I am writing this post to urge MikroTik to consider producing high-performance, production-grade Layer 3 switches. While MikroTik has somewhat capable Layer 2 switches, its Layer 3 switching capabilities are next to nil.

Layer 3 switching has been the de-facto standard in data centres for use-cases such as VXLAN/EVPN and for service providers i.e. MPLS/EVPN for more than a decade. These technologies allow data centres to scale to large numbers of devices and to isolate traffic between different tenants straight in the access layer using the spine-leaf architecture.

Unfortunately, MikroTik's current 'Layer 3' switches are not well-suited for these applications. They have limited routing table capacity, no hardware support for VXLAN/EVPN or MPLS, and relatively low performance. Therefore, data centre operators have no choice but to use Layer 3 switches from other vendors, such as Juniper and Cisco.

For an example point of reference, here's a switch that many DC operations and SPs around the around use for L3 switching, MPLS, VXLAN, EVPN at line-rate—The Juniper QFX5100:
https://www.juniper.net/content/dam/www ... asheet.pdf

This layer 3 switch supports a massive routing table, perfect for L3 switching and MPLS at scale:
• IPv4 unicast routes: 128,000 prefixes; 208,000 host routes; 64^4 ECMP paths
• IPv4 multicast routes: 104,000
• IPv6 multicast routes: 52,000
• IPv6 unicast routes: 64,000 prefixes

In comparison to MikroTik's current flagship switch CRS518-16XS-2XQ-RM, this speaks for itself, there is no Layer 3 switching benchmark:
https://mikrotik.com/product/crs518_16x ... estresults

The CRS518-16XS-2XQ-RM is not usable for production-grade layer 3 switching (even if MikroTik ROSv7 supported EVPN for MPLS/VXLAN).

Normis himself made it clear that MikroTik doesn't support layer 3 switching for scale:
viewtopic.php?p=1024626#p1024629


Even worse are engineers or experts who somehow have yet to figure out that layer 3 switching is de-facto implementation for spine-leaf architectures, VXLAN/EVPN, MPLS/EVPN and yes, even Telco-MPLS rings:
viewtopic.php?p=1024626#p1024630

It's a switch, so why should it compute routes (thousands of them)? How does "thousands of routes" relate to "distribution switch in smaller networks"? In my world it doesn't, in my world "distribution switch in smaller network", even if used as L3, switch, is about a few (less than 10) subnets with tens of hosts active in each subnet.

A high-performance Layer 3 switch from MikroTik would be a boon to the data centre and SP community. It would give operators more choice and would help to reduce the OPEX and initial CAPEX.

So bottom-line question as the tile suggests:
Why are there no production-grade L3 switches from MikroTik in 2023?

You do realize that your question, since asked in an user forum, is more or less a rhetorical one? You're asking about MT's business plans and that question really should be directed towards MT's marketing department directly, ideally backed up by a large past business and prospect of huge future business (if your wish becomes true).

And then there's space for discussion about your expectations about low cost of such device (just because it would be made by Mikrotik), I'm not such an optimist (because quite possibly there is no mass-market switch SoC supporting wanted functionality which is a prerequisite for low price of end product).

You do realize that your question, since asked in an user forum, is more or less a rhetorical one? You're asking about MT's business plans and that question really should be directed towards MT's marketing department directly, ideally backed up by a large past business and prospect of huge future business (if your wish becomes true).

I do not have an NDA/Contract with MikroTik for R&D, therefore, the best I can do is post on their public forum.

And then there's space for discussion about your expectations about low cost of such device (just because it would be made by Mikrotik), I'm not such an optimist (because quite possibly there is no mass-market switch SoC supporting wanted functionality which is a prerequisite for low price of end product).

Nobody is asking for $100 Layer 3 switching with MPLS/EVPN/VXLAN at line-rate from MikroTik. MikroTik's Layer 3 switching boxes can be half the price and performance of a Juniper QFX5100 - So a realistic pricing of $3000-$10000 is doable, depending on the specs/capabilities of the box from MikroTik.

Can you make a business case for a production-grade L3 switch ?

If you could make that BC [at a high level] then I suspect that might tweek Tik interetest ... Although I do suspect that they already have such BC and have decided that its not worth the investment.

Do you have any idea whatsoever the kind of gross margins manufacturing have to make ? I have been in that game and its very significant ...

Can you make a business case for a production-grade L3 switch ?

If you could make that BC [at a high level] then I suspect that might tweek Tik interetest ... Although I do suspect that they already have such BC and have decided that its not worth the investment.

Do you have any idea whatsoever the kind of gross margins manufacturing have to make ? I have been in that game and its very significant ...

I already mentioned the business use-cases, EVPN/VXLAN for data centre ops, MPLS/EVPN for Telco/ISP ops. Right now, what I need professionally? Layer 3 switches to build an MPLS transport ring, I do not want/need to use Juniper ACX or similar devices for MPLS transport that's largely stateless/minimal light-touch QoS and traffic engineering. What are my cheapest possible options? Juniper QFX5100 to my knowledge. What I want? A MikroTik option that does not exceed $10000 fresh off the shelf. Matter of fact, I can buy a Juniper MX204 refurbished for $10k, so for layer 3 switching? Tik can surely do $3k-$5k.

For gross margins on manufacturing, not my area of expertise, but I am willing to pay MikroTik thousands of dollars if they get their shit together.

In simple English, MikroTik boxes whether it's an L3 switch or plain router, can be 1/2 or 1/3 the performance of some Juniper boxes for 1/2, 1/3 or 1/4 the retail price – Example? CCR routers.

Can you make a business case for a production-grade L3 switch ?

If you could make that BC [at a high level] then I suspect that might tweek Tik interetest ... Although I do suspect that they already have such BC and have decided that its not worth the investment.

Do you have any idea whatsoever the kind of gross margins manufacturing have to make ? I have been in that game and its very significant ...

I already mentioned the business use-cases, EVPN/VXLAN for data centre ops, MPLS/EVPN for Telco/ISP ops. Right now, what I need professionally? Layer 3 switches to build an MPLS transport ring, I do not want/need to use Juniper ACX or similar devices for MPLS transport that's largely stateless/minimal light-touch QoS and traffic engineering. What are my cheapest possible options? Juniper QFX5100 to my knowledge. What I want? A MikroTik option that does not exceed $10000 fresh off the shelf. Matter of fact, I can buy a Juniper MX204 refurbished for $10k, so for layer 3 switching? Tik can surely do $3k-$5k.

For gross margins on manufacturing, not my area of expertise, but I am willing to pay MikroTik thousands of dollars if they get their shit together.

In simple English, MikroTik boxes whether it's an L3 switch or plain router, can be 1/2 or 1/3 the performance of some Juniper boxes for 1/2, 1/3 or 1/4 the retail price – Example? CCR routers.

How does a CCR2116 or CCR2216 not achieve what you want. L3 offload on all of the newer Tiks is pretty simple. If the input and output ports both have L3 offload enabled, and it can be routed statelessly, it can be offloaded. Doesn't really matter if it's MPLS, BGP, etc... configuring those routes.

For the few things you need to add a sprinkle of QoS and such on, these two devices have more than enough CPU to handle it.

I am writing this post to urge MikroTik to consider producing high-performance, production-grade Layer 3 switches. While MikroTik has somewhat capable Layer 2 switches, its Layer 3 switching capabilities are next to nil.

if you see that as an easy task then you are missing a good opportunity, raise some capital and startup your idea

i will be expecting that equipment built by your startup to revolutionize the market

You have the knowledge and the opportunity so go for it

The CRS1xx/2xx were in my opinion a flop due to the poorly thought out workflow, since then Mikrotik have been on a journey of discovery with the CRS3xx/5xx and CCR2xxx. They are doing really well, especially for the price point...

There are some areas where they could use improvement:

- Re-write of the legacy switch ACL's. Current ones are very limited, need to be able to match/rewrite VLAN tags including inner tags... Needs counter support and multiple action support. I could go on...

- MPLS/VPLS Push/Pop/Swap in Hardware

- VXLAN in Hardware

- Larger TCAM to support multiple full tables (on the CCR2xxx series)

Some of these would come for free as they can be done in software on the existing hardware, but others such as a larger TCAM come at a cost.

How does a CCR2116 or CCR2216 not achieve what you want. L3 offload on all of the newer Tiks is pretty simple. If the input and output ports both have L3 offload enabled, and it can be routed statelessly, it can be offloaded. Doesn't really matter if it's MPLS, BGP, etc... configuring those routes.

For the few things you need to add a sprinkle of QoS and such on, these two devices have more than enough CPU to handle it.

Did you bother to read the datasheet below?
https://www.juniper.net/content/dam/www ... asheet.pdf

Which box from MikroTik has 30+ ports with all the L3 offload stuff?

if you see that as an easy task then you are missing a good opportunity, raise some capital and startup your idea

i will be expecting that equipment built by your startup to revolutionize the market

You have the knowledge and the opportunity so go for it

Clearly this is sarcasm, not sure who's falling for this one.

Some of these would come for free as they can be done in software on the existing hardware, but others such as a larger TCAM come at a cost.

Many people are willing to pay that cost, though, is the point. A MikroTik box won't cost me $400k.

Some of these would come for free as they can be done in software on the existing hardware, but others such as a larger TCAM come at a cost.

Many people are willing to pay that cost, though, is the point. A MikroTik box won't cost me $400k.

I am one of those that is prepared to pay.

For gross margins on manufacturing, not my area of expertise, but I am willing to pay MikroTik thousands of dollars if they get their shit together.

The business - use case is one very obviouse and important consideration that you have articulated clearly --- the deal maker for any manufactures is the achievable sustainable profit margine and for that one needs to understand the ROI based on current/future market needs and potential ... that is wher the business case is the deal maker.

I suspect that if there is an opportunity in this specific area Tik would have already made the B case --- which is why so far they are not in this game for this specific class of hardware/asic/software/support equation.

How does a CCR2116 or CCR2216 not achieve what you want. L3 offload on all of the newer Tiks is pretty simple. If the input and output ports both have L3 offload enabled, and it can be routed statelessly, it can be offloaded. Doesn't really matter if it's MPLS, BGP, etc... configuring those routes.

For the few things you need to add a sprinkle of QoS and such on, these two devices have more than enough CPU to handle it.

Did you bother to read the datasheet below?
https://www.juniper.net/content/dam/www ... asheet.pdf

Which box from MikroTik has 30+ ports with all the L3 offload stuff?

CRS354, if gigabit on most ports is enough for you... /s

98DX85XX chips are used in the CRS518. Those could be made into a 48 port 10Gbps switch with all the bells and whistles. Or more realistically a 40 port 10gbps switch with two QSFP+ cages. This is the best, most capable, switch chip in the DX line. If you want it to do 25gbps, max port count drops to 24 ports. Hence why on the CRS518 you see 16x SFP28 + 2QSFP28 ports. Thats the most that switch chip can offer.

To do more, you need to jump to the Marvell CX line of switch chips. There is only one vendor i know of using these, they are out of shenzhen running sonic on them, otherwise they just aren't super available yet. We might see these find more prevalence with time, and im sure Mikrotik knows of their existence. However, that's going to end up being a vastly more expensive product.

The reality is, there's a gap in the switch chip market. And a $2-$3K 30+ port SFP28 switch with layer 3 features can't be made, the chips don't exist at that price point. A 30+ SFP+ port switch? Sure, tik could make that tomorrow using the chips they already have in the CRS518. Beyond that, they would need to move to the CX line of chips, which (just spitballing) probably puts it at $6K - $10K price tag.

Would i pay $6K - $10K for a MikroTik switch, that's the real question. Right now, no i wouldn't.

Add EVPN support to the BGP daemon, rework ACL's (or ideally make them dynamically generated based on filter rules, big undertaking i know), add hardware VTEP support, etc... to RouterOS and then maybe i'd consider it. EVPN and VXLAN VTEP at the very least are must have features for me to consider that price tag. Otherwise i'd just MLAG some cheaper switches and do layer 3 on a CCR above them.

"production-grade" is a strange term of use. What your are talking about is a serious datacenter-switch.

Yes, Mikrotik should implement EVPN support. And Yes, they should implement MPLS/VXLAN hardware support if possible.

TCAM space comes with a cost, and makes no sense for a regular enterprise/datacenter switch with a maybe hundreds/thousands of routes as most.

But there is no really no point to trying to directly compete against Cisco Nexus:es, Juniper QFX:es, Aristas or cheaper Broadcom white-box switches. If you are dealing with business critical datacenter stuff, then the price is not the primary point.

Even if they made the hardware, the lack of paid software support would limit the market for MikroTik. No serious operator is going to run hardware with no realistic prospect of a fix in a timely manner.

Even if they made the hardware, the lack of paid software support would limit the market for MikroTik. No serious operator is going to run hardware with no realistic prospect of a fix in a timely manner.

I think this is the real issue. Even if MikroTik made a REAL L3 switching box, say a box costs $8k, nobody's buying it because of their crappy support system and model.

On the other hand - you really can't buy a Nexus or QFX without a pricey support agreement (except for second hand)

On the other hand - you really can't buy a Nexus or QFX without a pricey support agreement (except for second hand)

Fair argument. MikroTik can sell reasonably priced support agreement. 1/2 the price of Cisco or Juniper.

MikroTik can sell reasonably priced support agreement. 1/2 the price of Cisco or Juniper.

Any business [especially manufacturers] that hopes to be alive for a long time operates with a sustainable gross margin model. What my be reasonable to techies is certainly not reasolnable to a business man ... The TiK business model has a very high dependency on 3rd party distribution

So the Chain looks liike the following:
OEM/TIK/Tik Distributors/Sub Distributores/reselers
Each wants a piece of the profit pie
OEM's Profit Margin -- volume determines GM [for example Intel [cpu] will only sell a minumum of 1 million unites per order otherwise they will not talk to you]
Tik Profit Margin - volume determines GM
Distubutors Profit Margin - volume determines GM
Reselers Profit Margin - GM is hit and miss all depending on popularity of items

Very easy to say "MikroTik can sell reasonably priced support agreement. 1/2 the price of Cisco or Juniper" but that is not how it works in business ...

if you see that as an easy task then you are missing a good opportunity, raise some capital and startup your idea

i will be expecting that equipment built by your startup to revolutionize the market

You have the knowledge and the opportunity so go for it

Clearly this is sarcasm, not sure who's falling for this one.

was not sarcasm

maybe sometimes the rude tone i perceive on your words is only the result of the translation. but i found your position not very thoughtful neither constructive

was not sarcasm

maybe sometimes the rude tone i perceive on your words is only the result of the translation. but i found your position not very thoughtful neither constructive

Yeah, like I care about what some random dude on the internet thinks about it. You can think what you want, we're never crossing paths in real life, I never worked with and will never work with Latin America network operations, mainly due to financial issues in your region, compared to North America, Europe, South Africa (country), APAC.

Yeah, like I care about what some random dude on the internet thinks about it. You can think what you want, we're never crossing paths in real life, I never worked with and will never work with Latin America network operations, mainly due to financial issues in your region, compared to North America, Europe, South Africa (country), APAC.

<rant>
Dude, these ad hominem attacks have got to stop. They add absolutely zero value to the conversation, and every disparaging remark you make completely erases any clout or respect you might have earned when sharing your expertise or opinion. I came to this thread expecting to see what users had to say about higher-end L3 switches, but it's digressed (surprise, surprise...) to grade-school name-calling.

You whine and complain about how others treat you (referring to your "delete my account" thread), and yet here you are treating people like dirt (again). Online, in-person—it doesn't matter. It's all "real-life" and, unfortunately, how people act online shows more about their true nature than the charade they may put up for others "in real life." You can say all you want about your background, education, and experience, but a professional you are not.
</rant>

I had hoped to discuss more of the idea and technical merits of higher-end switches, but this thread is not worth my time anymore.

<rant>
Dude, these ad hominem attacks have got to stop. They add absolutely zero value to the conversation, and every disparaging remark you make completely erases any clout or respect you might have earned when sharing your expertise or opinion. I came to this thread expecting to see what users had to say about higher-end L3 switches, but it's digressed (surprise, surprise...) to grade-school name-calling.

You whine and complain about how others treat you (referring to your "delete my account" thread), and yet here you are treating people like dirt (again). Online, in-person—it doesn't matter. It's all "real-life" and, unfortunately, how people act online shows more about their true nature than the charade they may put up for others "in real life." You can say all you want about your background, education, and experience, but a professional you are not.
</rant>

I had hoped to discuss more of the idea and technical merits of higher-end switches, but this thread is not worth my time anymore.

The “attacks” did not start from me, and nobody in this thread “name called”.

Anyway, 99% of MikroTik users agree MikroTik is never going to compete on big-vendor game, max they'll do is SOHO, WISP, small FISP and that's about it.

The whole idea of REAL layer 3 switching isn't a viable idea in this community nor platform.

Don't you worry, I'll eventually just stop logging in or commenting—My knowledge and expertise are valued elsewhere and certainly Linux based networking, just not MikroTik specific.

Don't you worry, I'll eventually just stop logging in or commenting—My knowledge and expertise are valued elsewhere and certainly Linux based networking, just not MikroTik specific.

Translation = F*** I'm Good, Just Ask Me !

Translation = F*** I'm Good, Just Ask Me !

Translation = I don't need praise from people in this forum. Money doesn't reach my bank account from here. Some people have appreciated my comments in this forum, some have not, doesn't matter to me either way.

Civil discussion seems to be impossible with some people.
(and no, I'm not targetting one single person here)

Thread locked.