What's the point of a server OS and server CPU - Page 2

For what it's worth, I have the 50/15 Comcast tier, but it's the residential version so I'm not supposed to run a web server. Not a problem because I am not a web guy.

I use hyper-v pretty heavily, and the new 2012 hyper-v (3.0?) that was released with Windows Server 2012 takes full advantage of my hardware. You run out of expansion room very quickly on single CPU systems.

http://www.overclock.net/t/1303032/uses-for-your-home-server

This link was posted earlier. This is a recent thread with discussion on uses of home server, and if you read through this you will find what I use my servers for...as well as Oedipus...as well as Manyak...as well as parityboy...as well as several others. This thread, if you take the few minutes needed to read it, will tell you all the different things you could do with a server.

Also, you don't need a "top of the line connection" to host a website. I host 3 websites, all WordPress (Apache)...and a wiki. I also have web interfaces for my SABnzbd, CouchPotato, SickBeard, Headphones, Plex, and SubSonic applications. I also have an RD Gateway set up, with RD Web Access for a personal VDI demo I am working on, as well as an Exchange server with Outlook Web Access. My firewall (Untangle) also has a web configuration page. That there is 13 different "web sites", if you will...and my connection is 50mbps down and 5mbps up. I do have a good connection...but all that stuff barely uses it. A personal self hosted WordPress site isn't going to use nearly any of your bandwidth (unless you have dialup or bandwidth limits or something).

And Oedipus, it is Hyper-V 3.0 in Server 2012. Isn't it great??
Edited by tycoonbob - 10/12/12 at 4:28pm

Yeah, I like it a lot. It's easy to use and, with 3.0, you can devote more than 4 virtual CPUs per VM. I haven't looked yet, but does hyper-v allow you to specify how those virtual CPUs are allocated? Consider a quad socket system with four quad core CPUs. If you want to allocate four virtual CPUs to a VM, you can, with other hypervisors, specify whether you want to allocate all four CPUs from the same physical CPU or one from each physical CPU. As far as the client OS is concerned, it would either see one quad core CPUs or four single core CPUs, depending on how they are allocated. Depending on the load on the client, there are some performance differences.

I may be wrong, and it could very well have something to the way it now handles Guest NUMA...but I don't think there is a way to set, like VMware, the number of CPUs and the number of cores. The idea behind Hyper-V is to be smart...and it's going to choose the 4 least busy cores (if you have 4 vCPUs set).

Why would you need a server OS? Because a) you want to use centralised services such as file storage/backups/media streaming around the home and b) you want something that doesn't have loads of desktop-class crap that you don't need, like a glossy desktop environment. This in turn will allow you to run your OS and services on lesser hardware, or make better use of the hardware you already have.

Why would you need a server CPU? The main features of a server-class CPU are:

1) ECC RAM. This is necessary to guard against data being corrupted in memory. Applications which greatly benefit from this are things like databases, simulations, scientific software and software RAID (notice how hardware RAID controllers ALWAYS come with ECC RAM for their cache?). If I was running Linux softare RAID or ZFS on a fileserver, I would use a cheap Xeon and ECC RAM, but then I'm a nerd.

2) Multiple CPU support. In the old days, even single-core desktop-class CPUs could support dual-CPU configurations (ahhh, the Pentium III "Coppermine", how I loved you... ), but Intel and AMD soon realised that they could make more money by restricting dually support to their server class line and charging the earth for it, and now with the advent of multi-core processors, desktop and even most server environments don't really need dual-CPU support.

3) Bigger CPU caches. The L2 and L3 CPU caches on the processor run at full speed, as opposed to main RAM which is several times slower, so the bigger the on-CPU cache the faster data can be processed because more of it can fit wholly in the cache and not have to hauled from main memory. Again this more suits a business environment where more than one user is hitting a server with requests, or for simulation or database software.

Hope this helps.

Linux has a low foot print meaning I can squeeze more out of cheap consumer hardware. After all, there's no point have compositing desktop effects (or even a GUI at all), sound nor anything else if you're just running services like networked file systems

A server OS has much lower overhead than a client OS. An idle windows 2008 server uses less than 100 MB ram if i remember correctly. A server OS is also build to be turned on permanently. A client OS slows down and can start having other problems if it didn't had a restart in a long time.

But that's just two reasons why you would want a server OS on a server. There really is a huge difference with a client OS.

might be (don't know) true for windows, but in linux/unix 'server software' would be defined mainly by how long the distro is testing before release, as just about any linux distro that isn't completly bloated with kde or unity or whatever is pretty light on resources.