I am interested in building a grade A tier 1 personals website. I
don't want to cut any corners though I'm not looking to burn through
cash wastefully. I received a fantastic answer to a previous question
(personals website question # 3) and want to build on that answer
here. In question # 3 the researcher broke the question into several
different points, please refer to that question throughtout your
research. I would like for you to refer to "personals website question
# 3" and the answers to number 3... Database development and tech
work, and number 4... Web hosting. Please expound on these answers. I
have also talked to a guy that handles computer work for me and he put
together a sheet on what he feels like I would need. He thinks I
should use a Dell 350 web server which he says would have a user
capacity of about 1,000 and another Dell 350 SQL DB server which would
have a user capacity of 10,000. He says that I should add another Dell
350 web server for each 1,000 additional users. The researcher who
answered question # 3 believes that I should initially rent and after
the site is doing well that I could look into buying. The computer guy
I know says that I should absolutely own all of the equipment and
initially let someone (i'm not sure of the terminology he used)
co-host it. If that sounds dumb or wrong then forgive me I am a
technical idiot. Anyway the guy I know says to own the equipment and
let this company we know of run everything until we grow to very large
site as which point I could consider putting together a data center
which he says can cost anywhere from $ 50K to 350K. Don't spend much
time on the data center aspect of what i've written here, that appears
to be way down the line. Additionally this guy I know said that at
10,000 users I would need a more robust SQL server. I know that many
of the sites allow everyone to browse for free and many allow everyone
that is eligable to join their site to post a profile and possibly a
picture or two without charging for membership until they want to
communicate with another member. Many of these sites say that they
have a million, two million or more people that browse their site and
post profiles and pictures. How does this play into the amount/cost of
hardware needed and hosting costs. Is the new server per 1,000 users
perspective talking about the ability to handle 1,000 users at a time
and not the thousands or millions of profiles that people have on the
system. I know I am skipping back and forth from answer # 3 and # 4,
please indulge my lack of knowledge and when responding put it back in
the proper context. I need to know your perspective of what the costs
of hardware would be if I buy and rental costs if I don't. I need to
know the monthly costs that it would take to host this site and your
justifications behind those costs. Now i'll show you how technically
ignorant I am..what about software..are there software costs for this
stuff beyond the development and if so what? would it have to do with
the number of users on the site? Okay back to the database development
and tech work...in question # 3 the researcher (John) estimated that
it would take about 500 hours at $ 60 an hour for initial development
with maybe another 100 hours of post-work, troubleshooting etc. The
computer guy I know feels like it will take about 150 to 250 hours
more overall and that the tpye of people needed would be more
expensive per hour. I need you to be more specific on these hours and
costs. Remember I want to build a grade A tier 1 site without burning
through unnecessary cash. The researcher from question # 3 said that I
might eventually consider migrating to Oracle or something more costly
in the future. The guy I know says I would never want to go to Oracle
because of the costs of the software and the costs of the people to
put it all together and maintain it. I need you to go through each
aspect of what i've written here, break it down and put it all back
together. I want a site that is technically proficient, processes
information quickly and provides quick and reliable results. Please
include a justification for the number of hours you feel are needed
and how much of a buffer you are adding to ensure success. I am
spending quite a bit of money on these questions and I expect to have
very professional, knowledgable and in-depth answers. If any part of
this question seems to be unanswerable then post your thoughts so that
I can see how I can receive what I need and you can feel confident
that you can provide the best answer in a way that I will feel
satisfied. Finally if you are able to communicate with other
researchers that are answering other questions you might want to
collaborate with the person that decides to answer my question
"personals website # 4".

Thank You,
Joel

---

Clarification of Question by joel1357-ga on 25 Oct 2002 01:16 PDT

Throughout my question I asked that you refer to my question
"personals website # 3..and the answers to # 3 and 4 from that
question. I made a mistake..I should have said...the answers to # 4
and 5...please take this into account when answering my question.

Thank You,
Joel

---

Request for Question Clarification by webadept-ga on 25 Oct 2002 02:15 PDT

Saw that Joel, got it covered, I'm working on your answer now. Should be done soon.

webadept-ga

Hi again Joel! 

Not sure if I'm happy or not about pulling this one from you, as I do
this type of stuff for my "day job". Happy because I know all the
answers :-) .. but it is what I do most of the time, (amongst other
related things) so it's not really a break. :-) Anyway, let's go
through this.

I think this question is referring to 4 and 5 of that answer you got
from : shiva777-ga


4) Database Development and Tech Work - This will be your largest
development expense. My best guess would be about 500 man hours at $60
an hour for initial development with maybe another 100 hours of
post-work, troubleshooting etc. The MySQL database is included with
most web hosters and should work fine for your needs. You can
eventually migrate to Oracle or something more costly in the future if
you need to.


5) Web Hosting - I don't recommend starting off buying your own server
when you can rent one much more easily and have all the technical
issues taken care of. When your site is absolutely cranking, owning
your own might be something to look into. To start off you can start
by renting a dedicated server for about $250 a month and scale up as
needed. Eventually it will cost you thousands of dollars a month, but
by that time you should have plenty of subscribers!

Let's start with your questions on #4 


a) Additionally this guy I know said that at 10,000 users I would need
a more robust SQL server.

MySQL is free and because of that many misunderstand it to be a
non-robust system. This isn't the case. If you were doing a Bank
Transaction type service then MySQL wouldn't work for you, the demand
would be for something like Oracal or DB2 or something of that nature.
MySQL is designed to be fast, light, powerful, and for the web.

MySQL Powers Yahoo Finance
http://www.mysql.com/articles/us/yahoo_finance.html

U.S. Census Bureau Reaps Awards from MySQL-based Web Sites
http://www.mysql.com/articles/us/us_census.html

MySQL Matters to Slashdot
http://www.mysql.com/articles/us/slashdot.html
-------
Slashdot utilizes four MySQL databases: a four-gigabyte, four-CPU
primary database that performs most of the data reads and writes,
another for replication, one for static content and another for
back-up. MySQL handles about 360 queries per second, running on Intel
hardware with the Linux operating system. "I have been using MySQL for
a number of years, and I continue to be amazed at how well it scales,
noted Aker.
--------

If you ever need something more powerful that MySQL for a webserver
you are doing something horribly wrong somewhere. Slashdot has well
over 10, 000 users and much more traffic than you could see in the
next few years. So, MySQL will be fine for a while.



b) I know that many of the sites allow everyone to browse for free and
many allow everyone that is eligable to join their site to post a
profile and possibly a picture or two without charging for membership
until they want to communicate with another member.

This isn't that much of a concern. The type of programming this is
called is "dynamic" pages. The user has a record in the database for
his/her profile. The page is a single template. When you ask to see a
user's profile, the single page is filled in with the record data and
presented to the browser for display. Using a language like PHP for
this you get really fast pages with little strain on the server.
Slashdot uses Perl for this, and you can go through some of the
profiles over there to see this type of thing in action. There might
be one or two static pages on that site but everything else you see
there is being pulled from a database and presented in templates. In
fact it's an OpenSource project and you can download all the code that
creates that site.

I have a site with a forum on it, which has a size of  509, 760 for
270 users in the user table. The whole database for this forum with a
total of 18119 articles is 26 megs total. This forum is very extensive
in the information it keeps. Its called PHPbb and you can see this
program (not mine, theirs) running at
http://www.phpbb.com

If you run a forum for these folks, then you might be looking at
similar stats for size per user. But otherwise, I wouldn't worry about
it too much.

By the way, you might want to look into a few OpenSource projects on 
http://freshmeat.net  
They probably have this made already and you can alter it instead of
developing from the ground up. Just a thought. I'll list a few
projects down below to get you started in that area.

c)  Many of these sites say that they have a million, two million or
more people that browse their site and post profiles and pictures.

Umm.. yeah, sure. That is rather an amazing number there. Two million
people on a site is quite a few. I would expect that the real figures
fall down into the 10k to 50k on a busy day. They may have had 1m or
2m users come and see their site, and even browse, since they started.
That's a real hard number to verify as well. I don't think Slashdot
has 1m users hitting them in one day, and I would be surprised if
Yahoo did. I'm not saying they are lying, just looking at the stats in
a way to show the most effect. Let's take my little website that I
host, (well one of them anyway) the one with the forum on it.

Now we already know that I have 270 registered users over there.
That's how many can really use the forum, and since its hidden and not
indexed on the Google or any place else, we can be relatively sure
that this is the greatest amount using it on any one day. Its really
much less than this, but let's give it the greatest number available.
Then lets add 20 users a day (way out there) who come to the main page
areas off Google and search engines or just stumble across the main
site.

The stats for this site currently for the month of October only are 

Unique visitors 	Number of visits 	Pages 		Hits 			Bandwith 
1842			5563			 100738		1005469		4.12 GB
			(3.02 visits/visitor)	(18.1 pages/visit) (180.74 hits/visit)
	(776.67 KB/visit)

Now if we look at the Number of Visits and the Hits, that looks really
impressive. But, Unique shows a much different story, and that's where
the real numbers are. So I can tell you that I have about 6k visitors
a month, but it's not really true. And the : Most users ever online
was 21 on Fri Jun 28, 2002 12:54 pm in the Forum area, the  busiest
area of my site.

Bandwith is another story. Like I said this site has some extensive
data being put in and read, so it gets used a lot. Much more than
probably most forums of its size. That's because the users are rather
prolific writers.


d)  How does this play into the amount/cost of hardware needed and
hosting costs.

With all this in mind we switch theories in mid stream. It is success
that will kill you faster than anything else. There are horror stories
of websites getting mauled by their users and crashing into horrible
deaths. Speed of Bandwidth is the thing to be most concerned with. A
good server with a couple of fast CPU's in there and 500m+ Ram will
handle a lot of users, much faster than a T1 can get them in there.
You want the fastest Network cards in there that you can get and the
fastest CPU's you can find. Disk space and RAM is cheap. Get a lot of
it.

Your ISP solution needs to be able to get you a dedicated T1 and be
able to go to a T3 if necessary in a rather short amount of time. Your
costs will be based on bandwidth used. Most ISP's give you a set
amount you can use a month and then tack a cost on for each GB used
after that. I don't know who you are using, so I can't do the math for
you, but with the stats I've given you above you should be able to
figure it out pretty easy.

e) Is the new server per 1,000 users perspective talking about the
ability to handle 1,000 users at a time and not the thousands or
millions of profiles that people have on the system.

I would say yes, this is 1,000 per hour probably or maybe even every
15 minutes at once. A hit is really fast stuff. A large(bulky) forum
page on my site is 119 KB (122,067 bytes)  ... * 1000 users at one
time is 119 megs, so our bandwidth is straining, but our server is
not. 120 megs every 15 minutes, for 18 hours (you'll have slow points
of at least 6 hours a day)  is 8,640 megs, or 8 gigs a day.  Remember
that "Slashdot utilizes four MySQL databases: a four-gigabyte,
four-CPU primary database that performs most of the data reads and
writes, another for replication, one for static content and another
for back-up. MySQL handles about 360 queries per second, running on
Intel hardware with the Linux operating system."  and Yahoo "Our needs
vary considerably, and that is one reason that MySQL works well with
us. It handles the high-volume, 260 million record tables just as well
as the small, low-volume ones. One of our servers handled over a
quarter of a billion queries in a month-and-a-half, and it still has
capacity to spare."


f) the researcher (John) estimated that it would take about 500 hours
at $ 60 an hour for initial development with maybe another 100 hours
of post-work, troubleshooting etc. The computer guy I know feels like
it will take about 150 to 250 hours more overall and that the type of
people needed would be more expensive per hour. I need you to be more
specific on these hours and costs. Remember I want to build a grade A
tier 1 site without burning through unnecessary cash.

If you have this type of traffic running through your site, and this
many users on there, you have a full time database guy on staff, who
is costing you $50.00 -$60.00 per hour. This SysAdmin guy knows the
primary language you are using for development (probably PHP), Perl,
and MySQL, as well as the Linux system. She is keeping tables clean,
updating indexes and creating new areas that were not thought of in
the first draft of the system. He's also running down bugs and keeping
the site moving as fast as possible.

The database isn't going to take 500 hours to develop. That's going to
be much faster than that. Tieing all the pages and areas and forums to
the site is going to be where the time is taken up. Shouldn't take
more than 80 hours to come up with a good database schema for your
site. That person is probably going to cost $60.00 per hour and will
design it so that new tables can be added in the future.  The 100
hours of post-work, troubleshooting etc is reasonable, perhaps even up
to 200 hours. Any project like this is going to have at least 5 things
go wrong, that just couldn't be seen at development time.

Your programmer(s) are going to be working with the database guy for
at least 2 weeks before he's ready, adding  80 hours to that as well.
That's the planning stage, going through the features you want and
trying to make that work for you. So total database development is
about 160-200 hours. If you want to budget 500 hours for him, that's
fine, but I would be wondering what the heck he was doing for all that
time.


e) The researcher from question # 3 said that I might eventually
consider migrating to Oracle or something more costly in the future.
The guy I know says I would never want to go to Oracle because of the
costs of the software and the costs of the people to put it all
together and maintain it.

I think we covered this pretty good already. Oracle isn't an answer
for you for this, but it is a lot of money.



a) He thinks I should use a Dell 350 web server which he says would
have a user capacity of about 1,000 and another Dell 350 SQL DB server
which would have a user capacity of 10,000. He says that I should add
another Dell 350 web server for each 1,000 additional users.

The Dell 350 is a good machine. There are several out there which are
just as good. If you are going to rent a dedicated server from an ISP
there are several benefits to this. One, if your machine goes down,
then they have backups which will "kick in" and keep your site up.
These machines run 24/7 and when they go down they go down hard.
Normally it means a hard drive or a system board has to be replaced.
If you own this machine and are hosting it at a data center, this
means you have no less than 6-8 hours of downtime. Not really a good
thing. If you are renting a dedicated server, then the down time is
normally much less, because they switch your IP over to another server
while they get yours fixed. This is a huge benefit and a large
argument for renting and not owning.

Your downtime is in travel time to the site, diagnostics, getting the
part ( you have those right?) getting them in there, bringing the new
system up, running tests and then going live. Like I said, if it went
down, then there is probably something very wrong. Uptime is
important, but that's kind of  a given. Some servers come with On-Site
repair warrantees which is great, IBM is famous for this type of
service, but you are still down while you wait for them to show up.

Linux is probably the best OS for this type of system, or a UNIX of
some sort. Linux will keep your costs way down and it's fast.
Microsoft is out for this. Not because I don't like Microsoft, just
because it's not the right tool for the job. Dell now supports Linux
installs and will sell them that way.

Dell 350 web server Review

http://www.epinions.com/content_56490036868
----
The cost of this server is one of its biggest strong points,
especially for a department/company that needs numerous servers for a
server farm or a department that does not have tons of money to burn.
The 350 is quite capable of being a nice workhorse server, and for the
price, there's nothing that even comes close.

I would recommend this server to anyone who needs quantity versus
quality, such as a beta-testing center that needs numerous computers
for little money. It's also a great web server. I would also recommend
it to anyone who needs a basic server.
----

I would recommend looking at a few more systems before making a choice
on this. Here are a few which have had some good success

Penguin Computing
http://www.penguincomputing.com/store/index.php?PHPSESSID=53100113398f5af45158abc4aa05dbb9&

IBM Linux xSeries
http://www-1.ibm.com/servers/eserver/linux/xseries/
with these
http://www-132.ibm.com/webapp/wcs/stores/servlet/CategoryDisplay?categoryId=2063529&storeId=1&catalogId=-840&langId=-1

I'm a big fan of IBM since that night when my AS400 tape drive went
out at 11:00pm and the guy showed up at 11:45pm  with a new one to
install for me. That was really cool, and I like service like that.
You pay for it, but there's no substitute for up-time.

A good server like this is going to take on 10k users a day, with room
to spare. Not 1k. Again, your limiting factor is bandwidth.



b) I need to know your perspective of what the costs of hardware would
be if I buy and rental costs if I don't.

I like the rental idea much better than the purchase idea to start.
The last thing you want to worry about when getting a site this large
up and running is hardware problems. Once the software is done and
working and you get 10k users on there pushing the code, then start
looking at taking on the extra strain of dealing with the hardware
area as well. Hardware is cheap. IBM's with all the fat on them is
costing you about 1.5k a server. That's not the worry. The worry is
the man hours maintaining those bad boys. System maintenance is a time
consuming project with 10,000 users on there. You are looking at
hiring at least one SysAdmin for $50.00 per hour to keep three servers
running as well as they can. Compare that to dedicated server hosting,
and really it's no contest.


c) I need to know the monthly costs that it would take to host this
site and your justifications behind those costs.

If you have 10,000 regular users plus the many thousand snoopers and
looky-loos you are looking at about 150 -200 gigs of bandwidth a
month. There's your big cost, and I don't see that happening for a
while. This is based on a site that has huge pages for text and
communication between people with lots of places to add more. If you
went over 5 gigs a month for the first 6 months I would be very
surprised, but we'll plan for success.

Option Description Type Amount 
Dedicated Server: Default Options monthly $80.95 

Bandwidth: 150 GB data transfer line speed 	5 Mbps Billed at $3 per
Gigabyte over 150 GB monthly $825.00

Processor: P4 1.8 GHz monthly $160.00 
Memory: 1024 Mb monthly $125.00

Storage: 1 x 40 GB EIDE 5400 RPM monthly $20.00 IP 
Space: /29 Network - 5 Useable IPs monthly $10.00 
Backup: 1 GB monthly $30.00 IP 
Monitoring: Disabled monthly $0.00 
Cold Remote Reboot: Disabled monthly $0.00 
Firewall/IDS/Traffic Shaping/Load Balancing: Disabled monthly $0.00
Email Accounts: 0 
Email Accounts monthly $0.00 
Email Quota (per Email Account): 0 Mb Email Quota monthly $0.00 
Operating System: Redhat Linux 7.x standard install one-time $25.00 
DNS Management: 1 Entry one-time $0.00 
Turnaround: 48 hour turnaround one-time $0.00 
Email Domains: 0 Email Domains one-time $0.00 
Setup Fee: 150 GB data transfer line speed 5 Mbps one-time $400.00 
Setup Fee Service Initiation one-time $127.28 
Setup Fee 1 x 40 GB EIDE 5400 RPM 
Initialization one-time $20.00 
Setup Fee 1 GB Initialization one-time $30.00 
Promo: (usa promo): 10% discount monthly $-96.60 
Promo: (usa promo): 10% discount one-time $-60.23 


Monthly Charges $1,125.85 
One-Time Charges $652.57


Got this off this site, they have a good calculator and the prices are
middle of the road. You can go there and input this info, and add
email address and that kind of thing
http://www.nyi.net/dedi_calc.php

That's one heck of a machine with oogles of bandwidth, and we are
looking at a weeks salary for one guy to take care of this after you
purchase the hardware and buy the bandwidth. Rent it. Really. Get a
good dedicated server at an ISP in your area and let them take care of
the day to day needs of the hardware and system. They have the
resources and you don't. Your Computer guy can probably do it, but we
are talking about a lot of work, and things happen. Put him to better
use with the database and the web design area.

Additional Areas 

a) what about software..are there software costs for this stuff beyond
the development and if so what? would it have to do with the number of
users on the site?

No. All your costs are in the development area. Everything else is
either in the system (linux) or available for free.

b) Please include a justification for the number of hours you feel are
needed and how much of a buffer you are adding to ensure success.

I think I covered this, but if you need any more information from me
in this subject area please don't hesitate to ask. I would be happy to
work with you on this as much as you need.


Open Source Projects out there that do this type of thing and could
help to save much of the develpment costs if you chose to got that
route.

Freshmeat -- A website full of OpenSource projects
http://freshmeat.net

GeekLog -- Very configurable and adaptable to almost any communication
site
http://geeklog.sourceforge.net/


Again, if I didn't cover something as much as you would have liked
then don't hesitate to ask. I'll keep an eye on this as much as I can
over the next few days.

Thanks

webadept-ga

---

Request for Answer Clarification by joel1357-ga on 25 Oct 2002 23:25 PDT

webadept-ga,

I have been extremely pleased with many of your answers in the past,
however I have to say that in this circumstance I am COMPLETELY LOST.
I'm sure there are plenty of technical people like richard-ga that
would say WOW great answer, but I have read this 4 times and I feel
like I have gone 50 steps backwards in my understanding of all of
this. When I read shiva777-ga's answer I showed it to several people
and they felt like this guy was really on top of it. Keep in mind that
my friends and I are businessmen and not technically literate. We are
nearing the time when we will dump some money into this but the
problem is I am having a VERY difficult time quantifying how much
money we are talking about. I very desperately need to know what the
up front and ongoing costs will be..i'm talking from day 1 through the
end of development, and into the 1st 1,000 users all the way up to
around 20,000 users a day (possibly at one time, with the ability to
store the information for a couple of million profiles and each
persons pic...in this question I am trying to find out what those
costs will be for answers 3 and 4 of personals question # 3..I've
spoken to these two friends of mine and I can't tell them if were
talking $5.oo or $ 5,000 on day one......$ 3,000, $ 30,000 or $
300,000 for the development..I can't predict how much money it will
take to host the site..or when that part of the costs kick in..is it
from the beginning or only after the development is done..will the
costs initially be $ 1.00 a month or $ 10,000 a month...how much will
these costs grow as more and more people come to the site..eg when I
mentioned the additonal server per 1,000...heck I don't know..Bottom
line I am ready to move forward with this yet I can't quantify the
results...shiva777-ga did a great job helping me understand when he
said...."so lets sum up these estimates"..and then he proceeded to put
it all down so that I could get an idea and be closer to telling my
other friends how much all of this will costs us. Could you please
help me.

Thank You,
Joel

---

Clarification of Answer by webadept-ga on 25 Oct 2002 23:58 PDT

Yes I can, but I can't get to this tonight or even most of the day
Saturday, as I have other engagments and I'm just about to leave. But
I will get this too you by the end of the day Saturday. I'll break it
out for you for both renting and purchasing and give you some good
figures.

Sorry I got too technical for you. I really tried to bring it down to
Executive level as much as I could. I guess it just got away from me.

webadept-ga

---

Clarification of Answer by webadept-ga on 27 Oct 2002 01:41 PDT

These are breakdowns for the areas you have asked me to cover in this
question. I hope that these clarify some of the questions I didn't
address.


Scenario One

You purchase everything and run the website at an data center which is
giving you bandwidth and a server room, but the server is yours and
all the maintenance on it is your responsibility.

Break Down of startup Costs

	Hardware

	Two servers to start with 	1500.00 x 2 :  $3000.00

Main Server and backup server. Either the Dell's your man suggested or
the IBM servers.  System relies to start on one server being both
database server and website server. To start the site and get it up
and running to the first 1000 users I would not suggest getting into
load balancing and other scenarios. When the need changes then add the
second server putting the database on its own and accepting the costs
then. This is not required, only suggested strongly. Routers and other
network connection is taken care of by the ISP datacenter. You just
supply the box and have a network card.

	Software

	Linux OS (Redhat, Mandrake) 	$120.00 (full package with all the
Server frills)
	Apache Webserver   : Included, no cost
	MySQL 		: Included no cost
	PHP		        : No cost
	Perl		        : No cost
	SSL(Secure Server for money transactions) : No cost. 

	Hosting ISP Setup

	Data center  with Bandwidth needs :  est $300 setup fees and access
                                                           est $400
per moth starting bandwidth

Personnel costs

	Setup Server 	Total : 15 hours  x 55 per hour :  $825.00
with database, apache, SSL, IP address, network, DNS, sendmail, Perl,
PHP, and security. Add necessary scripts to maintain, and get
everything ready for the programmers  and database admin to move in.

Total startup costs with this senerio is : $4645.00


Break Down of Costs to 1000 users (monthly costs)
	Hardware
No change

	Software
No Chage


	Hosting ISP Setup

	Monthly charge estimate :  $400 per month

Personnel charges:

	Webmaster for system. On call for break downs, backup and recovery,
daily maintenance and alterations to the system as needed by the
programers as they do their work.

	Full time position  est $50.00 per hour x 40 hours per week. 

	On call programmer working as needed to add features or trouble shoot
areas where things are not working quite right.
	est between  40 and 200 hours a month possible. 



Break Down of Costs from 1000 to 10,000 users. (monthly costs)

	Hardware
	 Add new  two servers:   $1500.00 x 2 :  3,000

	Software
System needs to be setup now for load balancing for the Apache server,
to handle requests. Database is moved to it's own server. Backup
server is changed to replace one of the main two.

	Hosting ISP Setup

	Amount of traffic now has increased bandwidth charges. You may be
paying as much as $1000.00 per month for bandwidth along with $200.00
per month basic charge.  est   $1200.00 per month.

Personnel Changes
	Increased activity now requires a Database manager, and a full time
programmer to make alterations and add/fix features. Should also be
able to be a backup for your main SysAdim.

	est $40 - $50 per hour. full time position. x 2


Scenario Two

This scenario is based on using a rented or leased dedicated server
rather than using in-house talent for server maintenance and backup
requirements.

Break Down of startup Costs

	Hardware
          No Cost

	Software
          No Cost

	Hosting ISP Setup

	One-Time Charges $652.57 
	Monthly Charges $1,125.85 


Personnel costs

Database manager to setup database and get it ready for programmers. 

	time charge 5 hours  @ $55.00  :  $275.00


Total startup costs with this scenario is : $2053.42


Break Down of Costs to 1000 users (monthly costs)
	Hardware
No change

	Software
No Change


	Hosting ISP Setup

No Change

Personnel charges:

	On call programmer working as needed to add features or trouble shoot
areas where things are not working quite right.
	est between  40 and 200 hours a month possible. 

Break Down of Costs from 1000 to 10,000 users. (monthly costs)

	Hardware
	 Add new server 

	Software
No change and not extra costs. 

	Hosting ISP Setup

	One-Time Charges $652.57 
	Monthly Charges $1,125.85 

Personnel Changes
	Increased activity now requires a Database manager, and a full time
programmer to make alterations and add/fix features. Should also be
able to be a backup for your main SysAdim.

	est $40 - $50 per hour. full time position. x 2


This break down is for the basic infrastructure of the website and
doesn't include development costs, or design and graphic artist costs.
It is just the areas that get the website up and keep it running on a
monthly basis.

I hope this answers your question. If you need more clarification
please don't hesitate to ask.

webadept-ga

---

Request for Answer Clarification by joel1357-ga on 27 Oct 2002 01:21 PST

webadept-ga, 

I need the development costs, or design and graphic artist costs.

Thanks,
Joel

---

Clarification of Answer by webadept-ga on 27 Oct 2002 01:42 PST

Okay, Do you want me to break those out in your #4 area like I did
here, or just add them to this question? It might be better in the #4
question so that you have the background explanation and the breakout
on the same page. Just to keep things together.

I'll start working on it Sunday afternoon and have it posted sometime
in the evening. If you don't want it posted to #4 send a
Clarification, otherwise look for it there.

Thanks

webadept-ga

---

Request for Answer Clarification by joel1357-ga on 27 Oct 2002 16:09 PST

Putting it in answer # 4 will be fine.

Thank You,
Joel

---

Clarification of Answer by webadept-ga on 05 Nov 2002 01:11 PST

Posted in #4

Thanks, 

webadept-ga

webadept-ga,

I have talked over your answer with people that work for me. They felt
like your answer was superb, and were able to break your answer down
to a level I could understand. I appreciate your thoughtfulness and
thoroughness. My only request is that in the future when answering my
questions that you would over-compensate for the fact that I am a
technical idiot.

Again Thank You,
Joel

Webadept's answer is top notch--where else but on Google Answers could
you get a consultant of his caliber to provide such a comprehensive
answer at this price.  Bravo!
Google Researcher richard-ga

I'd like to offer you some perspective, for what its worth.

Just because you can doesn't mean you should.  You're talking about a
huge infrastructure cost, one that you're unlikely to recoup if this
doesn't work (unless you don't mind throwing money away).  Putting it
mildly, its a big IF. Of course anything 'service' (programming)
related are sunk costs.

Before buying all this equipment, you may want to re-evaluate if a
month-to-month leasing arrangement for hardware wouldn't be better
suited to your operations.  Nobody goes into business thinking they'll
fail, but an exit plan should be in your busines plan.

Food for thought.

I think you got a great answer.  However I am not sure you are asking
the right questions.  If I were going to do this,  I would spend only
about $1k total atleast on the initial build and that is a maximum.
Go here
http://www.hotscripts.com/search/?query=dating&category=PHP&bool=AND
and buy one of the predone scripts.  To me E-MatchMaker looked good
and was $150 but you can find one you like.
If you need any changes to a script to customize it.  Go to elance.
http://www.elance.com you can add any changes you need.

Since you aren't technical pay someone to install the script probably
$25-$50.  You can probably get the place you buy it from to install
it.

Next pay someone to make the front end of the site.  Again go to
elance.  You can probably get a front end designed for about $500.
Then find a hosting service about $20 per month.  Go with a fairly
large ISP so you don't have connectivity problems.  Make sure they
have PHP and Mysql included. Also buy a domain name $10 to $20 based
on where you register it.

So here are your costs.
Script $150
Install Script $50
Web Design $500
Domain Name $20
Hosting $20 per month

Now here are the questions you should be asking. 

1. How are you going to be different from the 100 other major dating
sites on the net? And why are people going to want to use you over the
other dating sites?

2. The other important thing is how are you going to get traffic? 
Just because you build it doesn't mean they will come.  So what is
your advertising budget?  Unless you have a few million dollars don't
expect a lot of traffic.  How are you going to get traffic?  Don't
think just because there are millions on internet daily that they will
visit you.

Finally you are worrying about things that are basically non-issues. 
If you have so many users that your site can't support it then spend
the money on some custom solution.  Believe me you will have plenty of
time to work on a new program after you put the initial site out
there.  This isn't 1995 you aren't going to get a million people to
your site without a serious advertising budget and a lot of work.  90%
of your effort will be devoted to getting people to your site.

I don't mean to discourage you but don't waste your money on some
custom solution.  Spend your money after you know it is going to be
successful not before.  I haven't read your other 4 questions but if
you paid $200 per question you could have built my solution:-)

OK. Just a few thoughts.

First, I am not an expert. Repeat, I am not an expert. What I am is a
PC technician who has done a little bit of reading into servers, and
into the business side of things.

Not in any particular order here, but:

1)Linux is probably your cheapest O/S choice. I've done a few quotes
on servers recently, and I gotta say, the cost of Windows is often as
high as the cost of the server. Not that I am saying I think anything
about Linux, because I've never used it. All I'm saying is Windows
2000 server, Advanced Server, and DataCentre Server are expensive.

2)As a businessman, you and your partners have probably thought about
this already, but anyway: LEASE!!! If you buy now, generally the
equipment will be a capital cost expense that has to be gradually
depreciated. 3 years from now you'll still have a lot of value in the
books for the equipment. If you rent, you'll pay a monthly fee which
can be written off as an operating expense. However, usually the rent
isn't accrued or applied in anyway towards if you wanted to buy the
equipment. If you are on a rent-to-own kind of deal, it can end up
being very expensive. If you get yourself a good lease rate, you can
make monthly payments that can be written off, AND at the end of the
lease, you can either buy the equipment (usually at fair market value,
which isn't high on 3-year old equipment, even for servers.) Or you
can re-lease the same equipment, or you can get new equipment. The
great thing is, you can build "soft" costs (ie. non-hardware) into the
lease. Initially, that could include the setup of the hardware itself,
the service contracts, the backup systems, the power supply hardware
(you have to have redundant uninterruptible power supplies), data
backup systems, etc. Moreover, you can lease the backup server at the
same time, without signifigantly increasing your startup costs. If you
want to upgrade half-way through, instead of waiting until the end of
the lease, you can have what's left on the lease added to the new
equipment and the migration costs in the new lease. No major expense
involved.

3)I'd buy from a VAR/reseller (authorised, of-course) rather then
vendor direct (except in the case of Dell, where you cannot buy from a
reseller. Reasons: Often the resellers will sell below the list price
that the vendor charges. Also, if the equipment doesn't perform as
well as it should, or has a high failure rate, then you have someone
else on your side when dealing with the vendor. You may only have
bought a couple of servers, but the reseller that's getting pissed at
the vendor might be selling 10 thousand of them every year. Vendors
don't want to lose that kind of account. I've seen them bend over
backwards trying to help me out with a quote and with information,
especially if they know I'm preparing quotes from more then one
vendor. And sometimes it works. Just recently, a customer went with a
quote on an IBM server for a total of about US$25,000. Not because the
Compaq quote was more expensive and less powerful, which it was, but
because I couldn't complete the quote due to the account rep not
calling back. They took too long, I presented the IBM quote, and the
customer decided not to wait to hear what Compaq had to day.

4)Many servers include 3-year onsite warranty. You will, of course, be
getting the 3-year, 24 hour, 7 days a week, 4-hour onsite contract. If
your website is going to be as big as it sounds like, you cannot
afford downtime.

5)Speaking of that, your going to want redundancy. Example: Server
itself has redundant power supply. Hard drive in server is mirrored
via RAID. Main storage is external. Completely redundant (and probably
striped as well) RAID array. hot swappable drives. Reduntant power
supply. Server and storage are linked to a second server which is
backup to the main server. Again, all have redundant power supplies,
connected to independant UPS systems. Tape backup in each server or in
the storage unit for nightly or weekly backups. I'd do a local backup
nightly, and maybe every week, or every 3 nights or something,
depending on your needs, have a backup set taken off-site to a remote
location, to be stored in a fireproof safe. Even if a bomb blast kills
your entire setup, you can still get new equipment and be up and
running very quickly. All the netowrking equipment should be on the
lease as well. Probably want gigabit (to allow for future growth, the
day may come when your site hosts video as well as pictures) ethernet
equipment. All network equipment is redundantly connected to the
servers using redundant links. Redundant power supply for network
stuff. Connected to UPS of course. Try to get either secondary backup
connection (T1 or whatever) from ISP or from different ISP if
possible. When I say "network equipment" that would include switches,
firewalls with DMZ, etc. My personal choice is 3Com. It's your call.

Advantages of all this redundancy: If anything fails (and it probably
will) you DON'T lose money. For example:

Something in the server fails (ie. CPU, RAM, system board, etc.) the
backup server comes online immediately and takes over. Because the
storage is external, you don't have downtime beyond the switching time
for the second server to kick in. Within 4-hours, the technician shows
up, repairs or replaces parts in the server as needed. You either
switch back to it, or make it the backup and leave the 2nd one running
primary. Since they're identical, it's no problem.

Power supply in the server or storage unit or network equipment fails.
No problem. Even one of the UPS systems fails. No problem and zero
downtime. In fact, they're often hot-swappable, which means you (if
you have extras) or the tech that shows up within 4 hours (remember
your service contract) simply pulls the dead one and puts the new one
in.

Network switch or connection fails. No problem. Second switch or stack
takes over through redundant links. Virtually zero downtime. Netowrk
hardware company often has sameday or next business day advance
replacement of failed equipment.

Hard drive in server fails. No problem. Either the server runs with
it's one drive, or you flip to the backup server. Within 4 hours the
technician arrives and hot-swaps the drive. Don't even need to reboot.
Since it was mirrored, all data is copied back to the new drive. Zero
downtime.

Hard drive in data storage unit fails. No problem. Since you are
running a mirrored and striped system (can't remember the RAID level
number) you have no downtime and no lost data (super crucial for a
data-driven business). System keeps running, and within 4 hours, a
technician shows up and hot-swaps the dead drive. (BTW hot swap merely
means that without having to shut anything off, the part can be
removed and replaced.)

Link from ISP goes down. No problem, network equipment switches to
backup. Little downtime.

Another advantage of the redundant servers is that one can perform the
full nightly backup while the other runs your site.

You might ask why I seem so paranoid about downtime. That's where your
business experience comes in. Take a look at annual revenue (not
profit, just incoming revenue) for a major competitor that you want to
be challenging. Divide that by 8,760. That's how much revenue could be
lost if they went down for one hour. Bear in mind that the actual
amount could be much higher because not every day of the week is the
same, not every hour of the day is the same. There will be slow times.
Which means the loss during a peak time could be quite higher. And it
doesn't factor in future losses due to customers driven away during
the downtime. Then compare that to the cost of the extra equipment and
servicing. It's really your call but you did say "grade A tier 1...
without cutting corners"

I would also get some sort of tertiary backup system in a remote site.
maybe a server, maybe just a PC. something that can host a "sorry
we're having technical difficulties" type of thing that you can switch
to in case you ever have major problems. The key thing will be having
it set up so that your ISP can make your website's IP address point to
that server when necessary. That way when people key in
www.yoursite.com they would get the message instead of an error. When
you really grow, you may want to make the remote site a live
backup/mirror of your original site. Basically so that if the building
blew up, the city got nuked, or maybe just some idiot with a backhoe
cut off a major trunk line into the city, you'd still have a website.

Hardware wise, I won't get too specific. I know webadept gave you some
rough pricing, but realistically, that may have been under quoting.
I've built high-end game PCs that cost more then $1500. Especially
considering you're spending at least $1,000 just on 5 google questions
(assuming they're all the same price as this one.) I don't usually
deal in US dollars, but at a rough estimate, I'd think you'll be
leasing about 15 to 25 thousand US dollars of euipment, including the
service packs, network gear, and power equipment. But I could totally
off. Just a guess.

For the specs, I'd be thinking along the lines of:

Xeon processor, the faster the better. Probably dual CPU, preferably
with room to expand to even more. If it's available when you build,
look at the performance and price of the IA-64 stuff. But probably
Xeon.

1 gigabyte of RAM, minimum.

only around 9 to 18 gigabytes of hard drive for the servers themselves
(remember you'll actually be buying twice as much, 2 drives, for each
server.) Probably 10K rpm ultra 160.
Tape backup in the range of 60 to 200 GB (to allow future growth)

The storage unit will need probably 4 or 6 drives (again, remember
we're striping and mirroring it.) 16, 36, 74, or 146 GB per drive. All
identical. 15K rpm ultra 160 or ultra320 hot swappable.

Superstack 3 switches. maybe 4300 or 4400. You'd have to talk to a
network specialist.

APC for the UPS stuff. no questions there. Gotta be APC.

IBM or Dell are probably your biggest choices for the server. Dell
makes great stuff, but as I say above, I'd go through a vendor, which
would mean IBM. Avoid clone or "house" brands on anything. You need a
vendor that you can 100% guarantee will be around a few years from
now. You could look at Compaq, but in my experience they are more
expensive for the same or less performance level. I don't think, in
your volume range you could even get HP to talk to you about this
kinda stuff. I know they wouldn't talk to us when we had a client that
needed 2 or 3 servers and about 100 workstations.

Another reason for going with a VAR, especially one that handles more
then one line: They're making money on either brand. So if you have a
reseller you trust, and is authorised, then you're likely to have an
account rep that knows what they're doing, and will make the best
recomendationd for you.

I hope this has been of some help for you.

additional perspective...
i've spent 3 years as a systems engineer supporting fortune 500 web
companies.

a few comments.
almost no large web site uses open source.
slashdot is not a great example as its not a very busy site in terms
of internet traffic and well they have all the in house talent to
correct any issue.

from a business perspective saving $800 on licensing doesnt make sense
when something goes wrong. should a problem arise within an open
source package or kernel who do you call to demand someone to fix it
while you loose money? I dont think Linus takes emergency pages at
3am.

Sun microsystems and Microsoft are pretty much the only players in the
true web hosting game as far as operating systems go. linux is only
being rolled out in limited numbers for roles that cost too much for a
licensed server (SLAs are usually greatly reduced if its used in a
full production enviornment). (replication servers etc). Both Sun and
MS web servers can handle huge loads. There is honestly no simple or
accurate way to predict the amount of users this or that server can
support. Ive worked on sites that were speced to handle 10k concurrent
users that bombed out at 20 and vice versa.

The only reason apache and linux numbers look so high in web surveys
is those surveys include all the web survers out there. 90% of those
are single servers guys are running out of their houses. Check and see
what the largest companies with the most impressive web sites are
running.

For this reason application development is the first step followed by
extreme load testing in as close to a production enviorment as you can
afford to replicate.

As far as hosting no small time ISP can handle the traffic stated.
Even if they can make it work from the business side there it makes no
sense take the risk.
When users dont get through or see an error chances are they just
arent coming back, sad fact but true. Businesses that are intent on
becoming serious players on the internet are now requiring from
hosting companies uptime SLAs of 99.999. thats just minutes a month
before the hosting company starts refunding money.
Why bother with an ISP that is hops away from actually controlling
their networks. Select the web hosting companies that ARE the
internet. Worldcom owns Digex and Cable and wireless own Exodus. Both
own a large majority of the internet.

As mentioned before purchasing equipment is just a waste of money.
Hosting companies dont even purchase their equipment, its just bad all
around.

Co-location is and was a bad idea that is dieing off. Managed web
hosting makes much more business sense and requires less company staff
to use.

The two largest managed web hosting companies at this point are
www.Digex.com (hosts ford etc) and www.exodus.com (hosts yahoo etc).

I'm sure either would be happy to talk with you. Sales and Sales
engineers from either company are great "free" resources when planning
a large business venture on the internet.

Both run massive web sites for huge name clients but they also have
plenty of 1 and 2 server clients in their stables. The only thing most
managed web hosting companies wont provide is programming talent.

getting your site working is only a small part of the game keeping it
running is where you want the pros on your side.

as far as a costs estimate of running a 2 server site that can handle
20k concurrent users? 6-10k a month w/o programming staff costs (ever
wonder why there are so damn many ads on the internet?)