On Tue, Sep 09, 2003 at 03:16:08PM -0700, Ben Barrett wrote:
> On a related note, I just read a review for Omnisight, an enterprise
> log-file analysis framework, in the Sept. 1 InfoWorld (free sub), and
> was flabberghasted.  They go on about the impressive performance of the
> system, which costs a mere $124K, but don't mention the specs of the
> system they run their huge queries on.  The framework inserts logs into
> SQL, where analysis is easy (as SQL allows, at least).  What bugs me is
> that this seems like nothing -- it requires significant development to
I'm sure it is nothing.  They charge that much because they believe
people will pay it.  People will pay it because they don't know any
better, because they are being marketed to about some big time solution.
They will pay that much because people inherently believe that a product
with such a high price tag must be darn good, and do a lot.  We paid a
high price tag for such a product.  What we got was a pretty lame
product.  It took OSS for us to realize that price has no relation to
quality, functionality, security or performance.  For the suckers that
pay that price tag, they have some more lessons to learn.

> much.  It does seem to offer, built in to the "framework",
> cluster-friendly management and reporting... but gah, the basics seem so
Sure, framework, let's see ha-linux, postgresql, raid.

> It has a lot to do with education, of course; 
Yes, education that 1) we can and should expect programs and computers
to work in a certain way, like OSS does and much MS software does not,
2) price has no relation to the above list, 3) OSS is cool.

> Now for a fun story, slightly related:  While observing the network
> operations at burning man roughly a week and a half ago, there were some
> problems, which were somewhat mysterious (Clif might clarify, though) --
> someone had some kind of virus or worm on their windows system which
> somehow damaged the [debian] router's ability to maintain TCP routing;
> although UDP kept working fine, so the VoIP phones kept working
> perfectly.  
Based on this information, I would say the windows machine probably had
MSBlaster, Welchia or a variant and the debian router had it's state
table filled to overflowing.  There is no state in UDP.  VoIP and other
realtime media protocols usually use UDP because it is not reliable and
they don't want every packet to get there.  They want the packets on
time or not at all, no delays as allowed in TCP.  For a nat router it
requires state.  A worm like Welchia (which I had on my network a few
weeks ago) tries to ping 65,536 ip addresses within about 20 minutes,
looking for machines to infect.  Waits a little while, then tries again
with a new set of ips.  

I don't know about autodetection and autodefense around it.  I suppose
snort might do something, but that isn't an area that I've learned about
yet.  I discovered it with tcpdump.  I then alleviated the sitation by
entering in a firewall rule that disabled ICMP from that machine going
out to the internet.  If I were there, I'd do something similar:
tcpdump, then manually modify some firewall rules.

> Thanks for putting up with another lengthy post,
gaaah

Cory

-- 
Cory Petkovsek                                       Adapting Information
Adaptable IT Consulting                                Technology to your   
(541) 914-8417                                                   business
[EMAIL PROTECTED]                                  www.AdaptableIT.com
_______________________________________________
EuG-LUG mailing list
[EMAIL PROTECTED]
http://mailman.efn.org/cgi-bin/listinfo/eug-lug

Reply via email to