I know there are a few people here interested in networking... so my question to you is, what would be your ideal distributed computing setup for a OpenBSD/Linux Apache PHP MySQL setup?


Let's assume a budget of $10,000 to $50,000.


Based on my current level of knowledge and understanding.... here's roughly what I would want:

1. Firewall Running OpenBSD (1)
- Firewall equally distributes Port 80 requests among 3 web servers
- Firewall equally distributes Port 81 requests among 3 image servers

2. Web Servers Running OpenBSD, PHP, and Apache (3)
- Web servers locally connect via SOCKS to remote MySQL cluster head.
- We use IPSec to create a VPN to the MySQL cluster, and NTFS to mount the remote SOCK locally

3. MySQL Database Servers Running OpenBSD and MySQL Server (3)
- One server is elected as head, receives all requests
- Requests are then equally peered out and distributed to other MySQL servers

4. Image Servers Running OpenBSD and SCIS (3)
- Image Servers dynamically cache frequent images minimizing Disk IO

私は日本語を少し読むと書くをできます。あなたも？

I'm still only learning about linux and networks so my knowledge is limited, that does seem like a sweet setup though. The question is "what would it be used for", so you would want to evaluate the use there. I can see why you might want something like that for expansion but if it's for a small site that isn't very well known I don't see many advantages to seven big powerful servers running one small site. Also, the upkeep for those systems would be kind of high, what with hard drives, other hardware, software licenses etc.

Granted, it'd be over kill for ACDB right now. Haha. Currently we're running off one 90's era server.

What would it be used for.... hmm, that is difficult to summarize. Need to look at daily values for pages requested, data transfered, database queries and internet connections.

Our current ACDB server, I theorize will max out at:
10,000 daily users
300,000 daily page requests
1TB daily data transfers

The setup I came up with I'd expect would accommodate perhaps 10x the load. The current bottle neck is SQL queries....

私は日本語を少し読むと書くをできます。あなたも？

I did parallel processing in Unix, but nothing internet related, so I don't even know where to start.

but equally distributing anything would only be a good idea if each of them is equal in connection speed, otherwise it should go to the ones that are able. I dont think there would be much of a problem with calculations since its just read, write, and search for the most part.

'The unseen cast no shadow' ~THE ROAD, Cormac McCarthy

"Only a fool trusts his life to a weapon." Frank Jaeger, Metal Gear Solid

Regret I did parallel processing in Unix, but nothing internet related, so I don't even know where to start. but equally distributing anything would only be a good idea if each of them is equal in connection speed, otherwise it should go to the ones that are able. I dont think there would be much of a problem with calculations since its just read, write, and search for the most part.

I should explain a little more that got me thinking this way. I was reading through the OpenBSD pf.conf manual and saw in the examples section a way to load balance across multiple servers.

Let's say we have 3 HTTP servers:

172.16.2.2 ( HTTP Server A )
172.16.2.3 ( HTTP Server B )
172.16.2.4 ( HTTP Server C )

The Firewall, 172.16.2.1, gets a request for port 80. Let's say all three of our servers are clones, so we can fire the connection out to any one of them. The firewall checks its connection tables, and forwards the port 80 request to the server with the fewest established connections.

I'm still curious how to operate three cloned web servers, especially if incoming connections will use different servers over the course of their sessions. An example problem is, PHP uses temporary files to store session values. If these files are created on server A, then how do we get exact copies on server B and C, in the event the user's established connection closes, and is reestablished on server C? I'm thinking, mounting certain directories like those as NTFS may be best.

私は日本語を少し読むと書くをできます。あなたも？

Now, for something like the problem there wouldn't it be good to have the servers connected to a single data storage device? I know that there are setups out there that allow for one computer to have the data and other couputers (like web servers) access that information. That would fix that problem. ...now if I could remember the name for that stuff...

That may well be the ideal solution for that issue. I was thinking multiple servers with the same data for the sake of backups, but perhaps a single data server using an ideal RAID setup, could do the job just as well if not better.

私は日本語を少し読むと書くをできます。あなたも？