[e2e] TCP un-friendly congestion control

[Ibrahim Matta]

Injong Rhee wrote:
> Hi Ibrahim,
> 
> Please correct me if I am wrong (I read the previous version of your
> SIMD paper -- I haven't read your newer version). I like the paper and
> especially, the fact it utilizes history and its t^2 behavior. However,
> the protocol seems to target to match TCP throughput under a loss rate p
> by adjusting its CC parameters (alpha and beta).  But if it matches TCP
> throughput under loss rate p, then I am not sure how it can perform
> better than TCP under fast, long distance networks (where it is known
> that TCP is not doing well) while being fair (or compatible) to TCP.
> Maybe I need to read your paper again. Sorry if I am wrong. 

Injong, "better" would be in terms of convergence speed to
its fair share, or how fast you can grab available capacity...

ibrahim




> 
> Just curious and lazy...
> 
> Injong Rhee
> Computer Science Dept
> North Carolina State Univ.
> rhee at csc.ncsu.edu
> http://www.csc.ncsu.edu/faculty/rhee
> 
> -----Original Message-----
> From: Ibrahim Matta [mailto:matta at cs.bu.edu] 
> Sent: Friday, June 06, 2003 12:59 PM
> To: Injong Rhee
> Cc: 'Tom Kelly'; 'Michael Welzl'; end2end-interest at postel.org
> Subject: Re: [e2e] TCP un-friendly congestion control
> 
> Injong Rhee wrote:
> 
>>Hi.
>>
>>It is my understanding that in the environments where these protocols
>>(HSTCP, Scalable TCP, and FAST) are intended to operate -- fast
> 
> (>1Gbps)
> 
>>and long distance networks (>50ms) -- TCP friendliness is not a
> 
> concern
> 
>>since TCP cannot fully utilize the available network bandwidth. These
>>protocols try to overcome this under-utilization problem of TCP.
>>However, if we try to apply this to the general Internet environments
>>(as it is noted in the announcement by CalTech), I think there could
> 
> be
> 
>>some TCP-fairness issues.
>>
> 
> 
> It is possible to design more aggressive protocols
> that are also TCP-friendly. E.g. see our
> history-based SIMD proposal at
> 
> http://csr.bu.edu/simd/
> 
> A wider class is defined in:
> 
> http://www.cs.bu.edu/techreports/pdf/2002-027-spectrum-tcp-friendly.pdf
> 
> 
> Best regards, ibrahim
> 


-- 
--
Ibrahim Matta       Dept of Comp Sci, 111 Cummington St, MCS-271
matta at cs.bu.edu     Assoc Prof, Boston Univ., Boston, MA 02215
Tel: (617)358-1062, Fax: (617)353-6457, URL: www.cs.bu.edu/fac/matta/