Good Day,

What equipment would be required to packet shape a connection up to 500Mbps?

We currently have a 100Mbps wireless connection coming into our campus and are looking to upgrade to a fiber based line, and possible have up to 500Mbps of bandwidth.

We use a Blue Coat Packet Shaper at present, but I am wondering if we could achieve the same thing with a Mikrotik / RouterOS based solution?

Thank you!

the new RB1100AHx2 dual core device can route 1 million pps, or up to 4Gbit per second. If there is any router that can do this, it will be the RB1100AHx2

the new RB1100AHx2 dual core device can route 1 million pps, or up to 4Gbit per second

just a notice for ChildOTK: it's about routing, not shaping. what do you mean by 'shape'?..

I said "route". It's hard to say about specific configuration, depends on settings, number of users, number of queues etc.

I said "route"

yeah, sure, I meant ChildOTK's 'shape'

Thanks for the response guys...

The biggest thing we do with our current packet shaper is we limit each IP Address to say 2Mbit. I know RouterOS can do this with queues, although I am not highly experienced with this. I have played with "queueing" an entire interface but not specific to IP / MAC addresses. Otherwise we allocate priority to our Staff IP Address pools, our server ip address pools, try to block certain things like p2p or bit torrent.

If we do upgrade our current connection to more than 200Mbit we will need to upgrade our packet shaper, but these are ridiculously priced, and I am wondering if we can accomplish the same thing with a RB1100AHx2 or any RouterOS installed device really.

we have x86 Intel platform with some 4-core CPU (Xeon?), shaping 300 Mbps in evenings (about 2000 online users)

2 normis - new RB1100AHx2 can route 1 million pps at which packet size ? 64 or 1500 byte, and can i route this 1 million pps from one port to second ? no just sum pps of all ports

1Mpps of 1500-byte packets is 12 Gbps - I doubt you can pass it through one port

1Mpps of 1500-byte packets is 12 Gbps - I doubt you can pass it through one port

Smart Chupaka

)) 1Mpps its summ of all PPS on interfaces. ~180000 x 13 ports ~= 2340000 pps

)) 1Mpps its summ of all PPS on interfaces. ~180000 x 13 ports ~= 2340000 pps

no it's not a sum of all interfaces

ok, 1 Mpps its from port 1 to port 2 ?? or what ? and what size of packet when test this ?

the test results are here, just look yourself

http://routerboard.com/RB1100AHx2

thanks normis, thats what i need )

we have x86 Intel platform with some 4-core CPU (Xeon?), shaping 300 Mbps in evenings (about 2000 online users)

Hi Chupaka. Could you please share here some detail of yours:
1. What was your CPU load % at this moment?
2. If possible please let me know more detail about your hardware? CPU model? Ethernet controller model?
I want to know yours, because I have maximum 710 pppoe active client. At that moment, CPU was 40%, traffic was 120Mbps/30Mbps on one interface, but latency of next hops suddenly increases from 0ms to 200ms at all interfaces. My CPU is 8x2400Mhz, maximum RAM usage is only 5%, ethernet controller is Intel 82571EB and 80003ES2LAN.

There is a difference in the terms used and I know the terminology in the wiki isn't always consistent. I could try to explain it but it's easier just to borrow one of Cisco's diagrams.



It's normally used in situations like transitioning from a gigabit lan to a T1/E1/OC3/etc.
If you're wondering why I'm bringing up Cisco terminology that's because it's pretty much the gold standard for network training and engineering.

http://routerboard.com/RB1100AHx2

What about performance of source NAT (@1518 byte frames)? Imagine up to 2000 users downloading with 1Gbit/s. So, each user downloading with 500kb/s. Will the RB1100AHx2 be able to route this?

http://routerboard.com/RB1100AHx2

What about performance of source NAT (@1518 byte frames)? Imagine up to 2000 users downloading with 1Gbit/s. So, each user downloading with 500kb/s. Will the RB1100AHx2 be able to route this?

at the bottom of the page is throughput test results. we don't do more specific tests because there are too many variables.