cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
21035
Views
28
Helpful
24
Replies

Cisco 871 Maximum WAN speed

I've just connected via pppoe to a fibre 70/40mbs service but only get around 12mbs bandwidth. The FE4 interface reports running at 100mbs Full Duplex so was wondering if the dialer interface has some kind of limit around 12mbs?

If the router is limited to 10/12mbs of WAN throughout can someone recommend and similar router with greater WAN throughput more suitable for a fibre service.

24 Replies 24

I maintain a Cisco 871 at my office, and I consistently get around 30Mbps down and 8Mbps up, but that is due to my cable companies speed limitiations applied upon me.

I know this Cisco document (routerperformance.pdf) is old but I didn't know it was that incorrect.

According to this document, the 870, with CEF enabled, can push 12.8 Mbps (in a single direction and without any encryption).

     Well that document only shows the max speed for 64-byte packets. Those are not typical of all the packets on your network, they are AWK packets and stuff like that; there are a lot of them, but they are not all there is. If you look at a Cisco 860 in table 6 of the document I included (a Cisco 860 and 870 both get 25 Kpps with 64-byte packets), you can see that when you mix up the packet sizes to include 1500-byte packets, which are large packets, the throughput increases significantly; around 27Mbps is the max  throughput in an IMIX (translate to 'real world') environment.

      Even you want to talk crazy talk, you can look at the results when using all 1500-byte packet sizes in table 1, and see that the throughput increases to 197 Mbps. Of course this is literally impossible because the ports are only 10/100, and therefore could never get more than 98.7 Mbps in the the most perfect environment.

     In conclusion, the Cisco 871 is capable or routing at 27 Mbps in a real world environment.

Disclaimer

The  Author of this posting offers the information contained within this  posting without consideration and with the reader's understanding that  there's no implied or expressed suitability or fitness for any purpose.  Information provided is for informational purposes only and should not  be construed as rendering professional advice of any kind. Usage of this  posting's information is solely at reader's own risk.

Liability Disclaimer

In  no event shall Author be liable for any damages whatsoever (including,  without limitation, damages for loss of use, data or profit) arising out  of the use or inability to use the posting's information even if Author  has been advised of the possibility of such damage.

Posting

     Well that document only shows the max speed for 64-byte packets. Those are not typically all the packets on your network, the are like AWK packets and stuff like that; there are a lot of them, but they are not all there is. If you look at a Cisco 860 in table 6 of the document I included (a cisco 860 and 870 both get 25 Kpps with 64-byte packets), you can see that when you mix up the packet sizes to include 1500-byte packets, which are large packets, the throughput increases significantly; around 27Mbps is the max  throuput in an IMIX (translate to 'real world') environment.

All true for an 860, running IMIX in a specific configuration, as documented.  However, individual performance can often vary significantly which is why we often quote worst case numbers and which is also why Cisco, in that document, recommends an 860 for WAN links of only up to 4 Mbps (although my experience is this Cisco recommendation is very, very conservative).

Of course this is literally impossible because the ports are only 10/100, and therfore could never get more than 98.7 Mbps in the the most perfect environment.

Assuming you have two 100 Mbps duplex connections, you have 200 Mbps to deal with.  Also the 98.7 Mps is data payload rate.  If jumbo Ethernet was supported, date transfer rate increases (and PPS decreases too).

     In conclusion, the Cisco 871 is capable or routing at 27 Mbps in a real world environment.

Perhaps, perhaps not.  Depends on your real world environment.

Disclaimer

JosephDoherty is full of misinformation so it is no wonder that he feels the need to post a Liability Disclaimer as a preemptive form of deniability at the beginning of each and every one of his posts! Any reader taking his posts seriously does solely at reader's own risk.

     First of all, I do not know why you are following me around to every discussion trying to discredit my observations! You are dirtying up the discussions with your misinformation, please quit stalking me.

     With that said, in my real world environment, I am reaching the limits that my ISP will allow on my cable link. Please understand that the Cisco 800 series is a SOHO series, if you don't understand the significance of that, well I'm not surprised. My router is a pretty active SOHO router because we are WISP R&D branch office, and I have several services running on my Cisco 871:

  • PAT/NAT
  • ACL
  • Encryption
  • VPN
  • QoS
  • VLANs

Therefore we consistently have the full array of packet sizes. We are a pretty average IMIX environment. If you read the paper you linked me to, you'll see that that there is no real IMIX standard, but my results are similar.

     In conclusion, the Cisco 871 is capable or routing at 27 Mbps in a real world environment.

Perhaps, perhaps not.  Depends on your real world environment.

     Where is is the perhaps not?! This is proof that you don't know your elbow from a TCP packet. MY CISCO 871 IS GETTING 27 MBPS, THEREFORE THE CISCO 871 IS CAPABLE OF GETTING 27 MBPS! There is no perhaps about it, it is happening at this very moment. You just need to admit you are wrong. YOU ARE WRONG JOSEPH. Now please quit dirtying up these discussions with your misinformation.

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

chad patterson wrote:

Disclaimer

JosephDoherty is full of misinformation so it is no wonder that he feels the need to post a Liability Disclaimer as a preemptive form of deniability at the beginning of each and every one of his posts! Any reader taking his posts seriously does solely at reader's own risk.

     First of all, I do not know why you are following me around to every discussion trying to discredit my observations! You are dirtying up the discussions with your misinformation, please quit stalking me.

     With that said, in my real world environment, I am reaching the limits that my ISP will allow on my cable link. Please understand that the Cisco 800 series is a SOHO series, if you don't understand the significance of that, well I'm not surprised. My router is a pretty active SOHO router because we are WISP R&D branch office, and I have several services running on my Cisco 871:

  • PAT/NAT
  • ACL
  • Encryption
  • VPN
  • QoS
  • VLANs

Therefore we consistently have the full array of packet sizes. We are a pretty average IMIX environment. If you read the paper you linked me to, you'll see that that there is no real IMIX standard, but my results are similar.

     In conclusion, the Cisco 871 is capable or routing at 27 Mbps in a real world environment.

Perhaps, perhaps not.  Depends on your real world environment.

     Where is is the perhaps not?! This is proof that you don't know your elbow from a TCP packet. MY CISCO 871 IS GETTING 27 MBPS, THEREFORE THE CISCO 871 IS CAPABLE OF GETTING 27 MBPS! There is no perhaps about it, it is happening at this very moment. You just need to admit you are wrong. YOU ARE WRONG JOSEPH. Now please quit dirtying up these discussions with your misinformation.

Chad, chill out, relax.  First I'm not following you around trying to discredit your observations!

I did inquire/comment about a couple of your statements in a couple of your recent postings, but if you would examine more of my postings, you'll see I asked Glen, just today, about 3750 stacks automatically copying files between member stacks.  Also recently, I commented on a posting made by Leo about 3750 requiring all stack member to have the same IOS version and feature set.  Both Glen and Leo are very long and notable posters within these forums.  In other words, I'm not singling you out.  Additionally, others have asked or commented on my postings too, especially Paolo or Rick (both Hall of Fame).  My intent, and I think this is likely true for others too, is we all want to be as accurate as possible and indeed don't want other readers or the original poster to obtain other than the best information.

Yes, yes, you're 871 is getting 27 Mbps throughput.  I don't recall ever writing I doubted your veracity.  Nor do I recall writing the 871 is incapable of obtaining 27 Mbps, if fact, I recall writing (in your query posting) it might reach close to 200 Mbps (even 300 Mbps if 25 Kpps was the actual number for 1500 byte Ethernet packets).

To your question "Where is the perhaps not?", I believe some traffic mixes with some configurations may not obtain 27 Mbps.  So yes, I agree for some traffic mixes and some configurations, the 871 is quite capable of supporting 27 Mbps, but this is not guaranteed for ALL traffic mixes and ALL configurations; i.e. the perhaps not.  You disagree?  Ok, that's fine too.  We've both stated our reasoning, readers can agree or disagree with either of us.

BTW, I see my other postings in this tread have received "not helpful" ratings.  Chad, assuming these ratings are from you, you can rate any posting as you desire, but if you read the guidelines for using these forums, vindictive ratings might cause action against you.  So try to participate in the spirit of these boards, which is trying to help others.

Disclaimer

The  Author of this posting offers the information contained within this  posting without consideration and with the reader's understanding that  there's no implied or expressed suitability or fitness for any purpose.  Information provided is for informational purposes only and should not  be construed as rendering professional advice of any kind. Usage of this  posting's information is solely at reader's own risk.

Liability Disclaimer

In  no event shall Author be liable for any damages whatsoever (including,  without limitation, damages for loss of use, data or profit) arising out  of the use or inability to use the posting's information even if Author  has been advised of the possibility of such damage.

Posting

leolaohoo wrote:

I maintain a Cisco 871 at my office, and I consistently get around 30Mbps down and 8Mbps up, but that is due to my cable companies speed limitiations applied upon me.

I know this Cisco document (routerperformance.pdf) is old but I didn't know it was that incorrect.

According to this document, the 870, with CEF enabled, can push 12.8 Mbps (in a single direction and without any encryption).

Leo, what might be questionable is what's being measured for "bandwidth" for a specific forwarding PPS, not so much that it's incorrect.

Cisco documents the same minimum forwarding rates, but the old document assumes a packet is just 64 bytes, while the newer document references RFC 2544.  RFC 2544 specifies certain frame sizes on specific medium, for example 64 bytes for Ethernet and Token Ring.  Actual PPS rate, for wire-rate or line-speed Ethernet has overhead.

So 25 Kpps, if just 64 bytes per packet is, 25,000 * 64 * 8 = 12.8 Mbps, as noted on the old Cisco reference for "bandwidth".

But if you were going to send/receive minimum size Ethernet (64 bytes) at 25 Kpps, actual (Ethernet) bandwidth consumption is higher.  As wire-rate/line-speed 100 Mpps Ethernet is noted for 148.8 Kpps, 25 Kpps on this measure (Ethernet) "bandwidth" would be 25/148.8 or 16.8% or 16.8 Mbps.

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Hi all, this is slightly off topic for the original posted question, but I wanted to elaborate a little more about performance interpretation of "bandwidth" so expected device performance might be correctly understood.

We very often reference Cisco's (old) http://www.cisco.com/web/partners/downloads/765/tools/quickreference/routerperformance.pdf.  In it we see columns for Mbps, which the document's header explains (excluding 12000 Engines 0, 1, 2, 3 and 5)  is the PPS rate * 64bytes * 8bits.  This is all very well and good, but can be misleading when trying to determine whether a router can handle certain link bandwidth loadings.

If we wanted to know if a router's PPS could deal with 10 Mbps, full duplex, Ethernet, minimum size packets, we might just simple look at the Mbps column and see if the rate was listed at 20 Mbps or higher.  Seeing a value less than 20 Mbps, we might then conclude that the router was unable to handle 10 Mbps, full duplex, Ethernet at line-speed or wire-rate.  However, just looking at the Mbps column, for 20 Mbps, we would be incorrect.

Wire-rate/line-speed 10 Mbps Ethernet, for minimum size packets, is 14.88 Kpps.  So 10 Mbps (20 Mbps total) duplex Ethernet would need a PPS 2x 14.88 Kpps or 29.76 Kpps.

However using the Cisco's document's bandwidth conversion, 20 Mbps needs 20,000 / 64 /8 = 39.0625 Kpps.  I.e. much higher than we need!

Conversely, Cisco's document, for 29.76 Kpps, would define bandwidth, for 29.76 Kpps, as (only) 29760 * 64 * 8 = 15.24 Mbps, incorrectly implying this router could not deal with line-speed/wire-rate minimum size packets for 10 Mbps full duplex Ethernet.

Is the Cisco document wrong?  No, but it's wrong to just apply bandwidth values in that document to media bandwidths.

Basically then, Cisco routers have more actual media "bandwidth" capacity than the listed bandwidths as the document's bandwidths don't account for L2 overhead, which actually lowers the PPS requirement (especially for small packets where L2 overhead is proportionally greater).

In the above, I only used Ethernet as an example, but other media have different L2 overheads which will skew the required PPS rate to support line-rate/wire-speed.

Also in the above, only PPS rate for simple packet forwarding is addressed, but of course, CPU is also used for other functions of the router.  How much CPU might be required for these other functions is also very, very difficult to predict.  Impact of actual traffic can be also very, very difficult to predict, even how much might be process switched (which on software based routers can use much, much more CPU per packet).

Generally, when estimating performance, you often want to be very generous (this may also explain Cisco's often very conservative recommendations) or you'll need to try your config with your traffic on an actual device and see whether it performs as needed.

Disclaimer

The  Author of this posting offers the information contained within this  posting without consideration and with the reader's understanding that  there's no implied or expressed suitability or fitness for any purpose.  Information provided is for informational purposes only and should not  be construed as rendering professional advice of any kind. Usage of this  posting's information is solely at reader's own risk.

Liability Disclaimer

In  no event shall Author be liable for any damages whatsoever (including,  without limitation, damages for loss of use, data or profit) arising out  of the use or inability to use the posting's information even if Author  has been advised of the possibility of such damage.

Posting

Just one more item I wanted to mentioned about a router's performance, for the same hardware and same traffic, different IOS versions may (unlikely though) perform (very) differently.

For a WAN link of 70/40 Mbps capacity, you'll probably want at least an 890 series.

BTW, what's the CPU usage look like when you hit peak performance?

     Mr. Joseph Doherty, You seem to be rather argumentative. You dance around the truth. Either a car can go 30 mph or it cannot go 30 mph; there is no perhaps or perhaps not about it. If a car can go 30 mph, then you are incorrect to say that the car cannot go 30 mph, which is what "perhaps not" implies. In truth, you are limited to saying that "the car is not going 30 mph, but it is able to go 30 mph". If the car is not going 30 mph, then you need to shift gears (or perhaps get on a paved road).

     You Joseph, refuse to make a similar truthful statement about the Cisco 871 router. The Cisco 871 router IS capable of going 30 mbps. Period. It may not be going 30 mpbs, but it most certainly capable of this. How can I make this statement with such confidence? Because my Cisco 871 is doing 30 mbps as we speak. So there you have it. The cisco 871 is capable of going 30 mpbs. If the router is not going 30 mbps, then you need to adjust the configuration of the router (or perhaps adjust the network).

      And this should most definately be the final word on this issue

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Yes, completely understand how information contrary to your beliefs can appear argumentative.  When you write "You dance around the truth.", objectively, might not that be seen as flaming me?  So who really is being argumentative?

Regardless, the truth is, on software based routers, throughput performance varies, often much, by the "kind" of traffic and how your router is configured.

Since you used an analogy of a car being able to go 30 MPH, what about on an uphill grade.  Is there any speed difference between having just one passenger, or all passenger seats filled and luggage too?  How well does it run at 10,000 above sea level?

There's a big difference between whether your car is capable of doing 30 MPH and whether it can do 30 MPH regardless of any other factors.

Are these considerations real?

Last year I was involved in analysis of a 7206 with a NPE-G2 being crushed by a couple of 100 Mbps of bandwidth (on a gig link).  Normally this router has no difficulty handling this volume of traffic.  What was different this time?

Well, VoIP had just been activated (different kind of traffic) and another engineer had installed a lengthy ingress service policy to count every possible DSCP marking.  Simple removal of the service policy saw the CPU go from 100% to 30%.

I was also involved in analysis of a pair of 2821 highly loading their CPUs (60 to 80%) with only 20 Mbps of traffic.  These devices have much more performance capacity than 871s.  What was the issue this time?

These routers were supporting GRE/IPSec tunnels.  Someone had added, downstream, about 10 security camera, all streaming about 1 Mbps each.  Problem was, this traffic was UDP and didn't care about PMTU.  I.e. all the large UDP packets were being fragmented.  Fix was enabling JumboEthernet on the "WAN" side, so standard 1500 MTU could be supported without fragmentation.

The truth is an 871 can't guarantee 30 Mbps under all possible conditions.  That's why in the attachment, Cisco recommends a similar performing 860 for only 4 Mbps and the even faster 880 for only 8 Mbps.

BTW, I'm the principle engineer for about 600 Cisco routers and switches, ranging in size from 871s up to 6500s with close to 700 ports each.  Probably have over a hundred 871s, so I'm not totally unfamiliar with the platform or its performance.

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community:

Review Cisco Networking products for a $25 gift card