIN switches.. Wide input DC power. It'd be nice to be able to mount small switches on walls in closets, cabinets, backboards, industrial situations, etc...

Good idea!

Also something like what Nexans is doing with their Microswitches for FTTO.
I've used them in a few buldings and while the idea and the hardware is great, their management software (nexman) is kind of mediocre.

I've used them in a few buldings and while the idea and the hardware is great, their management software (nexman) is kind of mediocre.

Many industrial and DIN-mounted stuff is horrible from a software and management perspective. The are mostly built to last and not be flexible. And you can't build big network with just plain switches.

I just saw a proposed triple RJ45 connector from 2012.

35.20mm high.

1U height: 44.45mm

Can you guy's make a rackmount version of the netpower 16P ? ans also a DC version of the 328P?

This is just a crazy theoretical concept to spark a discussion. What would you guys think about a switch that is aligned like this, with ports facing upwards, taking up a 2U space. The device would be able to slide out, to acess the ports. Much more ports than on a regular 2U switch.

Image is pure concept art, nothing real.

I think it would be better if switches would be sideways and in a tray: you pull out the tray and "easily" work on the cablin. This would require some extra decimeter of cabling and a cable holder inside the tray. It would still be a bigger hassle compared to traditional switches.
What I don't "like" about traditional switches is that I feel I'm wasting space in the rack unit, since switches are small. When I can I put something on the other side (front or back) of the rack unit, like a small UPS, but I can do that only when I can access both front and back of the rack.

I do not like it

I don't think this would work with current rack designs but if the entire rack design was to change I think it could work. I believe some of the hyper scalers are already redesigning racks to be wider etc so there really is no reason that an entire new design for super high density couldn't be a thing with enough buy in from others.

A better option (for current racks) might be splitter cables (I think these already existing and are called break out cables) where you can take existing cables and combine them into one large capacity cable (I think this is how GPoN works, one cable with multiple wave lengths).

May be add some mtp 24 ports for use with this splitters?

May be add some mtp 24 ports for use with this splitters?

b05d375438b16cb7186bd0ad26c12c17.image.500x500.jpg

Sadly, breakouts of more than 4 cables/ports is only supported starting with 400G ports. 400GBASE-SR16 (802.3bs), 400GBASE-FR8 (802.3bs), 400GBASE-LR8 (802.3bs), and 400GBASE-ER8 (802.3cn) being the relevant standards.

What could hoverever be done is a solution along the lines of a 802.1BR port extender.
Imagine a PoE++ switch (802.3bt) with 10GBASE-T ports running a 802.1BR Controller Bridge, and Breakout-Cables (1x 10GBASE-T expanding to 8-10x 1GBASE-T) with a simple switch chip running 802.1BR port extender. Power supply could happen via PoE-PD, possibly with PoE-passthrough.
Could also work with copper-to-fibre or fibre-to-fibre.
Should be able to achieve quite a bit of port density with that. No idea how feasable it is from a financial point of view, though...

Power supply could happen via PoE-PD, possibly with PoE-passthrough.
Could also work with ... fibre-to-fibre.

Took slightly out of context on purpose ... but how would PoE happen in this case? ;-)

Power supply could happen via PoE-PD, possibly with PoE-passthrough.
Could also work with ... fibre-to-fibre.

Took slightly out of context on purpose ... but how would PoE happen in this case? ;-)

The "port expander adapter", with its 803.1br switch chip, will need some power, being an active cable. Since you'd want 10GBASE-T on the side facing the switch, you're bound to 802.3bt-PD (power delivery) for the port to be usable as "PoE-in".
The optional PoE passthrough could then the handled passively, or actively (802.3at or 802.3bt). We already see that happening in several routerboards.
Depending on whether enough power can be supplied via the . Probably not, if one were to look for maximum port density.

[*] Controller Bridge Switch with 48x 10GBASE-T PoE++ (802.3bt) & 2x QSFP56 (200G) Uplinks in 1U chassis.
[*] Connect each 10GBASE-T port to one "port expander adapter"
[*] Assuming 1-to-8 adapters, you would supply 48*8 = 384 1GBASE-T Ports from 1U
If no PoE passthrough is needed, they should be able to get that done with less than 1000W PoE budget on the Controller Bridge Switch.

Example for a 1-to-8 port expander adapter: Obviously, instead of RJ45 "output", you could maybe go with direct fibre pigtails as well. Then again, we're only starting to see fibre connected directly to the switch chip in the first high end devices, so it may be quite a while until it's available in this form factor, or another.

By the way: EARS cause a lot of problems when mounting, demounting from rack, full of 1U servers and cables.
Especially, if you have some switches on top of the rack and you have to place under other switches.
Would be cool if first you mount light rails and then you mount a switch without ears inside rails.

Small DIN switches.. Wide input DC power. It'd be nice to be able to mount small switches on walls in closets, cabinets, backboards, industrial situations, etc...

+1
Would be a nice option!

I would like to see a DC powered -48 volt ( telco power neg 48 volts ) and or a switch that can do either -48 or +48 volts.

something with 12 to 24 POE out ports
something with some 10-Gig SFP ports ( 2 of 4 sfp+ ports )

Although Mikrotik does have some switches/stuff that can run on 48 volts , none of them can be connected to a telco neg 48-volt battery power system :(

--- also --- to compliment a plus and/or minus 48 volt switch ...
- An optional outdoor enclosure ( mid-span or tower mounted enclosure )
- An optional a utility power AC ( 110 volt AC ) to 48 volts which can run the 48 volt ( +48 or -48 volts ) and also charge a 48 volt battery bank.
* all 48 volt devices should include an on-board 24-volt supply
* all DC powered device should never have the negative leads or positive leads grounded. The power should be isolated and a ground lug should be used to reference ground if wanted.

--- also - not wanted :(
switches/devices that do not require two DC power wall-wart power supplies ( a 48 volt and a 24 volt power supply --- both at the same time :(
48-volt DC powered switches that are internally negative grounded.

FYI - connecting a Mikrotik device that is internally negative 48 volts grounded to a telco 48 volt positive ground battery power system will result in smoke and fire or blown fuses if any thing touches any ground ( such as a tower ).

I think the concept has merit but I would worry about cooling. The application of this high density switch worries me due to a single point of failure.

I think the concept has merit but I would worry about cooling. The application of this high density switch worries me due to a single point of failure.

Re cooling:
In my opinion , a router/switch with an internal power supply will always run warmer/hotter than an identical device with an external DC power supply.

Re single point of failure:
With external power supplies. The motherboard still has the same number of connectors from an internal power supply verses the same motherboard with external DC power supplies. Also , with external power supplies , it's pretty easy to connect up n+1 power supply redundancy.

I think the concept has merit but I would worry about cooling. The application of this high density switch worries me due to a single point of failure.

Re cooling:
In my opinion , a router/switch with an internal power supply will always run warmer/hotter than an identical device with an external DC power supply.

Why? It is quite easy to put a stronger exhausting fan, and use the PSU as part of the ventilation system. If You are worried about wear and tear of the PSU fan, just make it fanless, and use a shroud to guide the air from the discrete internal fans. Just like is done with CPU coolers on 1U chassis.

I think the concept has merit but I would worry about cooling. The application of this high density switch worries me due to a single point of failure.

Re cooling:
In my opinion , a router/switch with an internal power supply will always run warmer/hotter than an identical device with an external DC power supply.

Why? It is quite easy to put a stronger exhausting fan, and use the PSU as part of the ventilation system. If You are worried about wear and tear of the PSU fan, just make it fanless, and use a shroud to guide the air from the discrete internal fans. Just like is done with CPU coolers on 1U chassis.

Well - yes --- it is always possible to do some fan modification/changes to get the type of cooling you desire.
For me, the issue is not cooling.
For me, the issue is getting up to or greater than 7 days of backup battery run time when the power goes out.
Typically a 110 VAC UPS powering a some Mikrotik APs and a Mikrotik switch on a 110 VAC UPS ends up with the UPS consuming more power from the batteries just to run the UPS itself. What is the cost for an external 110 VAC UPS and additional batteries that is able to run half-a-dozen Mikrotik devices on battery power for about 1-week without utility power ?
By staying DC you don't have the inefficient battery-to-AC-inverter to get 110 VAC then other 110 VAC power supplies to get back to board level DC power.

Well - yes --- it is always possible to do some fan modification/changes to get the type of cooling you desire.
For me, the issue is not cooling.
For me, the issue is getting up to or greater than 7 days of backup battery run time when the power goes out.
Typically a 110 VAC UPS powering a some Mikrotik APs and a Mikrotik switch on a 110 VAC UPS ends up with the UPS consuming more power from the batteries just to run the UPS itself. What is the cost for an external 110 VAC UPS and additional batteries that is able to run half-a-dozen Mikrotik devices on battery power for about 1-week without utility power ?
By staying DC you don't have the inefficient battery-to-AC-inverter to get 110 VAC then other 110 VAC power supplies to get back to board level DC power.

But You talked about cooling!
"In my opinion , a router/switch with an internal power supply will always run warmer/hotter than an identical device with an external DC power supply. "

Yes, I agree with you: it would be great to be able to power it directly from DC. I think Mikrotik should do a DC-DC PSU, just so we could use it with more devices.

Well - yes --- it is always possible to do some fan modification/changes to get the type of cooling you desire.
For me, the issue is not cooling.
For me, the issue is getting up to or greater than 7 days of backup battery run time when the power goes out.
Typically a 110 VAC UPS powering a some Mikrotik APs and a Mikrotik switch on a 110 VAC UPS ends up with the UPS consuming more power from the batteries just to run the UPS itself. What is the cost for an external 110 VAC UPS and additional batteries that is able to run half-a-dozen Mikrotik devices on battery power for about 1-week without utility power ?
By staying DC you don't have the inefficient battery-to-AC-inverter to get 110 VAC then other 110 VAC power supplies to get back to board level DC power.

But You talked about cooling!
"In my opinion , a router/switch with an internal power supply will always run warmer/hotter than an identical device with an external DC power supply. "

Yes, I agree with you: it would be great to be able to power it directly from DC. I think Mikrotik should do a DC-DC PSU, just so we could use it with more devices.

Mikrotik does have some optional DC replacement power supplies that are Mikrotik made to replace a Mikrotik 110 VAC power supply. However , those power supplies are not able to power a 24 port or more all PoE switch. And the power supplies for something like their 48-port PoE switch & 4 sfp+ & 2Q ports switch has two fans dedicated just to cool the AC power supply and another 2 fans to cool the motherboard.

I assume you have seen some of my recent posts on the DC mods I've done on the 24-port-PoE and their 48-port-PoE devices ( convert from 110 VAC to 24-VDC. I know they run cooler now because their is now no longer any internal AC power supplies - but I did keep the power supply fans connected just because they were already there.

Mikrotik does have some optional DC replacement power supplies that are Mikrotik made to replace a Mikrotik 110 VAC power supply. However , those power supplies are not able to power a 24 port or more all PoE switch. And the power supplies for something like their 48-port PoE switch & 4 sfp+ & 2Q ports switch has two fans dedicated just to cool the AC power supply and another 2 fans to cool the motherboard.

I assume you have seen some of my recent posts on the DC mods I've done on the 24-port-PoE and their 48-port-PoE devices ( convert from 110 VAC to 24-VDC. I know they run cooler now because their is now no longer any internal AC power supplies - but I did keep the power supply fans connected just because they were already there.

But these PSU can't be used on all models with an internal PSU! One thing I think they should do is exactly the same thing the ATX standard did for PC PSUs: modularity. Even if they did one standard to hot swapabble and another to "normal" ones. This would be great.

About the CRS354-48P-4S+2Q+RM: It is rated to 800W PoE! Even with a PSU rated to 95% efficiency, we are talking about 40W of waste heat! No wonder they put two fans in it. Even the CRS328 can push 500W on PoE - at about 95% efficiency this is about 25W of waste heat.

About your mod: Yes, I read it. Don't remember the details now. It may very well run cooler now - it all depends on the internal design. If the PSU is open (fully exposed on the inside of the chassis), then removing it will make a lot of difference. If the PSU is closed (like an ATX computer one), then it shouldn't make much difference - since the hot air would be blowed outside.

For ultra high port density may be go for co packaged optics ? Integrate optics into the switch.



Make it modular: 1U chassis and small modules to select port speed / transceiver type .

I would like to see a CRS326-24G-2S+RM with 4 SPF+ or 2 QSFP+ cages. This would close the gap between CRS326-24G-2S+RM and CRS326-24S+2Q+RM.

So, Normis, what is the conclusion?
Where have you moved to?

Although higher port density would be nice, the design doesn't seem feasible to me. In my opinion you should better focus on new RB or CCR/CRS models which can be dual-mounted in 1U. I definitely see a market for such devices in cases where redundancy is required but budget and space are limited (for example on customer premises, branch offices, etc.).

A RB4011 with smaller width and two SFP+ instead of one would be perfect:
- One fiber (ISP A / ISP B) to each RB as upstream
- One fiber between both RBs as example for running iBGP
- 8-12 copper ports for attaching customers switches, etc.

https://mikrotik.com/product/rb5009ug_s_in
Not dual SFP+, but SFP+ + 2.5G ethernet

I would like to see a CRS326-24G-2S+RM with 4 SPF+ or 2 QSFP+ cages. This would close the gap between CRS326-24G-2S+RM and CRS326-24S+2Q+RM.

That would be a CRS328-24G-4S+RM

If Mikrotik would build this, just ship them to Baltic Networks and take our money!

What would you guys think about a switch that is aligned like this, with ports facing upwards, taking up a 2U space.

It's creative. But not sure how useful it be have looking to actually mount it a "rack"... e.g. "able to slide out to acess the ports" as that seems like a cabling nightmare.

Where I think this concept might be super useful is if you had plywood wall to mount it on, and had NO rack – thinking old phone room as modern day "66 block" for ethernet, and avoid actually needing a rack for ethernet switching. The "grid" of ports might be useful to help organize stuff into floors or areas visually on the switch (vs. 2+ rows of 24+ ports) too. As other have suggested, if it come with cable management, perhaps "around the outside", certainly be more useful.

Now "stacking" RB5009s VERITICALLY with some new/different mounts, achieve the same density today. So kinda agree with @schadom adding an 2nd SFP to medium/smaller switch let you daisy-chain your way to density in odd places.

In my opinion you should better focus on new RB or CCR/CRS models which can be dual-mounted in 1U. [... with] two SFP+ instead of one would be perfect

If you're looking to get density INSIDE a rack, perhaps a "double-end" switch with ethernet ports on "front" and "back" (power in middle, accessed from side?). That might get more density-per-U and solve an actually issue I run into that some rack equipment use "front" ethernet ports, while others are on "back". A switch that use both sides might be more useful that trying to adapt a rack to "top mounted" ethernet ports...

Some sweet pics! Free beer and pizza if anyone wants to help me wire up my Patch Panel. Airline tickets not included!

How would you access the ports? Do you plan to make a special insertion and removal tool? https://www.youtube.com/watch?v=H1TIVh3VD4Q
Whatever you do, put activity lights near where the cable plugs in. Having indicator leds (multi color for diffent speeds) is a very useful troubleshooting tool, as you can quickly see if the cable you plugged in is active. Remember where these are going to be installed, not in your living room, so wife factor is less important than functionality. TP-Link got this right on the TL-SG108E but screwed up on the TL-2008v3 (lights are on opposite edge from cables). The CSS106-5G-1S and hEX got it better than opposite edge, but still the top edge on opposite side makes it hard to see if these are on a shelf or stacked. The Ubiquiti ER-X has the same problem. The RB5009 got this right.
And isn't a console port a worth while thing any more? I am not interested in bluetooth access.

What about 3 rows in 1U?
https://www.youtube.com/watch?v=5kILCRsachk

Or at least 5 rows in 2U
120 ports are not bad for 2U
I suppose 19" can fits 30 ports on a single row (total 150 ports in 2U).

Its a dust trap! A problem waiting to happen. I would not use one, What I need are SFP and SFP+ ports. A 100Gig with 10 SFP+ ports would be of interest to me all my customers are remote and fiber is the only option for me. Due to heavy trees and mountainous terrain fiber is the only option and I install 1 Gig connections every time but only give 1 Gig if customer is willing to pay for it.

My thought on this

Richard

I could sure use at lease four 16-port 100-Gig Mikrotik switches ( or even better 16 100-Gig ports and 16 10-gig ports ).

North Idaho Tom Jones

In the next few years we will be wanting 200 Gbit/s QSFP56 then a few more and it will be 400 Gbit/s QSFP-DD. The demand for bigger and bigger links is growing.

In the next few years we will be wanting 200 Gbit/s QSFP56 then a few more and it will be 400 Gbit/s QSFP-DD. The demand for bigger and bigger links is growing.

Today , I need/want qty 4 to 8 100-Gig many port switches ( 2 and 4 port 100-Gig switches do not even come close to what I now need ).
We have a few dozen servers and a few dozen switches that need to be upgraded or replaced. We have grown way beyond 10-Gig networking and from now on we will only order servers & switches for our NOCs with 100-Gig ports.

North Idaho Tom Jones

Hello Guys,

I recently did a deep dive in all Mikrotik Switches and compilations I could use for the upcoming new data-centers.
Probably not the standard use, but some more switches should be useful for other situations.

I have some suggestions for other Switch devices with more and other higher bandwidth ports:

CRS528-24XG-4Q+-RM

• 24x 10G Ethernet ports

• 4x 40G qsfp+ ports

• dual hot-swap power supplies

CRS514-12Q+-2XQ-RM

• 12x 40G qsfp+ ports

• 2x 100G qftp28 ports

• dual hot-swap power supplies

CRS512-12XQ-RM

• 12x 100G qftp28 ports

• dual hot-swap power supplies

These Switches would be very nice to have for a bunch of different setups and would make other 100G Mikrotik products so much more useful.