OK that's fixed
Mr Self Destruct
tagada at tagada.com
Wed Jun 9 16:45:56 EDT 2004
I finally read the script after some rest :o) ... and I find an error in
it...
On Wed, 09 Jun 2004 00:18:46 +0200, Mr Self Destruct wrote:
> On Wed, 09 Jun 2004 00:06:37 +0200, Mr Self Destruct wrote:
>
> I forgot some important things :
>
> I'm using the MIT sources, compiled with the option not to install
> kerberos 4.
>
> 2) I'm using the kerberos ftp deamon for my benchmark on kftp (not proftpd)
>
>> Hello,
>>
>> I'm currently writing benchmarks to compare the speed of ftp vs kftp. I
>> have three computers on a 100 mb/s LAN. One is the kdc, one is a client,
>> the third is the ftp server. All computers are running on Debian 3.
>> 1) The benchmark for ftp is based on the command "time". I'm timing the
>> whole process of logging till bye is typed. All is made with automatic
>> Perl scripts even the ftp commands typing. The server uses the proftp
>> deamon.
>> 2) I'm using the same process for kftp. I do the kinit before
>> timing.
>>
>> I have a sample of files of different size. Each file is downloaded ten
>> times, then I only keep the mean of the time transfer.
>>
>> Here are the results in seconds :
>> FTP
>> 16 bytes 0.0404
>> 128 bytes 0.04054
>> 1 KB 0.040154
>> 8 KB 0.0412154
>> 64 KB 0.05382154
>> 512 KB 0.119682154
>> 4 MB 1.5084682154
>> 32 MB 11.86024682154
>> 256 MB 148.321224682154
>> 2 GB 1197.84332246822
>>
>> KFTP
>> 16 B 0.0277
>> 128 B 0.03217
>> 1 KB 0.032817
>> 8 KB 0.0318817
>> 64 KB 0.03598817
>> 512 KB 0.084298817
>> 4 MB 0.5235298817
>> 32 MB 11.27425298817
>> 256 MB 142.527725298817
>> 2 GB 1153.79037252988
>>
>> As you can see, the transfer time looks slightly better with kftp,
>> a fact that I find hard to believe because there's more data transfer due
>> to the TGS request and replies and the session key transfer between
>> client and server.
>>
>> As someone any explaination or hints for me ??
>>
>> Thank you !!
>>
>> Mr Self Destruct
More information about the Kerberos
mailing list