Re: [UDT] General questions about performance

[Yunhong Gu <yu...@la...>]

On Thu, 4 Jan 2007, Yunhong Gu wrote:

>
>
> There are many factors that can affect the transfer performance. In addition, 
> UDT is not necessarily always faster than TCP, especially on long bandwidth

Sorry, I meant "low bandwidth environments".


> environments. I do believe that you can modify the UDT code by yourself to 
> reach better performance, unfortunately this requires some understanding to 
> the code:)
>
> There are available tools using parallel TCP to boost transfer speed for home 
> network users.
>
> Yunhong Gu, Ph.D., Research Scientist
>
> National Center for Data Mining
> University of Illinois at Chicago
> SEO 700, M/C 249, 851 S Morgan St
> Chicago, IL 60607-7045
> T (312) 413-9576
> F (312) 355-0373
>
> On Tue, 2 Jan 2007, Yang wrote:
>
>> hi all, i tested sendfile/recvfile between two hosts on the same
>> university network (~1ms lag), but found the performance to be poor.
>> using http, i was able to transfer a 6.6MB file in 7.7s (~860KB/s),
>> but this took 17s using udt. one host is linux (just used 'make' to
>> build), the other is windows xp (built using visual studio).
>> 
>> i was wondering:
>> 
>> - is this result expected? i realize that my hosts have low latencies,
>> but i didn't think this explains the magnitude of the hit...nor that
>> the performance of udt would/should in any situation be worse than
>> tcp. my understanding is that a primary aim of udt is to maximize
>> throughput.
>> 
>> - is there anything (reasonably effortless) i can do to see better 
>> performance?
>> 
>> - are there other file transfer apps that do offer improved bandwidth
>> usage over tcp? i glanced at some of the papers' related works, and
>> the only other released software i found was tsunami (which i have not
>> tried).
>> 
>> - is there any reason why i should not expect to find such an app? (i
>> have zero background in networking. things i can imagine are:
>> userspace implementations have too much overhead to compete with
>> kernel tcp stack; isp's/routers prioritize tcp packets, which play
>> fairly, over other packets; etc.)
>> 
>> to give some background/a better idea of what i seek: i was ftp-ing
>> large files today between two hosts across the US at disappointing
>> rates when i got the idea to look into alternatives to tcp. i realize
>> that people have long researched protocols and congestion controls to
>> maximize throughput, particularly targeting the case of high
>> bandwidth*latency. i didn't want to subject myself to too much pain in
>> this endeavor (e.g., rebuilding kernels would be going too far), so i
>> set about looking on various open-source software sites for real-world
>> implementations using udp - and that's how i found udt. i haven't
>> found any other apps, though (except for tsunami, but i imagine that
>> udt is an improvement thereon).
>> 
>> thanks very much in advance!
>> 
>> -------------------------------------------------------------------------
>> Take Surveys. Earn Cash. Influence the Future of IT
>> Join SourceForge.net's Techsay panel and you'll get the chance to share 
>> your
>> opinions on IT & business topics through brief surveys - and earn cash
>> http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
>> _______________________________________________
>> UDT-general mailing list
>> UDT...@li...
>> https://lists.sourceforge.net/lists/listinfo/udt-general
>> 
>