Somewhat confused as to the practicality of both of these products, since they have different CPU architecture
CCR's have more cores but do they still have better throughput than the RB1100AHx4?
I know that CCR's used to have issues particularly with IPSec, but i've heard that has been fixed?
Essentially what is the intended purpose of each device, and for core networks would it be better to use the RB1100's or the CCR's?
-
-
IntrusDave
Forum Guru
- Posts: 1286
- Joined:
- Location: Rancho Cucamonga, CA
Re: RB1100AHx4 vs CCR series for core/distribution?
The CCR routers have MUCH higher throughput than the RB1100AHx4. The CCR's are intended for core routing, with multiple networks attached. The RB1100AHx4 is geared more towards a larger business network. You can check out the block diagrams for each to see the design differences. On the CCR's, each port is directly connected to the CPU, providing a full wire-speed throughput. The RB1100AHx4 uses switch chips, each connected with a 2.5Gbps link.
Re: RB1100AHx4 vs CCR series for core/distribution?
I agree that even a low model CCR should be considered more powerful than the 1100AHX4.
The 1100AHX2 and 1100AHX4's individual cores however are, AFAIK, a fair bit more powerful than a single core of the CCR. The CCR's only win because they have so many cores. This can have an impact if your particular application will place a heavy load on a single core. You could theoretically hit a limit more rapidly with a CCR for such an application, even though for most things a CCR would be significantly faster/better.
An example is 'global' attached queue trees - my understanding is that these are all handled by one CPU core, so theoretically if you use a large queue tree setup attached to 'global' then the 1100AHX4 may perform better than a CCR in that very limited use case. However, also keep in mind that most of these situations can be solved by proper planning around the limitations (ex. using simple queues or interface-attached queue trees).
The 1100AHX2 and 1100AHX4's individual cores however are, AFAIK, a fair bit more powerful than a single core of the CCR. The CCR's only win because they have so many cores. This can have an impact if your particular application will place a heavy load on a single core. You could theoretically hit a limit more rapidly with a CCR for such an application, even though for most things a CCR would be significantly faster/better.
An example is 'global' attached queue trees - my understanding is that these are all handled by one CPU core, so theoretically if you use a large queue tree setup attached to 'global' then the 1100AHX4 may perform better than a CCR in that very limited use case. However, also keep in mind that most of these situations can be solved by proper planning around the limitations (ex. using simple queues or interface-attached queue trees).
Re: RB1100AHx4 vs CCR series for core/distribution?
Not only CPU is different, but how the ports are connected to it. Consider these block diagrams. CCR has all ports directly connected to the CPU. The RB1100 has switch chips in between. The connection between switch chip and the CPU is less than the sum of all port speeds.
You do not have the required permissions to view the files attached to this post.
Re: RB1100AHx4 vs CCR series for core/distribution?
Hi,
I bought 3x RB1100AHx4 Dude Edition routers to replace some TP-Link TL6120 in each of our 3 offices locations.
My main point for choosing the RB1100AHx4 was the idea that it could support 2.5Gbps throughput on one port. So I could achieve the theoretical maximum of 1Gbps of the port speed connection.
We are having at each location 1Gbps WAN speeds, so the TP-Link 350Mbps was a limiting factor.
But I tested now two RB1100AHx4 connected together (same port on both, ip addresses on each interface) with the bandwidth test provided by RouterOS interface, and for receive or transmit, individually, I get almost 1Gbps (975,4Mbps). If I do both, the speed is capped at ~650Mbps.
I've tried again after several hours, and now the speed gets to 975.4Mbps on both ways, but the lost packets increase dramatically.
Why?
Shouldn't the 2.5Gbps speed be full-duplex? 2.5 both ways, same time?
Would I was better off by going with the CCR1009?
I'm a little disappointed, as it's twice the price we paid for the TP-Links 4 years ago, and only twice the performance...
I bought 3x RB1100AHx4 Dude Edition routers to replace some TP-Link TL6120 in each of our 3 offices locations.
My main point for choosing the RB1100AHx4 was the idea that it could support 2.5Gbps throughput on one port. So I could achieve the theoretical maximum of 1Gbps of the port speed connection.
We are having at each location 1Gbps WAN speeds, so the TP-Link 350Mbps was a limiting factor.
But I tested now two RB1100AHx4 connected together (same port on both, ip addresses on each interface) with the bandwidth test provided by RouterOS interface, and for receive or transmit, individually, I get almost 1Gbps (975,4Mbps). If I do both, the speed is capped at ~650Mbps.
I've tried again after several hours, and now the speed gets to 975.4Mbps on both ways, but the lost packets increase dramatically.
Why?
Shouldn't the 2.5Gbps speed be full-duplex? 2.5 both ways, same time?
Would I was better off by going with the CCR1009?
I'm a little disappointed, as it's twice the price we paid for the TP-Links 4 years ago, and only twice the performance...
Re: RB1100AHx4 vs CCR series for core/distribution?
The ports are 1Gbit ports, so you can't get more than 1Gbit full duplex on one port. The test results show CPU/unit processing power, they are done through multiple ports at the same time. One port to one port is limited by the actual interface speed.
Re: RB1100AHx4 vs CCR series for core/distribution?
Maybe you are doing a lot of processing on the packets that you are routing, e.g. NAT, firewalling, fragmentation, etc?But I tested now two RB1100AHx4 connected together (same port on both, ip addresses on each interface) with the bandwidth test provided by RouterOS interface, and for receive or transmit, individually, I get almost 1Gbps (975,4Mbps). If I do both, the speed is capped at ~650Mbps.
I've tried again after several hours, and now the speed gets to 975.4Mbps on both ways, but the lost packets increase dramatically.
Why?
This could use a lot of CPU and there could be the limit, rather than at the interface speed or the link between switch and CPU.
Sometimes the processing can be optimized by a few simple steps, so please show us the configuration you use.
Re: RB1100AHx4 vs CCR series for core/distribution?
Ok, let's start again. Blank standard RB1100AHx4 routers.
Router1 - port 6 (IP address: 10.0.0.1/30) connected to Router 2 - port 6 (IP address: 10.0.0.2/30), Cat6 patch. Nothing else configured.
Router 1 -> Tools -> BTest Server -> Enabled
Router 2 -> Tools -> Bandwidth Test -> Test to: 10.0.0.1 (Random data enabled).
Which ever test (receive or send) I select, I get 650Mbps. If I take out the random data checkbox, it increases to 975.4Mbps, and the Lost Packets counter stays at 26000+.
If I test to 10.0.0.2 (so the output interface, so the packets are routed only internally), I get 650Mbps. If I take out the random data checkbox, I get 2.5Gbps and no Lost Packets.
Is it a normal to act like this? Is the CPU capped at 650Mbps/port?
Router1 - port 6 (IP address: 10.0.0.1/30) connected to Router 2 - port 6 (IP address: 10.0.0.2/30), Cat6 patch. Nothing else configured.
Router 1 -> Tools -> BTest Server -> Enabled
Router 2 -> Tools -> Bandwidth Test -> Test to: 10.0.0.1 (Random data enabled).
Which ever test (receive or send) I select, I get 650Mbps. If I take out the random data checkbox, it increases to 975.4Mbps, and the Lost Packets counter stays at 26000+.
If I test to 10.0.0.2 (so the output interface, so the packets are routed only internally), I get 650Mbps. If I take out the random data checkbox, I get 2.5Gbps and no Lost Packets.
Is it a normal to act like this? Is the CPU capped at 650Mbps/port?
Re: RB1100AHx4 vs CCR series for core/distribution?
You already said that random data unchecked brings it to port maximum, ~1Gbit.
What are you asking really?
When router needs to generate random imaginary traffic on the CPU, it takes a toll one the CPU resources. For a true test, you must test THROUGH the device, not ON the device.
This device has three switch chips ether1-ether5, ether6-ether10 and ether11-ether13 and on each of them total throughput is maximum 2.5Gbps
Each separate port supports maximum 1Gbps.
What are you asking really?
When router needs to generate random imaginary traffic on the CPU, it takes a toll one the CPU resources. For a true test, you must test THROUGH the device, not ON the device.
This device has three switch chips ether1-ether5, ether6-ether10 and ether11-ether13 and on each of them total throughput is maximum 2.5Gbps
Each separate port supports maximum 1Gbps.
Re: RB1100AHx4 vs CCR series for core/distribution?
BTest is ancient history, limited to one cpu core, use this:Ok, let's start again. Blank standard RB1100AHx4 routers.
Router1 - port 6 (IP address: 10.0.0.1/30) connected to Router 2 - port 6 (IP address: 10.0.0.2/30), Cat6 patch. Nothing else configured.
Router 1 -> Tools -> BTest Server -> Enabled
Router 2 -> Tools -> Bandwidth Test -> Test to: 10.0.0.1 (Random data enabled).
Which ever test (receive or send) I select, I get 650Mbps. If I take out the random data checkbox, it increases to 975.4Mbps, and the Lost Packets counter stays at 26000+.
If I test to 10.0.0.2 (so the output interface, so the packets are routed only internally), I get 650Mbps. If I take out the random data checkbox, I get 2.5Gbps and no Lost Packets.
Is it a normal to act like this? Is the CPU capped at 650Mbps/port?
https://wiki.mikrotik.com/wiki/Manual:P ... _Generator
Re: RB1100AHx4 vs CCR series for core/distribution?
Done. Thank you for the suggested test macgaiver.
Connected R1 to R2 as the link suggested, (port 2, 6, 11 and 12, on both routers), set router R2 as traffic generator (6 streams, 3 ports), request 500Mbps/stream (as there are 2 streams for Rx and 2 streams for Tx per port):
Obtain 484 to 500Mbps per stream.
R1 reported:
Add streams for the fourth connection (port 12), request 330Mbps/stream (as now there are 3 streams for Rx and 3 streams for Tx per port). R2 set as:
R1 reported:
Probably one could increase the number of ports, interconnection and streams to force the router to it's maximum capability, but for the two 1Gbps WANs that I want to use seems fine for me. I am now sure the 7.5Gbps is achievable per the specs sheet/drawing.
Thank you all for your help!
Connected R1 to R2 as the link suggested, (port 2, 6, 11 and 12, on both routers), set router R2 as traffic generator (6 streams, 3 ports), request 500Mbps/stream (as there are 2 streams for Rx and 2 streams for Tx per port):
Code: Select all
[admin@Bai210] /tool traffic-generator> quick tx-template=r12,r13,r21,r23,r31,r32 packet-size=1500 mbps=500R1 reported:
Code: Select all
[admin@Maz201] /interface> monitor-traffic aggregate,ether2,ether6,ether11
name: ether2 ether6 ether11
rx-packets-per-second: 246 095 82 032 82 028 82 027
rx-bits-per-second: 2.9Gbps 987.0Mbps 986.9Mbps 986.9Mbps
fp-rx-packets-per-second: 246 070 82 021 82 020 82 021
fp-rx-bits-per-second: 2.9Gbps 984.2Mbps 984.2Mbps 984.2Mbps
rx-drops-per-second: 0
rx-errors-per-second: 0
tx-packets-per-second: 244 794 80 723 82 030 82 029
tx-bits-per-second: 2.9Gbps 971.2Mbps 986.9Mbps 986.9Mbps
fp-tx-packets-per-second: 246 070 80 706 82 014 81 979
fp-tx-bits-per-second: 2.9Gbps 968.4Mbps 984.1Mbps 983.7Mbps
tx-drops-per-second: 0
tx-queue-drops-per-second: 1 361 10 5 1 346
tx-errors-per-second: 0
Add streams for the fourth connection (port 12), request 330Mbps/stream (as now there are 3 streams for Rx and 3 streams for Tx per port). R2 set as:
Code: Select all
[admin@Bai210] /tool traffic-generator> quick tx-template=r12,r13,r14,r21,r23,r24,r31,r32,r34,r41,r42,r43 packet-size=1500 mbps=330Code: Select all
[admin@Maz201] /interface> monitor-traffic aggregate,ether2,ether6,ether11,ether12
name: ether2 ether6 ether11 ether12
rx-packets-per-second: 325 515 81 926 82 025 79 527 82 027
rx-bits-per-second: 3.9Gbps 985.7Mbps 986.9Mbps 956.8Mbps 986.9Mbps
fp-rx-packets-per-second: 327 689 82 483 82 574 80 055 82 572
fp-rx-bits-per-second: 3.9Gbps 989.7Mbps 990.8Mbps 960.6Mbps 990.8Mbps
rx-drops-per-second: 0
rx-errors-per-second: 0
tx-packets-per-second: 324 354 78 579 81 929 81 818 82 026
tx-bits-per-second: 3.9Gbps 945.4Mbps 985.8Mbps 984.4Mbps 986.9Mbps
fp-tx-packets-per-second: 327 689 79 104 82 483 82 401 82 563
fp-tx-bits-per-second: 3.9Gbps 949.2Mbps 989.7Mbps 988.8Mbps 990.7Mbps
tx-drops-per-second: 0
tx-queue-drops-per-second: 1 134 0 0 651 483
tx-errors-per-second: 0
-- [Q quit|D dump|C-z pause]
Thank you all for your help!
Re: RB1100AHx4 vs CCR series for core/distribution?
Have a ressurected CCR1009 on the desk and was wondering the same thing.
Purely computational wise, RB1100AHx4 is light years ahead of CCR1009
Bandwidth test on 127.0.0.1 for both I think shows that
CCR1009 RB1100AHx4 Also note that the CCR1009 was blank, without any firewall rules or traffic, while RB1100AHx4 had a lot of firewall rules and was in use at the time of test.
Purely computational wise, RB1100AHx4 is light years ahead of CCR1009
Bandwidth test on 127.0.0.1 for both I think shows that
CCR1009 RB1100AHx4 Also note that the CCR1009 was blank, without any firewall rules or traffic, while RB1100AHx4 had a lot of firewall rules and was in use at the time of test.
You do not have the required permissions to view the files attached to this post.
Re: RB1100AHx4 vs CCR series for core/distribution?
How can this be?Have a ressurected CCR1009 on the desk and was wondering the same thing.
Purely computational wise, RB1100AHx4 is light years ahead of CCR1009
Bandwidth test on 127.0.0.1 for both I think shows that
CCR1009
CCR1009.png
RB1100AHx4
1100AHx4.png
Also note that the CCR1009 was blank, without any firewall rules or traffic, while RB1100AHx4 had a lot of firewall rules and was in use at the time of test.
Is there a logical explanation for this?
Re: RB1100AHx4 vs CCR series for core/distribution?
No clue. This is the only CCR1009 I can test with.
Anybody else with one can test on local with random?
Maybe my unit is in a coma.
later edit:
found another 1009. Same story
Anybody else with one can test on local with random?
Maybe my unit is in a coma.
later edit:
found another 1009. Same story
You do not have the required permissions to view the files attached to this post.
Re: RB1100AHx4 vs CCR series for core/distribution?
How can this be?
Is there a logical explanation for this?
yes
RB1100ahX4 despite being a aparently smaller device has some advantage because of the follwing differences in CPU:
ccr1009 has 9 cores at 1.200 mhz using in-order architecture
rb1100ahx4 has 4 cores at 1.400mhz using out-of-order execution architecture
the advantage of out-of-order execution architecture allows the rb1100ahx4 to do more work for each CPU cicle, because of that you can see the rb1100ahx4 challenging or beating the ccr1009 despite having less count of cores beacuse each core is more productive specially if the task can take advantage of single core performance which is around 2X times in favor of rb1100ahx4
sumarizing
the 4 core cpu of the rb1100ahx4 can compare with 9 core cpu of ccr1009 in multi core loads
for single core loads rb1100ahx4 is able to concentrate in one core the processing capabilities of 2 ccr1009 cpu cores
let's extrapolate this phenomenon to newer ccr2116 which has the follwing improvements over the rb1100ahx4
modern more effective out-of-order CPU
higher cpu clock up to 2.000 mhz
16 cores
wider ddr4 ram memory bus (vs ddr3 on rb1100ahx4)
for single core loads ccr2116 cpu is able to concentrate in one core the processing capabilities of at least 3 ccr1036 cpu cores, so in single core loads you can see up to 3x improvements
because of that you can see ccr2116 16 core CPU improving over 30% percent in heavy processing scenarios over the ccr1036 with 36 tilera cores in multi core loads
Re: RB1100AHx4 vs CCR series for core/distribution?
The biggest drawback of the rb1100ahx4 is the lack of SFP ports otherwise a great router.
Re: RB1100AHx4 vs CCR series for core/distribution?
rb1100ahx4 is a niche product specifically designed to fit as a replacement for rb1100ahx2 deploymentsThe biggest drawback of the rb1100ahx4 is the lack of SFP ports otherwise a great router.
For any other scenario you have the RB4011 which uses the same CPU and has sfp+ interface, or ccr2004-1g-12s+2xs with lots of sfp interfaces and more performance
or ccr2004-16g-2s+
Re: RB1100AHx4 vs CCR series for core/distribution?
The CCR2K series is not an option in certain cases because it does not support RouterOS v6. And as is well known, there are features that are not fully supported in RouterOS v7
Who is online
Users browsing this forum: aboiles and 17 guests