Dedicated Server

[THL]

Just got a dedicated server from 1and1 how should I have my server settings ?

WebDav on off ACTIVE
FastCGI on off INACTIVE
SSL support on off ACTIVE
Perl as Apache module on off INACTIVE
PHP module on off ACTIVE
Memory usage kB 131072 kB
Process timeout seconds 60 seconds
Maximum simultaneous processes 100



Should perl as apache module be active? Is the memory usage ok I have 1GB setup, what about process timeout too...

Thanks for your help! :)

[bfarber]

Unless you use any perl scripts, keep it off (for security reasons).

1GB memory isn't that much - how many active users do you have? If it's more than 100 simultaneous users or so you'd want to upgrade that. Memory is the key really - processors vary and most server processors work out fine, but if you don't have enough memory you will start swapping and server will seem extremely slow, if it doesn't crash.

[rct2·com]

The optimum amount of memory also depends on the operating system. ;)

Asking whether an amount of memory is adequate is what I call a "how long is a piece of string" question.The answer is always "it depends". ;)

Details of:

• Typical # of Posts per day
• Peak users online (and duration of peak)
• Operating system (and version)
• Number of sites on server
• Any 'bloatware' running (PLESK, WHM, CPanel, etc)
• Processor speed
• Disk interface, # of disks, speed of disk
• LAN interface speed
• etc.
  

would all help you get a better answer to your question.

For the record, we run 14 sites, mostly static HTML sites, but 2 with IP.Board running, 1 of those with 340 users per hour (for 4 hours), on RedHat 3, PHP 4.x, MySQL 4.x, on a 2.4GHz Intel P4 with a single 80Gb IDE hard disk, and 512Mb memory. There is NO paging of memory to disk because of the way we have configured Apache, and server load (as reported by Linux TOP command) rarely exceeds 6.

[bfarber]

I'd be interested in hearing how you have Apache configured - also, do you have cpanel or anything running on that box?


While you are correct, there are dozens of factors that can affect performance, I was basing a generalization. I've noticed 80-100 users you might get away with 1GB RAM (I'm talking active simultaneous users, as opposed to active in a time period), while 100+ users the server might start to choke.

[THL]

Well I'm in the process of scrapping the 1and1 account rubbish support and misleading information about their servers...

I've now went with a managed liquidweb server and the spec is as follows, hopefully should be fine for the one forum I run with around 200 members and 150 guest at peak times...

Athlon 64 X2 4600
2000gb ddr ram
100mbs port
x2 160gb sata HD

[Andrew.S.]

Blah THL, I could have hooked you up with a better dedicated server for a cheaper price B)

[rct2·com]

I'd be interested in hearing how you have Apache configured - also, do you have cpanel or anything running on that box?

Yes forgot to mention, we have PLESK 7.5.1 running on the server.

There is nothing like XWindows or GNOME or other GUI, all our access is command line putty. There's only a limited number of development tools like gcc and make, needed for compiling RPMs and the like.

In January 2006, when we took the server build from our hosts (am I allowed to mention it was ev1, who are now ThePlanet?) we really only did 5 things: {MaxClients}*{Memory allocated to a running httpd process} <= {physical memory in machine} As I understand it, [I am a self-taught Linux/Apache technician], by default, each new visitor to your site causes a new instance of Apache to start running to process all that visitor's 'clicks'. Once there are {MaxClients} versions of Apache running, visitors have to share the versions of Apache that are running. This means that the visitors' clicks get 'queued' and if the number of visitors is excessive, they will get server timeout and too long to respond messages. However, if {MaxClients} is too high, all the Apache instances need more memory than there actually is, and the server starts 'paging' memory out to disk. If the paging area is on the same disk as your MySQL database, your server load goes through the roof, and almost inevitably users start to get '500 internal server error' as Apache completely crashes. There are lots of way to manage this, and it really depends on your server layout. IP.Board supports locating your database on an entirely separate server for example, there are 'light' versions of Apache, you can have more than one disk so that paging can be on one disk, and the database on the other, more memory allows you to raise the MaxClients limit, etc. But the thing that really seems to kill your server is paging of memory to disk. The formula above can mean that visitors see timeouts, but at least your server doesn't crash.

1. [*]Updated Apache from 1.3 to 2.x [*]Updated MySQL from 3.x to 4.x [*]Updated PHP from 4.x to 4.a different x! [*]Changed PHP's memory allocation to 16MB from 8MB [*]Changed the Apache MaxClients setting
   
   
   
   
   
   
   
   
   
   
   

It was the last one of these that gave us the real performance improvement. I have posted about this before in Ineo topics somewhere. You really need to make sure that [on Linux]: