[Cryptography Team] Re: SecureSession published to Cryptography repository

Robert Withers robert.w.withers at gmail.com
Tue Dec 8 14:53:23 UTC 2015


Ok, less confused now - it IS 27500%!  That's with me single-core 900Mhz 
machine. An i5 or i7 will have much better throughput me thinks, at 
least the clocking...I would love to hear some numbers...it is 13 Mbs 
now, in loopback; can we reach 50 Mbs?

One known issue: when calling setsockopt: (squeak protocol is 
Socket>>#setOption:value:, one of the values fails. Here is the code 
where  Iam calling this...

ConnectionHandler>>#newListener
     | listener |

     self flag: #todo. "listener _ self socketClass newTCP."
     listener := Socket newTCP.
     listener setOption: 'SO_REUSEADDR' value: 1.
     listener setOption: 'SO_LINGER' value: 0.
     "backlog is how many simultaneous connections to accept at the same 
time"
     listener listenOn: portNumber backlogSize: 10.
     ^listener

The stdout/stderr output from this call is:

setsockopt: Invalid argument
setsockopt: Invalid argument

Note that I also refactored teh pipeline to eliminate a few calls, 
change the protocol and rename some classes. I updated the diagram at 
https://github.com/SqueakCryptographySquad/Mushroom/wiki.

There is now a version 5 of the SecureSession package in the repository.

Thanks,
Robert




On 12/08/2015 08:33 AM, Robert Withers wrote:
> Is that a 27,500% increase? I am a little confused.
>
> On 12/08/2015 08:07 AM, Robert Withers wrote:
>> Good morning,
>>
>> I managed to separate the SecureSession code out of Mushroom and I 
>> have published it into the SqueakSource Cryptography repository. 
>> Version 4 is this release, though I expect name refactoring to occur, 
>> so just take the latest. This requires the loading of Cryptography, 
>> though I did not set a package dependency in Monticello for 
>> SecureSession. Should I?
>>
>> Note to Pharo: Please load version 37 of Cryptography.
>>
>> WARNING: the Performance test will fail without the crypto plugins 
>> installed.
>>
>>
>> Resulting performance data, with and without crypto plugins...there 
>> is 275% increase in performance with crypto plugins.
>>
>> With crypto plugins:
>>
>>     100 KB data chunks received: 10
>>     send time: 272 ms, rate: 3.676 MB/sec.
>>     receive time: 600 ms. rate: 1.667 MB/sec.
>>
>>
>> Without crypto plugins:
>>
>>     100 KB data chunks received: 1
>>     send time: 8469 ms, rate: 0.012 MB/sec.
>>     receive time: 17233 ms. rate: 0.006 MB/sec.
>>
>> Regards,
>> Robert
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.squeakfoundation.org/pipermail/cryptography/attachments/20151208/80017eed/attachment.htm


More information about the Cryptography mailing list