I was looking for CCR2004 review to find you review. To correct your information regarding the TILERA based 1000 series CCR. Until the TILERA production itself is halted, the MIKROTIK has no issue regarding the dropped TILERA support in LINUX kernel, since MIKROTIK is using theiw own, custom kernel in ROUTEROS, and only version 7 switches to newer version.
Hey there, I am comparing the CCR2004-16G-2S+ to the CCR2004-1G-12S+2XS and I'm not sure which one to get. The 16G has many 1Gb Eth ports which I don't really need. However it boasts that it's the fastest router in their entire CCR2 lineup. I'd rather get the 12S, but I'm concerned it's single-core routing will not be as high performance, even though they have the same CPU. I figured with your hands-on knowledge you can speak to this. On these models, would they experience the same router performance? Neither router will be under heavy switch load but it will be connected to a 1Gb+ Internet connection. Thank you!
Hi Chad, there is a "Passive Inteligent Port Extender" in a block diagram of the CCR2004-1G-12S+2XS. It sounds like a Mikrotik shortcut which usually leads to all sorts of problems. In a case of CCR2004-16G-2S+ at least you know you are dealing with Marwell switch chips. Maybe a later would be a better choice if you do not need 25Gbps or something like this.
It would appear to me there are problems with hashing algos between the 2 switches. Perhaps you want to test again with both server trunk left to LACP/L3+4, but the inter-switch trunk to one of the balance modes (rr or alb)?
I think I was testing all the combinations, there was no big difference. Will try to test this again, when I have appropriate equipment. I'll try the 25 Gbps cables too.
I'm sure that 25GBit/s can totally be switched, I suspect that the Router can't hw offload the LACP and thus has high cpu utilization while running the benchmark. It would be interesting if using BGP+ECMP would be any better.
@@nejcsu very much looking forward to it ! Also - if you route on a LACP interface - will bottleneck be the switching capacities shown in this video you think?
@@nejcsu I just did some test myself on CCR2004. I am getting similar results as you do. I am testing LACP, which obviously goes through the CPU because it has no switch chip. If I route through LACP interfaces, when using iperf3 --bidir -P128 I can push about 11Gbps one way and 7Gbps the other way, while CPU goes up to ~68%. On 7.1beta5 that is, because transmit-hash-policy is broken in 6.x (it will only send via one interface). I have made an excel sheet if you are interested in seeing my test, let me know :)
@@nejcsu Also I did one test without LACP. Setup: - two machines with two interfaces each - each interface in separate subnet - make route to remote machine via one interface each - mikrotik ccr2004 with on interface in each of all 4 subnets. now make iperf3 --bidir -P16 via each subnet total of 30.4Gbps. CPU at 88%. 30Gbps is what I get if I remove ccr2004 and just --bidir -P16 directly via my machines. So bottleneck is my CPU. (8 core Intel(R) Xeon(R) CPU E3-1245 v5 @ 3.50GHz) But I am not sure why ccr2004 would route much faster than it can using lacp? Because traffic goes through CPU either way if its just routed or LACP+routed or just bridged?
It's in audio range of a desktop PC, it would be no problem to put it into office environment. Audio in this video was recorded while everithing was running...
@@VirtuaitSisec Do you expect the RouterOS 7 version to be free for the mikrotik router owners or will i have to buy it as soon as it is on the market? Because on the productpage it says that only 2 of the 4GB s of RAM are avaible on RouterOS 6 for some reason
@@Kappi1997 if they ever finish it, I think it will be free for current users. There are some alternatives like VyOS and OpenWRT, i do not think they can afford to charge for it.
Even if you are in bridge mode did you disable the firewall rules .. as there will be some default rules .. that will apply on interfaces.
I was looking for CCR2004 review to find you review. To correct your information regarding the TILERA based 1000 series CCR. Until the TILERA production itself is halted, the MIKROTIK has no issue regarding the dropped TILERA support in LINUX kernel, since MIKROTIK is using theiw own, custom kernel in ROUTEROS, and only version 7 switches to newer version.
QUERY ? A ccr2004-1G-12S+2XS SHOULD FAN THE EQUIPMENT WHEN TURNED ON OR WHEN THE EQUIPMENT OVERHEATS AUTOMATICALLY, DOES THE FAN ACTIVATE?
Hey there, I am comparing the CCR2004-16G-2S+ to the CCR2004-1G-12S+2XS and I'm not sure which one to get. The 16G has many 1Gb Eth ports which I don't really need. However it boasts that it's the fastest router in their entire CCR2 lineup. I'd rather get the 12S, but I'm concerned it's single-core routing will not be as high performance, even though they have the same CPU. I figured with your hands-on knowledge you can speak to this. On these models, would they experience the same router performance? Neither router will be under heavy switch load but it will be connected to a 1Gb+ Internet connection. Thank you!
Hi Chad, there is a "Passive Inteligent Port Extender" in a block diagram of the CCR2004-1G-12S+2XS. It sounds like a Mikrotik shortcut which usually leads to all sorts of problems. In a case of CCR2004-16G-2S+ at least you know you are dealing with Marwell switch chips. Maybe a later would be a better choice if you do not need 25Gbps or something like this.
Need a suggestion CCR1009-7G-1C-1S+ vs CCR2004-16G-2S+ which one is better for radius, with 500 static ip users
Hello all right, please do you have the maintenance electrical scheme of this ccr2004 I need to do repairs. Thanks
It would appear to me there are problems with hashing algos between the 2 switches. Perhaps you want to test again with both server trunk left to LACP/L3+4, but the inter-switch trunk to one of the balance modes (rr or alb)?
I think I was testing all the combinations, there was no big difference. Will try to test this again, when I have appropriate equipment. I'll try the 25 Gbps cables too.
Hello sir. That's support POE in sir? Thanks
Which model of these servers?
The servers are Supermicro SuperServer E300-9D-4CN8TP. They have built in SFP+ ethernet.
i used switch independent trunks with hyper-v, because there is no hw offloading with lacp on the ccr2004. i moved the lacp ports to crs354
Can you try again with iperf3 -B? Curious if you're hitting an issue with packet transport, throughput or something deeper in routing.
I'm sure that 25GBit/s can totally be switched, I suspect that the Router can't hw offload the LACP and thus has high cpu utilization while running the benchmark. It would be interesting if using BGP+ECMP would be any better.
Yes, please do the routing aswell! Also, have you tried v7? Any differences?
Ok, have to order some equipment, will happen soon.
@@nejcsu very much looking forward to it ! Also - if you route on a LACP interface - will bottleneck be the switching capacities shown in this video you think?
@@nejcsu I just did some test myself on CCR2004. I am getting similar results as you do. I am testing LACP, which obviously goes through the CPU because it has no switch chip. If I route through LACP interfaces, when using iperf3 --bidir -P128 I can push about 11Gbps one way and 7Gbps the other way, while CPU goes up to ~68%. On 7.1beta5 that is, because transmit-hash-policy is broken in 6.x (it will only send via one interface). I have made an excel sheet if you are interested in seeing my test, let me know :)
@@nejcsu Also I did one test without LACP.
Setup:
- two machines with two interfaces each
- each interface in separate subnet
- make route to remote machine via one interface each
- mikrotik ccr2004 with on interface in each of all 4 subnets.
now make iperf3 --bidir -P16 via each subnet
total of 30.4Gbps.
CPU at 88%.
30Gbps is what I get if I remove ccr2004 and just --bidir -P16 directly via my machines. So bottleneck is my CPU. (8 core Intel(R) Xeon(R) CPU E3-1245 v5 @ 3.50GHz)
But I am not sure why ccr2004 would route much faster than it can using lacp? Because traffic goes through CPU either way if its just routed or LACP+routed or just bridged?
@@sliddjur Hi, please do share the file, you can send it to info@virtua-it.si
Hello,
please, what servers are You using for traffic generation?
Thanks in advance and have a great day!
Hi we were using Supermicro Superserver E300-9D-4CN8TP servers. 32GB Ram
A bit of a strange question but how loud is it? I was thinking about placing this in an office enviroment. Will this disturb people while working?
It's in audio range of a desktop PC, it would be no problem to put it into office environment. Audio in this video was recorded while everithing was running...
@@VirtuaitSisec thanks a lot for the information.I guessed it but sometimes microphones can filter very well
@@VirtuaitSisec Do you expect the RouterOS 7 version to be free for the mikrotik router owners or will i have to buy it as soon as it is on the market? Because on the productpage it says that only 2 of the 4GB s of RAM are avaible on RouterOS 6 for some reason
@@Kappi1997 if they ever finish it, I think it will be free for current users. There are some alternatives like VyOS and OpenWRT, i do not think they can afford to charge for it.
nice one
What is your DAC cable type sir?
Hi, it was the one from FS.COM. But that should not matter, they are all the same.
@@VirtuaitSisec what do you meant they all same.. so dac for cisco is usable in application for mikrotik sir?
@@kemalrizkywahyudi6709 I do not know about Cisco, but they do work with Fortinet, Aruba/HP and Ubiquity.
It works even using cheap mikrobits worth as $50 or fewer
Es español
Which model of these servers?