Question about IIS performance on lower-spec dedicated hardware?

[dotitab398]

Hey everyone, I’m hoping to get some advice from those of you who have been running IIS for a while. I’ve mostly been a Linux/Apache guy, but I’m currently setting up a small, dedicated Windows box for a client who insists on an ASP.NET environment.

I’m working with a relatively low-profile setup—specifically a micro-server that’s running on a Server Power 200-Watt PSU. Because the power overhead is so tight, I’m trying to keep the hardware footprint as lean as possible. I’ve noticed that when I start layering on a few Application Pools, the worker processes (w3wp.exe) seem to spike in memory usage much faster than I anticipated.

One specific point I'm curious about is the "Idle Time-out" setting in the App Pool advanced settings. I’ve read conflicting things about whether aggressive recycling actually helps with resource conservation or if the constant "cold starts" end up putting more strain on the CPU in the long run. Since I’m trying to keep this build efficient and cool, I don’t want to overwork the processor unnecessarily.

In my personal experience, I’ve found that Windows Server can be a bit of a resource hog if you don't strip away the unnecessary roles and features right at the start. I spent all of yesterday just disabling services I knew we’d never touch, which felt like a win, but the IIS overhead still feels a bit heavy for this specific hardware.

Has anyone else here managed to successfully optimize IIS for low-wattage or constrained hardware environments without hitting major latency issues? I’d love to hear how you handle App Pool management when you don't have infinite RAM to play with.

[gizmola]

Sounds like a penny wise pound foolish exercise. 

Of course things will be slow if processes have to be started frequently.  Of course there will be additional power usage.  The question is: what is the expected average and peak usage?

You state this is for a client.  If the server PS is underpowered, then it's likely to fail.  What will the cost be, when the server PS burns out, perhaps damaging the entire machine?

Servers typically feature redundancy so that they don't die when a single component has a fault (ie. having multiple PS's, RAID controllers, RAID configuration and hot swappable drives.). Why would you use a box that isn't designed to be a server, as a server, and then attempt to micro optimize and skimp on the PS?

If the usage of this environment is such that there is low to no usage during specific periods, then perhaps you can economize by configuring the box to sleep and wake on lan.  You didn't provide any information on what services it will be providing.

I don't have much else to offer other than that IIS is infamous for its instability, particular if you require additional IIS Modules.  If you've gone through the exercise of trying to reduce overhead, and remove/turn off non essential services, then you've already optimized the server to reduce OS and cpu utilization, and if the box isn't doing anything then power consumption will already be low.