Re: GSoC idea

[dave b <db.pub.mail () gmail com>]

On 28 March 2010 06:54, David Fifield <david () bamsoftware com> wrote:
> On Sun, Mar 28, 2010 at 05:48:50AM +1100, dave b wrote:
> > Make zenmap use *less* memory and be faster.
> >
> > Currently if there are 4094 hosts (and only performing a service scan
> > on two ports) in an nmap xml scan file, 800mb + of memory can easily
> > be used.
>
> I think reducing memory usage in Zenmap is a good goal. We've added a
> lot of capabilities to it and haven't been measuring the impact on
> memory usage.
>
> > By moving zenmap to cElementTree massive amounts of memory should be
> > saved in cases where there is a large number of hosts. This should
> > hopefully also make zenmap faster in these cases.
>
> Can you substantiate your claim that cElementTree will use less memory
> than the SAX parser we use now? cElementTree, it seems, is going to
> build a complete representation of the XML tree in memory, while our SAX
> parser only looks at one fragment at a time. I doubt that XML parsing is
> the biggest user of memory; it's probably the size of the structures
> Zenmap builds from the XML (in which case changing parsers would not
> help anything). A good first step for someone interested in reducing
> memory usage would be to find a Python memory profiler and use it.
>
> David Fifield

No I can't substantiate those claims because I haven't been through
the zenmap code yet. You may be correct that it isn't the xml parsing.
So yes I think reducing the memory that zenmap uses would be a good
goal. imho the xml parsing is probably a good place to start
profiling.
_______________________________________________
Sent through the nmap-dev mailing list
http://cgi.insecure.org/mailman/listinfo/nmap-dev
Archived at http://seclists.org/nmap-dev/