[e2e] packet-pair probe implementation

[dpsmiles at turing.acm.org]

They are actually different, as Raghu has rightly pointed out. Quoting from
a writeup by Kevin Lai(Formerly at Stanford(Mosquitonet project), now at
Berkeley):

"We distinguish between the bottleneck bandwidth and the available
bandwidth of a route. The bottleneck bandwidth of a route is the ideal
bandwidth of the lowest bandwidth link (the bottleneck link) on that route
between two hosts. In most networks, as long as the route between the two
hosts remains the same, the bottleneck bandwidth remains the same. The
bottleneck bandwidth is **not affected by other traffic**. In contrast, the
available bandwidth of a route is the maximum bandwidth at which a host can
transmit at a given point in time along that route. Available bandwidth is
**limited by other traffic along that route.**
 
The question of which is the better metric can only be answered by the
application. Some applications want to know which route will give them the
minimum delay or want to use an estimate taken longer than a few seconds
ago. For these applications bottleneck bandwidth is probably the best
metric. Some applications are only interested in the best average
throughput. For these applications, available bandwidth is probably the
best metric. 

...[snip].. bottleneck bandwidth is a more stable metric and is therefore
useful over a longer period of time, and ... it bounds the available
bandwidth and can therefore be used later to more accurately compute
available bandwidth ...." 

The relevant URL is :
http://mosquitonet.stanford.edu/~laik/projects/nettimer/publications/infocom
1999/html/node2.html . I am sure there are many more.

Best,
Durga

Original Message:
-----------------
From: Raghurama 'REDDY rreddy at psc.edu
Date: Fri, 9 May 2003 9:17:54 -0400
To: nuk at comp.leeds.ac.uk, END2END-INTEREST at POSTEL.ORG
Subject: Re: [e2e] packet-pair probe implementation


>> From:	SMTP%"nuk at comp.leeds.ac.uk"  "S Puangpronpitag"  9-MAY-2003
08:08:44.93
>> To:	dpsmiles at turing.acm.org
>> CC:	end2end-interest at postel.org
>> Subj:	Re: [e2e] packet-pair probe implementation
>> 
>> > > I should have stated it as estimating the path or bottleneck
capacity.
>> > > "Available bandwidth" was definitely a poor choice.
>> > >
>> > > Thanks for the correction.
>> > >
>> > > -Joseph
>> >
>> > I think it's a common mistake to confuse between available bandwidth
and
>> > bottleneck bandwidth. In any case, available bandwidth can also be
better
>> > estimated from after astimating the bottleneck bandwidth.
>> > 2 cents!
>> 
>> Are they different? How?
>> I would explain the available bandwidth as
>>    o "the lowest link speed on the network path connecting a sender to a
>>       receiver", which would be bottleneck bandwidth".

    bottleneck bandwidth - What you can tranfer if there is no competing
    traffic on the path.

    available bandwidth - What you can transfer when there is cross
    traffic on the path.  Obviously this varies depending on what other
    traffic is going through the bottleneck and varies with time.

    In the absense of cross traffic then they are equal.

>> >From Vern Paxson Ph.d. thesis,
>>    the bottleneck bandwidth means "the fastest transfer rate the path
>>    can sustain".
>> 
>> Or, look at Keshav's and other few papers' explanation.
>> 
>> I think they would be the same thing, isn't it?


--Raghu Reddy


--------------------------------------------------------------------
mail2web - Check your email from the web at
http://mail2web.com/ .