MikroTik

http://www.ubnt.com/edgemax#EdgeMAXhardware
http://dl.ubnt.com/Tolly212128UbiquitiE ... kroTik.pdf

what will you say?

Yes, wondering about this too. Lower price of 1100AH series and Cloud Router for $1000?

Also, I'm wondering what they are not telling you. Ubiquiti wireless radios seem very interesting at first but when you want something as simple as having multiple SSID's, you can't do it. Multiple SSID's with different VLAN's is a must I would say in many situations, but their software just doesn't support it and no word on when it will. Not saying Ubiquiti is bad but Mikrotik has proven to be more versatile.

I do not believe.

price of $ 99!
D-Link have price more!

EdgeMax router have new OS - EdgeOS

Packed with Features

IPv4/v6 addressing, DHCP client and server,
VLANs Static routes, OSPF, RIP, BGP
Firewall (ACL-based and zone-based), NAT, QoS
VPN: IPsec, L2TP, OpenVPN, PPTP client and server
PPPoE client and server, bridging, bonding, GRE,
VRRP
Dynamic DNS, DNS forwarding, DHCP relay

zdraste zdraste Oleg

Yes, wondering about this too. Lower price of 1100AH series and Cloud Router for $1000?

Also, I'm wondering what they are not telling you. Ubiquiti wireless radios seem very interesting at first but when you want something as simple as having multiple SSID's, you can't do it. Multiple SSID's with different VLAN's is a must I would say in many situations, but their software just doesn't support it and no word on when it will. Not saying Ubiquiti is bad but Mikrotik has proven to be more versatile.

Ubiquiti is famous for over-promising and under-delivering and I find latest hardware to be of much lower quality than previous generations (coming to think of it - it's the same with mikrotik). In any-case, competition is great and I'm looking forward towards cheaper mikrotik products.

EdgeOS does not appear to have dynamic access lists, packet marks, PCC load balancing and PCQ. It probably has better dhcp, dns and HTB just because mikrotik is outdated in that area.

Hi Max


we will wait for the real tests, prices and reviews

http://www.youtube.com/watch?v=PgjKvJ2rdJY


cool numbers on machine CYSCO (subtle hint)

Biggest difference from what I can see is that the Ubiquiti device can do some kind of hardware acceleration and thus achieve the high speeds. Most of the time this works great, except when you wish to do some fancy feature work then the hardware layer cannot support it and it still needs to be done through software aka CPU.

Same difference between a hardware layer 3 switch what can do hardware layer 3 routing at wire speed, but if you wish to do more with your packets, you'll need a device that does it in software because of the versatility.

But it will be very interesting to see what it can and cannot do. I'm sure for some configurations it'll perform quite nice, but probably if you wish to keep that kind of performance, you'll be limited in what you can do. And thus mikrotik will probably still win with features and flexibility. Still, it's a very nice product for a very nice suggested price. Interesting times!

router with hardware acceleration with price 99$ ??????


There's Always Free Cheese In A Mousetrap.

In that article they say that:
"Ubiquiti's performance was unaffected, demonstrating low and in some cases lower latency than without the firewall"
It's a joke right!? How this can be possible?
Maybe they reinvented how routering works.
And this is only on a Lite Router

I don't get the negativity on here tbh . Competition is always good for the consumer . Who knows you might get a new hardware revision or hardware facelift on the 1100/1100AX series or perhaps get niceties like SFP's thrown on in the future from MTK ?

I think of Ubiquiti as cheap, simple plastic devices with limited functionality. The new products may change my mind.

Physically the new EdgeMax product line looks very professional. But being based on the Vyatta IOS clone is not a good thing IMO, im not a fan of the IOS CLI, they should have cloned RouterOS or JunOS CLI if they wanted to clone something good.

Im a wait and see kind of guy. Im waiting to see if CCR can perform, and fixes a bunch of issues we are having with L3VPN/VRF's.

That comparison is flawed.
Compare a 3 port router with a 13 port one and say "look, the 3 port one is cheaper". This one should compare to a RB433G, better yet to an hypothetical RB430G.

In this case, compared to the $99, you get 3 Gb ports and 3 miniPCI slots for $139. Somehow on par, don't you think?


On the other hand, Ubiquity uses plastic, but they work quite reliable. I have some Rockets and some Nanostations M5 up for 2 years now, and they work and look as new. And compared to other outdoor products they look pretty nice.

I don't get the negativity on here tbh . Competition is always good for the consumer . Who knows you might get a new hardware revision or hardware facelift on the 1100/1100AX series or perhaps get niceties like SFP's thrown on in the future from MTK ?

At this point all you can do is being cynical since. for one, product is not available, two, the test was commissioned so one product looks better than other.

EDIT: However, as I posted here: http://forum.mikrotik.com/viewtopic.php?f=2&t=65526, RB1100AH Tolly,s tests revealed interesting trends:

1. Somehow ROS is unable to process large number of packets on a single interface (it can process a lot more packets is they come through different ports)
2. You have to pay attention which port you use to get maximum throughput.

It sounds like the RB1100AH is a 5 port router, 5 Gig-E ports to the CPU? The fact that 2 of those ports are connected to embedded 6-port GigE switches, leaving 5 open ports on each switch is a bonus.

For maximum throughput on RB1100AH pass traffic between combinations of ether1, ether6, ether11, ether12, ether13? Or are ether11 and ether12 one port to the CPU, due to the passthrough capability? I think not but...

The RB493G is a 2 port router connected to one embedded 6 port Gig-E switch and one embedded 5 port GigE switch?

For maximum throughput on RB493G pass traffic between ether1, ether2?

If that is correct, it doesn't bug me. I just would like to be made more aware of that in the documentation, not that anyone reads documentation. But if we did, we could make more of an effort to ensure that the most utilized path through the router does not end up with both links on the same embedded switch.

We should probably also spread our connection(s) to our VLAN capable switches across more physical connections to the switch with high throughput VLANs ending up on different CPU ethernet hardware ports, rather than 1 arming it. The config will end up being more complex, but the throughput should be better, right?

If you search the forums and MUM presentations you will find this about RB1100AHx2:

switch group 1 - include ether1-5, have 1Gbps up and 1Gbps down shared connection to CPU
switch group 2 - include ether2-6, have 1Gbps up and 1Gbps down shared connection to CPU
ether11 - direct connection to CPU - fastest port on the board, should always be used as backbone port
ether12,ether13 - PCI-E ports - connected to CPU via PCI-E bus (more resource demanding and not as fast latency wise as directly connected ports.)

So usual setup should be - ether1-10 clients, ether11 backbone port, ether12-backbone backup (ether11 and ether12 have hardware bypass functionality), and ether13 management port.

to get max number of 64byte packets you need to use less "expensive" ports and minimum number of ports. so for this test you should go ether11<--->any one of <ether1-10>
to get max throughput, first you should go for 3port test first (example ether11,ether6,ether1), if max 3Gbps is reached you can try to connect ether12

@lambert:
Maybe I have a wrong impression, but there are 9 independent individual gigabit ports on the 439G.
The switch chips are there to allow them to be optionally switched, and are bypassed by default.
So it is a 9 port router, not a 2 port router + a switch. Each of the 9 ports with its IPs, routes and so on.

I never used a 493, but the RB450G at least behaves like this: 5 individual Gb ports + optional HW switching between ether2-5.
Not 2 ports + switch.

Now if the switch chip is used to access the phys for each port and transfer data to the cpu - that's another story: it is still not a L2 only device for those ports. "Hub", "Switch", "Bridge" and "Router" refers only to its functional capability on OSI stack level, not to its internal hardware architecture.

Think like this:
- Hub: 1 single collision domain
- Switch, Bridge: 1 single broadcast domain, individual collision domains per port
- Router: individual broadcast domains per port

What Ubiquiti Edge Router has the bandwidth management PCQ + queue tree like mikrotik?
I love mikrotik becouse queuetree+PCQ

If you search the forums and MUM presentations you will find this about RB1100AHx2:

switch group 1 - include ether1-5, have 1Gbps up and 1Gbps down shared connection to CPU
switch group 2 - include ether2-6, have 1Gbps up and 1Gbps down shared connection to CPU
ether11 - direct connection to CPU - fastest port on the board, should always be used as backbone port
ether12,ether13 - PCI-E ports - connected to CPU via PCI-E bus (more resource demanding and not as fast latency wise as directly connected ports.)

So usual setup should be - ether1-10 clients, ether11 backbone port, ether12-backbone backup (ether11 and ether12 have hardware bypass functionality), and ether13 management port.

to get max number of 64byte packets you need to use less "expensive" ports and minimum number of ports. so for this test you should go ether11<--->any one of <ether1-10>
to get max throughput, first you should go for 3port test first (example ether11,ether6,ether1), if max 3Gbps is reached you can try to connect ether12

11, 12 and 13 ports are useless unless the router is located in a bomb shelter. In the network I have about 20 RB1100xx routers and in most of them mentioned ports are not working. on ~10 routers altogether I dont have port 11, as it does not exist.

@lambert:
Maybe I have a wrong impression, but there are 9 independent individual gigabit ports on the 439G.
The switch chips are there to allow them to be optionally switched, and are bypassed by default.
So it is a 9 port router, not a 2 port router + a switch. Each of the 9 ports with its IPs, routes and so on.

I never used a 493, but the RB450G at least behaves like this: 5 individual Gb ports + optional HW switching between ether2-5.
Not 2 ports + switch.

The 493G is definitely a router with 9 Gig-E ports. I'm not sure it is a 9 Gig-E router. (All trees are plants but not all plants are trees.)

I suspect that the CPU/PCI bus(es) has(have) one GIg-E connection to each switch group. The CPU/PCI bus port doesn't show up in the view which RouterOS gives us. Instead RouterOS may be showing us the "port based VLANs", my concept picture, and calling them ether1,ether6-9 and ether2-5. I did not find documentation that supports that view for the RB4xx series. It is just my theory which could be completely wrong. I do get the "port based VLAN" impression from what I have read about the RB1100xx series.

You won't get 1Gbps from ether2 to ether3 and 1Gbps from ether4 to ether5 unless the ports are in "switch mode" and the packets never traverse the CPU port. If the packets traverse the CPU, they will be bottle necked by the 1Gbps connection from the switch to the CPU. It doesn't matter how beefy the CPU is in that hardware configuration.

In the end, it doesn't make a lot of difference if all of the GigE ports are connected directly to the CPU or are used as "port based VLANs". The CPU still can't saturate the GigE interfaces it has. I'm happy getting > 100Mbps on some of the ports in my network which need the speed. I don't need 200Mbps on one interface in most locations. Having all those ports is very nice for us because it reduces the number of discrete devices at a location. Also you might get GigE speeds between servers on the same LAN if you do use the ports in switching mode. Haven't tried yet.

...
If the packets traverse the CPU, they will be bottle necked by the 1Gbps connection from the switch to the CPU. It doesn't matter how beefy the CPU is in that hardware configuration.
...
The CPU still can't saturate the GigE interfaces it has...

You are probably right. But even if that internal interface is capable of more than 1 Gb, the CPU will still be the bottleneck.
I don't think that a single core CPU e.g. at 680MHz can do more data processing since it has a data handling capability around 4bytes/machine cycle.
And there are more than 1 instruction needed per processed dword. So even if the ports would connect directly, the CPU would still be a bottleneck.

IMHO the only way to allow multiport wire speed multi-Gb routing are L3 switches, like e.g. the Brocade Iron switches and MLX routers which achieve 10+Tb routing capacity, but using dedicated ASICs. These are of course in a total different price range.

Sounds like a great option, especially if one gets:

• OpenVPN that actually supports more than 10% of the spec.
  An IPSec policy match.
  A scripting language that isn't so fragile. [Like, gasp, perl or bash]
  Allows you to drop into IPTables to build FW rules, with something like FWBuilder.
  Etc...

Obviously we'll see how it pans out, but I love to see something motivate MikroTik to really make some progress on nagging issues that have never been resolved and are really terrible/junky implementations.

More cheerful acceptance and acknowledgement of problem issues and less abuse of customers would be a nice topping too.

-Greg

If you search the forums and MUM presentations you will find this about RB1100AHx2

I've searched everywhere, among the MUM presentations and elsewhere, about the actual inner wiring of the RB1100AHx2 but I've been very unsuccessful...

Can you please point me to a documentation/presentation that explains how the "port group" work on this board?

Thanks!

Always use ports with same l2mtu to connect if possible. Never use 11,12 and 13.

If you search the forums and MUM presentations you will find this about RB1100AHx2

I've searched everywhere, among the MUM presentations and elsewhere, about the actual inner wiring of the RB1100AHx2 but I've been very unsuccessful...

Can you please point me to a documentation/presentation that explains how the "port group" work on this board?

Thanks!

one that i remember is this:
http://www.tiktube.com/video/JGdn3goDdE ... wolLonKGo=
there were few more sources of information from Normis and sergejs here on forum.

to ropebih:
You are talking about old RB1100, i was talking about new RB1100AH and RB1100AHx2 if i remember correctly then had not only different structure, but also had different CPUs (similar to RB1000)

Thanks all for sharing your opinion about the news. But let's keep the forum discussion about MikroTik products.