Hi All

I'm a newbie setting up a site and I'm having some trouble determining my hardware/service provider requirements.

WHAT I'M TRYING TO ACHIEVE:
My father and brother are photographers who run a small business publishing books and postcards featuring their work. They are very keen to start an online image library and as the family member with the IT skills I've joined forces with them for this project. However, my forte is (java-based) software development and I know very little about the world of servers.

We have around 2000 images at present and our server will need to provide users with some fairly standard functionality:

* Image search: results returned as thumbnails (~5K) linking to high-quality jpegs (~ 100K)
* Image download: (after purchase): each image is a TIF file (~50Mb)
* Mailing-lists: monthly newsletters, etc.

It is important that our site is responsive and that users can download their purchases quickly. I intend to use a
Linux-based server.

WHAT I NEED TO KNOW:

* Server requirements: I have no idea how to determine what level of machine is needed to ensure rapid response/good download times. What process do others follows to make an estimate? Any recommendations? Single Pentium III 1GHz processor? , Dual Pentium III 933MHz processor?

* Connectivity requirements: dedicated-server and co-location packages seems to discuss only throughput (i.e. how many gigs
are served each month. My concern is the speed at which my site will be served/images downloaded. What is the term for this
quantity (connection speed? I thought this is what the term bandwith applied to)? How does one determine this requirement and how do you determine what a service provider will deliver? Any recommendations?

* The storage requirements for this system are large. We will need ~100Gb to go live. I have not been able to identify any
suitable storage medium other than the hard-drive. It seems that servers usually use SCSI drives which appear to come in
36Gb units, whereas non-SCSI drives are available in 100Gb units. Am I correct in thinking that 4x36Gb SCSI drives will
provide far superior performance to a 100Gb non-SCSI drive, given that the SCSI drives reduce the work of the server's CPU?

I know that this is a mouthful, but answers to any of these questions will be hugely appreciated.

Thank you for your time.

looks pretty much like you need something like the following:

1ghz processor
some chunky SCSI hard disks (RAID Array)
fast internet connection - 100mbit id suggest

use something like apache for the web server probably with some sort of compression mod_gzip

id say getting 200-300GB transfer a month would suffice to start off with to get going.

most hosts will offer this but it will cost you a lot of money.

id bargain on at least $800 a month for a decent host with good quality reliable bandwidth.

post a request on the web hosting requests forum and im sure a few hosts ill be willing to provide you with an adequate service/system.

Regards

-XT

mod_gzip won't do any good, as it can't compress images which are already compressed, unless your site will also feature a lot of writing then I recommend installing it.

I don't mean to get off topic, but what would someone do with the 50MB tif file after downloading it?

Are you selling to book and postcard publishers then?

If I were an average photo fan, I'd want to buy it already printed.

Anyway, I'd think that IDE would be fine if you can't afford SCSI.

If cost is no option, get a quad xeon processor machine with a huge array of SCSI disks and a DDS-4 tape backup.

I'm sure that you'd be fine with a single processor (Athlon or Pentium III) 750-1GHZ machine with 2-4 100GB+ 7200RPM 100Mb IDE drives.

They'll only be downloading the 50GB file when they purchase, right? Of course you didn't say how many visitors and downloads (resource requirements) you expect, but I wouldn't expect too many.

Thanks for the replies :)

ASIDE:
We hope that fans will buy our books and postcards. Those purchasing our images digitally will typically be using them for reproduction in their own products (e.g. advertisments, magazines, etc.).

The industry standard for digital images suitable for high quality reproduction is ~ 200 dpi and this makes for some pretty big file sizes. So downloaded images will typically be turned into printed material by a professional printing house.

QUESTION:
So far the recommendations are for a single 750-1GHz processor. What I don't follow is how one determines this: what effect, for example, do 10 users downloading 50Mb files have on the responsiveness of the server to general web browsing requests. I appreciate that this depends on many variables not least of all the software being used. But are there any rules of thumb or formulas? Anyone know the way to articles on this subject?

If you're just serving files then a single processor (not even fast) will do the trick. Serving files doesn't tax the CPU, but if you have all sorts of scripts running, chugging along constantly with many users on the server at once, then the server needs to do something. You could have 200 people just downloading a 50 MB file and I don't think it would show up on the charts.

I agree with Chicken. A Pentium III 700 MHz should suit your needs just fine. Webserving for something like static pictures is farily easy work for the processor -- just read the data and send it. I would however recommend 256 Megs of RAM. Anything less than that doesn't really work well for webserving.

You referred to how FAST the connection is and mentioned bandwidth. Bandwidth means the total amount of data that is sent from your server. This has nothing to do with the speed of that data to individual users. The best way to estimate who has the fastest network is through simple ping'ing. Find the domain name of someone who has a dedicated server from the companies you're looking at (or ask the companies themselves for an example) and ping that site to get the response time. You do that by going to a DOS prompt window and typing "ping thedomain.com". This gives the travel time from your computer to the server and back -- Lower times are better.

As for the actual image software, I wrote almost exactly what you described about 2 years ago -- I may be able to help you out with that -- PM me.

<edit: I didn't mention anything about drives. I think 2 60gig IDE drives should work more than great for your particular setup.>