Best router for CCNA?

Street · 16 Dec 2009 at 13:43

I managed to get myself a couple 2611 routers fairly cheaply on the bay to play around with.. Just need to get some other kit now!

Might just stick with my current router for my house for now anyway..

Arthalen · 16 Dec 2009 at 14:02

A top tip I found when I first started delving into the world of CCNA and Cisco routers, is that the type of chassis of the router was largely unimportant!

Whats more important is the version of IOS it supports and what Wan Interface Cards (WIC) or Network Modules (NM) the device comes with or what it supports.

For example, one of my 2610's has a ISDN Pri port on a NM for use with an ISDN simulator (not that I can get my hands on one) ...

Also, unless you get the transceivers avoid the old 2500 routers, as there ethernet ports tend to be AUI port based, and not RJ-45 based.

bigredshark · 16 Dec 2009 at 15:48

DRZ said:
I have read that same PDF, but in my testing with real-world packet sizes I get far more!

An 871 for example is rated at something like 8Mbit, but can route at 90mbit and NAT at 40mbit with an internet MTU. Clearly that tails off massively as you add extra stuff into the path, but an 1841 will easily do more than 37Mbit in the real world.

Depends what you do, add policy based routing and 37Mbps becomes wishful thinking under some circumstances. You may get more but I'll stick to what the supplier says it'll do myself...

DRZ · 17 Dec 2009 at 00:36

bigredshark said:
Depends what you do, add policy based routing and 37Mbps becomes wishful thinking under some circumstances. You may get more but I'll stick to what the supplier says it'll do myself...

Its what it says it will do with teeny tiny packet sizes though?? Real-world is a completely different ballgame, even when you throw in all sorts of stuff?

Honestly, do you really believe anyone would buy an 877 if it couldnt even route/nat basic ADSL2+ speeds?

Our 100mbit leased lines come with 2811 routers, which according to that PDF cannot do the job... yet they do, day in day out?

My old 1760 could only do 8mbit accoring to that same PDF, yet it handled my ADSL connection just fine even when doing advanced NBAR-based QoS amongst other things.

I have no idea why Cisco choose to use the figures they do, but personally I will choose to use what I know to work in the real world!! If you do the sums with the packet per second throughput with a decent MTU, you start to see what you see when testing in the lab!

Arthalen · 17 Dec 2009 at 09:40

How is every calculating the packet throughput in relation to MTU size?

Are we talking simple bandwidth calculations, reviewing interface counters on the router or some kind of mathematical wizardry that I'm unaware of?

Please share so we can all learn

Yamahahahahaha · 17 Dec 2009 at 11:00

Cisco issued a document outlining the minimum throughput with 64-byte packets in packets per second, for a selection of routers.

http://www.cisco.com/web/partners/downloads/765/tools/quickreference/routerperformance.pdf

"Numbers are given with 64 byte packet size, IP only, and are only an indication of raw switching performance." So this has nothing to do with throughput, only handling packets of a minimal size.

Anyway, this is very off topic now.
CCNA is quite basic in the necessities for the feature set. If you get a newish 800 series, that will have many of the important features.
You will also need two switches that support VLAN trunking.

paradigm · 17 Dec 2009 at 11:33

Whilst we are on the subject, I've been searching for a definitive answer to this, but what would I be looking at to implement a CCNP lab?

I currently have:

1x 877
1x 2651XM
1x 2611XM
2x C3548-XL

Yamahahahahaha · 17 Dec 2009 at 12:20

http://www.ciscokits.com/ccnp-lab-suggestions/

4 switches, 2 x layer 3 switches, 2 x normal switches.
More routers the better.

bigredshark · 17 Dec 2009 at 13:48

DRZ said:
Its what it says it will do with teeny tiny packet sizes though?? Real-world is a completely different ballgame, even when you throw in all sorts of stuff?

Honestly, do you really believe anyone would buy an 877 if it couldnt even route/nat basic ADSL2+ speeds?

Our 100mbit leased lines come with 2811 routers, which according to that PDF cannot do the job... yet they do, day in day out?

My old 1760 could only do 8mbit accoring to that same PDF, yet it handled my ADSL connection just fine even when doing advanced NBAR-based QoS amongst other things.

I have no idea why Cisco choose to use the figures they do, but personally I will choose to use what I know to work in the real world!! If you do the sums with the packet per second throughput with a decent MTU, you start to see what you see when testing in the lab!

My judgement is based on in house testing with industry standard test gear, Cisco's figures are a fairly good guide as they indicate a level of performance it will provide in almost all circumstances. If you think you can get away with less then go for it, personally I'm not going to buy a product that can't do the job according to the manufacturers figures, despite what the reality might be I want it certified to do the job my the people who made it.

Arthalen · 17 Dec 2009 at 16:06

Sorry to derail the thread again (gimme a break I did answer the OP question

), but I'm guessing this is about right?

Cisco 2610XM router = 20,000 Packets Per Second @ Fast/CEF Switching

So, instead of the quoted 64Byte packet size, we use the lowest common denominator MTU and its maximum possible data payload, which would be:

Ethernet MTU (1500 bytes) - IP header (20 bytes Minimum) - TCP Header (5 Bytes minimum)

Therefore,

((20000 x 1475 x 8) / 1024) / 1024 = Approximate Speed in Mbps?
= 225Mbps .. or there abouts

However, surely the inverse is true? Cisco's quoted packet size of 64bytes is simply the smallest payload size of an Ethernet frame, and thus, the smaller the packet the quicker the routing of packets, so in an abirtrary kind of way may be Cisco's quoting there own 'best case' scenario.

Edit: Or is Cisco's quoted PPS rating simply the best possible no mater the data payload size?

bigredshark · 17 Dec 2009 at 20:32

Arthalen said:
Sorry to derail the thread again (gimme a break I did answer the OP question ), but I'm guessing this is about right?

Cisco 2610XM router = 20,000 Packets Per Second @ Fast/CEF Switching

So, instead of the quoted 64Byte packet size, we use the lowest common denominator MTU and its maximum possible data payload, which would be:

Ethernet MTU (1500 bytes) - IP header (20 bytes Minimum) - TCP Header (5 Bytes minimum)

Therefore,

((20000 x 1475 x 8) / 1024) / 1024 = Approximate Speed in Mbps?
= 225Mbps .. or there abouts

However, surely the inverse is true? Cisco's quoted packet size of 64bytes is simply the smallest payload size of an Ethernet frame, and thus, the smaller the packet the quicker the routing of packets, so in an abirtrary kind of way may be Cisco's quoting there own 'best case' scenario.

Edit: Or is Cisco's quoted PPS rating simply the best possible no mater the data payload size?

payload size is largely irrelevant, packets per second is the true measure of throughput on a router really. For most purposes a middle value like 500 bytes will give a decent idea of real world throughput.

The work is in determining what to do with the packet, actually doing it is easy (which is why high end routers have separate forwarding engines and routing engines). It's also why doing anything like policy routing or similar hurts performance so much...

There is a semi standard test called IMIX which is combination of packet sizes, specifically the latest I believe is...

64 bytes - 38%
174 bytes - 23%
750 bytes - 16%
1500 bytes - 23%

This is very popular for testing firewalls but it's hardly used for routers.

Arthalen · 17 Dec 2009 at 21:00

bigredshark said:
payload size is largely irrelevant, packets per second is the true measure of throughput on a router really. For most purposes a middle value like 500 bytes will give a decent idea of real world throughput.

The work is in determining what to do with the packet, actually doing it is easy (which is why high end routers have separate forwarding engines and routing engines). It's also why doing anything like policy routing or similar hurts performance so much...

There is a semi standard test called IMIX which is combination of packet sizes, specifically the latest I believe is...

64 bytes - 38%
174 bytes - 23%
750 bytes - 16%
1500 bytes - 23%

This is very popular for testing firewalls but it's hardly used for routers.

Fantastic, thanks for that. So in confirmation, I can take any devices PPS rating, combine it with my chosen L2 protocols MTU and find maximum possible bandwidth from that bandwidth over that medium.

What I do like about this is that were talking availble device bandwidth here, rather than available link bandwidth.

You know, slowly this is all coming together, like finally that 10,000 peice jigsaw puzzle is starting to assemble itself...

bigredshark · 17 Dec 2009 at 21:32

Well, the PPS rating is under optimal circumstances, which basically means a small number of static routes, no NAT, no ACLs etc...and if you take the maximum MTU then that's a theoretical maximum, in the real world the average MTU is going to be much lower.

Basically, it's difficult to know exactly, which is why Cisco quote those low numbers officially...