From: Marvin Greenlee (marvingreenlee@yahoo.com)
Date: Fri Sep 01 2006 - 03:16:34 ART
Although the ratios between the queues are the same,
there are some additional items to consider.
Delay
With queue sizes of 1000 9000, if both queues have
traffic queued up, traffic will be sent 1000 bytes
from queue 1, and then 9000 bytes for queue 2.
With queue sizes of 2000 18000, the relative
percentages stay the same, however traffic will be
sent 2000 bytes from queue 1, and then 18000 bytes
from queue 2. On an ethernet interface, the time to
send 18000 bytes would be ~15ms, but on a slower link,
could take much longer.
Queue Depth
Queue sizes are also affected by other values. Queue
depth specifies how many packets will actually be
queued up. If you don't have a large enough queue,
the traffic sent will be affected. The default queue
depth for CQ is 20 packets. If the size of your queue
is large, you may need to increase this. For example,
if your packet size were 300 bytes, a queue depth
limit of 20 packets would limit you to being able to
queue a maximum of 6000 bytes, even though your
byte-count for the queue is set to a value higher than
6000.
IMIX
Not all traffic is the same size. Many tests use a
mix of traffic referred to as IMIX, to emulate real
world packet size distribution. Traffic is sent using
a set distribution of packet sizes, A% at 64 bytes, B%
at 500 bytes, C% at 1500 bytes, etc. Different
vendors use different values, but for the most part,
the average packet size used works out to somewhere
around 450 bytes.
FTP
In general, assuming a reasonably sized file is being
sent, FTP is probably going to use large packets.
Skewing
With CQ, the router will not send a partial packet, it
will send the entire packet. So if your queue size is
1000 bytes, and you have a 1500 byte packet, 1500
bytes will be sent, not 1000.
So, with 1000 9000, the actual traffic sent would be
1500 9000, which would be closer to 14%/86% instead of
10%/90%.
Let's look back at the two examples, assuming the
following:
Overall traffic mix with an average packet size of 450
bytes.
10% of this traffic is FTP at 1500 bytes.
90% of this traffic is other sizes, with average
packet size of 330 bytes.
(These assumptions are just being made for
demonstration purposes. In the lab, make sure to ask
the proctor if clarity is needed. If they are not
specifying any information for the packet sizes in the
question, you don't need to worry about it. If packet
sizes are given, however, make sure that your queues
are sized correctly.)
With these assumptions, neither of these two would
provide the desired 10%/90% for traffic.
Configured:
1000 / 9000
Result:
1500 bytes sent for FTP (Skewing)
6600 bytes sent for other traffic (Skewing, Queue
depth)
Relative %:
18.5% / 81.5%
Configured:
2000 / 18000
Result:
3000 bytes sent for FTP (Skewing)
6600 bytes sent for other traffic (Queue Depth)
Relative %:
31.3%/68.7%
Recommendations:
If you are given packet sizes, queues should be a
multiple of the packet size for that queue. Also, be
aware of queue depth, and make sure that you increase
the depth if necessary for enough traffic to be queued
up by the route.
See also - Packet size normalization for custom
queueing -
Cisco - Congestion Management -
http://www.cisco.com/en/US/products/sw/iosswrel/ps1828/products_configuration_guide_chapter09186a00800ca595.html#xtocid1151315
--- sabrina pittarel <sabri_esame@yahoo.com> wrote:
> One more question....a very dumb one.
> With byte counts:
> 2000
> and
> 18000
>
> for example I would have got the same result, isn't
> it?
>
> Sabrina
>
> ----- Original Message ----
> From: Gianpietro Lavado <gianpietro1@gmail.com>
> To: uyota oyearone <spycharlies@hotmail.com>
> Cc: ccielab@groupstudy.com
> Sent: Thursday, August 31, 2006 11:08:11 AM
> Subject: Re: byte-count conversion
>
> Think of the CQ byte-count as weights; with that
> configuration FTP will
> send 1000 bytes of every 10000 bytes of total
> traffic (9000 + 1000). So FTP
> will always send 10% of the total traffic
> (1000/10000) regardless of the
> interface's speed, which in this case is 100Mbps
> (10% = 10Mbps).
>
> Regards
> Gianpietro
>
>
> On 8/31/06, uyota oyearone <spycharlies@hotmail.com>
> wrote:
> >
> > Hey guys, i been trying to do a simple conversion,
> it keep me up all
> > night
> >
> > task:
> >
> > R4(e0/0) <-->(g0/0) R6(s0/0)<-->(s0/0)BB1
> >
> > Users in vlan46(btw r4&r6) have been complaining
> about slow ftp trasfers
> > when accessing servers behind BB1. To resolve
> this, you've decided to
> > config r6, so that ftp traffic is guaranteed at
> least 10Mbps on its
> > ethernet interface connected to vlan 46.
> >
> > Donnot use CBWFQ, Assume host in our network uses
> only active FTP.
> > (ref:IEWB vol2-lab7)
> >
> > Sol:
> >
> > interface GigabitEthernet0/0
> >
> > custom-queue-list 1
> >
> > !
> >
> > queue-list 1 protocol ip 1 tcp ftp-data
> >
> > queue-list 1 default 2
> >
> > queue-list 1 queue 1 byte-count 1000
> >
> > queue-list 1 queue 2 byte-count 9000
> >
> > The issue i have is, how does 10mbps equals 1000
> byte ????
> >
> > Thanks
> >
> > Uyota
> >
> > Uyota Oyearone,CCNA,MCDBA,MCSE(Messaging/Security)
> > IT Consultant (Freelancer)
> > Computer Integrated Solutions
> > 35 Fountainhead rd,Unit 617
> > Downsview, ON,M3J 2V7, Canada.
> > Tel:(1) 416 3177045
> > uyota@hotmail.com
> > Network Architecture Technology Consultants
> Technical Support Sales
> > & Repair
> >
> >
>
------------------------------------------------------------------------
> >
> > Get MSN Messenger on your cell phone.
> >
> >
>
This archive was generated by hypermail 2.1.4 : Sun Oct 01 2006 - 16:55:39 ART