[NCLUG] Infrastructure changes (slightly OT)

dobbster dobbster at dobbster.com
Mon Jan 14 20:21:48 MST 2002


Hello all,

I might have to seek a consultant for this, but I thought NCLUG would be worth a
try. :-)  I hope this isn't too far off-topic.

Here's the situation: I am the sole programmer/WWW developer/IT person in a
small company which sells various products (oriented mainly towards "amateur
scientists" and hobbyists.)  The business started when a friend and I
collaborated about three years ago to sell geeky stuff on a single site.  Now we
are an s-corp with seven employees, a large variety of products, and numerous
web sites.  We do the vast majority of our sales on the WWW, using homemade
perl/CGI scripts to take orders, handle inventory, etc.  We have about 25 DNS
domains, but only a few of these domains are sales oriented; most are just
informative.

Our business has been successful; our sites are getting more hits and our sales
are increasing - We are steadily growing.  Obviously this is good, but we have
reached a point where we need to revise our infrastructure to handle the sales. 
I consider myself reasonably knowledgeable about a variety of things, but I have
some pretty big holes in my knowledge, as a result of learning everything alone
(NCLUG has saved my life a few times, even though I haven't been to a meeting
yet.)  We have to change things, and I am not sure what the optimal approach is
- There are so many possibilities, and I'm on my own here.

Right now our web sites run on two servers, both running Apache.  "server #1" is
a FreeBSD system managed by a major web hosting company, with a single, but very
active, web site.  This is the one that we started our business with.  "server
#2" is an old Mandrake 5.2 system I built myself (I know, not the ideal choice),
and is colocated at a local ISP.  This machine runs all of our other sites. 
Both machines run order processing software which I designed.

This arrangement has worked tolerably well, but has its limitations.  Among
other things, my proprietary software doesn't have "shopping cart" capabilities,
and it doesn't track customer data.  Most importantly, we want to have ALL of
our products available through a single ordering system so our customers don't
have to jump back and forth between our sites.  Rather than rebuild the wheel,
we have decided to migrate to a "canned" ordering system called "EdatCat"
(http://www.edatcat.com).

Here's where the problems start.  The hosting service doesn't permit root
access, runs an old version of Apache, and doesn't have the needed perl modules
to run EdatCat, so I will have to run the software on the colocated server #2. 
server #2 is expensive to colocate - We already far exceed our bandwidth quota,
so we have to pay LOTS extra for it.  This is partially a result of web page
hits and partially because I rsync server #2 with a machine in my house (call it
"server #3") running on a local DSL-based LAN.  If we place EdatCat on server
#2, our bandwidth will increase dramatically and it will be even more expensive
to maintain.  Also, server #2 lacks high availability.  If it crashes, it will
take at least a few days to rebuild from the rsync'ed server #3, which will cost
us a lot of sales, lost email, etc.

So...  I have been thinking about various options, and wondering what our best
route is.  One possibility I've considered is to get a fractional T1 (or
something comparable; DSL won't do) in our office, create a single LAN with the
servers and our office workstations, and combine all of our current servers into
a more powerful machine.  This seems like it would be great: It would make our
office machines faster (right now they use IP masquerading to access the Net via
a 56K modem, which is a drag, but the office is somewhat remote), and we could
add high availability via RAID and/or mirroring (something we can't do to the
colocated server #2.)  The problem is, I have no idea how much this would cost,
or if it's even feasible.  I've never worked with a T1, and I'm guessing we'd
need a pretty good router as well.

Other options I've considered are: Using a web hosting service for everything;
colocating elsewhere; continuing to colocate with our current ISP, but
renegotiate our contract and upgrade the server to have high availability; or,
some combination of the various possibilities.

I estimate our total outgoing bandwidth average (on servers #1 and #2) to be
about 200 Kbps, but it peaks at a lot more.  My estimate could be way off, too;
I'm basing it on the ISP's quota sheets and output from Analog.  

Here are my basic questions:

1. Does anyone have any feedback about the various options I am considering? 
What would work best?

2. Is a T1 or fractional feasible?  How much does it cost to set this sort of
thing up?  If it is feasible, where do I start?

3. Are there other technologies which would be more appropriate in this case?

4. Has anyone out there used EdatCat?  What is your opinion of it?

5. If I get lost, I'll have to consult someone.  Any suggestions there?  I've
heard that tummy.com is good...  :-)

6. Any other suggestions?

7. Am I clueless?  (Never mind...  I've already guessed the answer to this one.)

I know I am asking a lot here; any feedback would be gratefully appreciated.  In
general, thanks for all of the help through the years.  If I can pay back NCLUG
in any way, let me know...

Thanks,

Mark A. Story
dobbster at dobbster.com



More information about the NCLUG mailing list