OS and Software → No, I Will Not Fix Your #@$!! Computer > When Windows? When *nix?

When Windows? When *nix?

As you all know, several midsize and large networks have a mix of Windows and UNIX/Linux servers. Ill hear about how a server for a particular service will be *nix, such as FTP for example, and Ill wonder "why is this particular server *nix, and the other servers, like file or web servers, Windows?". Certainly for certain specific applications, a server has to be Windows or UNIX, but for something as basic as FTP, web, fileserving, or so on, how do you make the decision on what OS to use? I cant seem to find a pattern.
--
You're so vain... I bet you think this post is about you.

Simple ..

Exchange - windows.
Anything else - Unix.

Done.

Now, if you want to get into WHICH Unix for a particular task (HP-UX, AIX, Solaris, Linux ..) ... that discussion could take days.

^ Indeed. Especially if it's public/Internet-facing.

@dennismurphy probably said it best. As far as the web goes, the internet was developed and runs on *nix so always go with *nix there.

There are a couple factors to consider.

The first is who is going to support these servers? For some organizations with a lot of Windows expertise, then using Windows when it comes to FTP, web, fileserving, and so on would make the most sense. After all, the last thing any company wants to do is spend consulting dollars to administrate Unix if they don't have the expertise in house.

The second is the overall direction of the company IT Manager. I have administrated Windows and Unix servers in a variety of situations, from internet facing to internal only, and I can tell you that a properly administrated server will be very very unlikely to be exploited. That being said, IIS works very well and I am more than willing to support a Windows IIS server without hesitation. Microsoft Exchange and SQL are awesome in themselves. Then I have supported mainframe ERP systems that are solid as a rock. Same goes for Unix. The IT leadership's ability to look at the 1, 3, and 5 year plan for IT and make some good decisions on what platforms and software the company needs to invest in.

In short, making good decisions that benefit the company first before personal preference is the best decision.
--
My domain - Nightfall.net

I have to agree with the others here. Some applications demand certain OSes, Exchange being one of the most common. When there are choices, such as MySQL or Web/FTP serving (which can easily run on *nix or Windows), then you should choose whichever you have the skills to administrate.

I run primarily Windows systems but I do run a few *nix or *nix like systems because they either run better on it or the app I chose to run needs it. My firewall of choice is pfsense, that runs on FreeBSD. My phone server is Asterisk (PBX in a Flash really), that runs on CentOS. Can I find similar apps that would run on Windows? Sure but I don't feel comfortable administrating those (even though I feel more comfortable with Windows over *nix), nor do I feel comfortable getting RRAS or 3CX to replace pfsense/PiaF (though I'm sure they could do the job, I just prefer to stick with what I know).

I agree with this Nightfall.

Also add that they need to look at the evolution of the hardware as well/ And if its a fairly large environment like mine you have to consider power and cooling as well.

I like linux for all my needs due to its cheaper and can run more services with less need to scale the hardware.

Our support staff is less then good with linux but my 5 year plan is to get these guys up to speed with linux and help them advance their careers.
--
"It's always funny until someone gets hurt......and then it's absolutely friggin' hysterical!"

One more vote for Nightfall's explanation.

Or, as I would put it;

Money, experience and direction... get those three working smoothly/well thought out and the rest will follow.

I run a network that needs a 150k a year VM expert...

And it sho don't have one.

Now we pay the price because we are trying to keep a massive and complicated network running on decent Tech pay along with consultants and pricey support contracts. Staff is frustrated because the ferrari isn't running as well as it should when all we really needed was a Honda minivan to begin with.

That's what happens when you are missing (in this case) two out of the three to start with (experience and direction) and maybe too much of the other one. ($$)
--

Mom was right.... I NEED fiber!

It should be called Windows H8

From experience I see this all the time !

Just pay the vm guy 150 and ride the infrastructure till its out of warranty... for some reasons businesses just don't see the loss from the equipment not running at peak performance. They rather nickle and dime everything and every one, instead of nickle and diming the infrastructure.
--
"It's always funny until someone gets hurt......and then it's absolutely friggin' hysterical!"

I'll agree with you on HPUX, but Solaris is far from legacy.

Especially if you're running Oracle databases ...

Oracle is pushing their Oracle Enterprise Linux over Solaris when they aren't pushing their own Exadata appliance. While you can certainly run Solaris/Linux on the Exadata, their company prefers to have people sign up for Linux when there's a choice.

I think that HPUX is on its death bed. AIX survival will depend largely on how well the PureFlex hardware platform will do. Linux is taking off like a rocket. Red Hat seems to have won the corporate Linux wars. Oracle Enterprise Linux will ride on the coat tails of that success since it's a rebranding of Red Hat (with some other modifications). Basically, a RHEL admin can easily support OEL.
--
A government big enough to give you everything you want, is strong enough to take everything you have. -MLK

It sounds like I got the basic answer I was looking for, and thats using *nix for public-facing machines if possible/supportable. Funny how that never occurred to me before. Thank you!
--
You're so vain... I bet you think this post is about you.

You're probably right. I'm not seeing what you're seeing, but I don't exactly see the entire industry.

I was talking to a SAN admin who thinks that SAN is going to die eventually. He was telling me about some vendor seminars he was going to. The SAN is really just the network. It gets hooked up and then the Sys Admin ends up actually managing the storage pools. The vendor gets a contract to take care of any of the physical side of the SAN. The network guys hook up the fiber everywhere. Once it's dropped in place, the SAN admin wouldn't really have anything to do.

VMware is an example of that. IBM's PureFlex hardware means you don't have to go to a SAN admin for AIX either eventually.

Edit: I wouldn't see this happening for a very long time. Those kind of solutions have only been around about two years according to the vendors. There's still a ton of current setups and even future setups that are going to require a SAN admin. Our admin sees it now with Windows on VMware. He just assigns one large LUN, basically, and the Sys Admin manages how the disks are assigned to the hosts.
--
A government big enough to give you everything you want, is strong enough to take everything you have. -MLK

I'm seeing a trend back to local storage in a lot of cases.

Some distributed architectures (like Netezza, Vertica, Hadoop et al) are much more cost and performance efficient with local disk.

SANs are a good idea for certain things, but I'm seeing a move away in some pockets.

Someone needs to give netapp a run for their money in the NAS space. I haven't seen anyone else with a really credible offering. That'll accelerate the move away from SANs more than anything else.

This is the first I've ever heard about anyone going to local disk. I've heard about cache cards that are being used to speed up I/O speeds but never local disk.
--
A government big enough to give you everything you want, is strong enough to take everything you have. -MLK

Why is this? What if you need more storage than what's available inside? This all baffles me. I understand why there's local disk in the old school standalone servers, but in a virtualized environment where you have countless TB of storage being used, why?
--
A government big enough to give you everything you want, is strong enough to take everything you have. -MLK

Netezza/IBM had a failed project that attempted to work with SAN vendors to get the space they needed, the end result was utter failure when it came to performance. They moved back to internal disks and even attempted flash usage, but good flash is just to expensive.

Im seeing a gigantic push to internal drives here. The 10 gig fiber links are saturated and 40 gig is still pricey, internal raid controllers and 24 15 sas drives can crush the 10 gig links now.
--
"It's always funny until someone gets hurt......and then it's absolutely friggin' hysterical!"

1 word... performance.

Indeed, I get to see it every day

I also have seen the trend to local storage, and it's about speed. Raw capacity is pretty cheap now, add in the ability to replicate between hosts for HA failover, and it can be an attractive solution.

The biggest area I've seen local storage pushes so far is for VDI deployments. Still have the SAN on the back end for servers/multi TB storage, 6-12 SSD's in the host for VDI images.
--
How lucky am I to have known someone who is so hard to say good-bye to.

I have 32 slots for 1.8 in SSD's in my servers all 512 GB, capacity isn't an issue any more. It's performance.

My VM hosts all use SSD and I replicate to a JBOD with RAID. The only thing I won't run there is our databases.

But 64 core 1.5 TB ram beasts for the vmhosts never run into any issues besides disk i/o. Once I went to onboard ssd, I don't have that issue any more. Hell I have a box full of sas adapters that I pulled because they just were not fast enough.

Heck even my N6240 could not keep up with the i/o demand, and those are a beast on fiber channel.
--
"It's always funny until someone gets hurt......and then it's absolutely friggin' hysterical!"

The only things i run on Windows are:
- Active Directory: managing PCs in a windows world does not get easier than this. And when compared to others like Novell, AD comes out to be cheaper..
- DNS: Bind is nice and powerful, but configuring it for AD is a PIA..! Not an easy task.
- DHCP: mainly the same as above.
- File Server: this is only because of DFS. If anything in the Nix world can act as a "DNS" for file shares and just redirect the user to where the file actually is (just like DFS in a way), i would like to hear about it..! - I think replication can be easily achieved, but without DFS if a user is connected to a replicated share and that server crashes, the user will lose connection. (maybe gluster has this in the roadmap..?)
- Anti-Virus Server: I wish Sophos would make an Enterprise console for Nix.
- WSUS: doesnt get easier than this.
- ERP Applications: This is my biggest headache.. The ERP we use runs on HP-UX and nothing else, great..!! Except, that since 2007, it requires IIS and .NET for the end user UI...!!
- Warehouse Inventory Control: yep, let's reinvent the wheel. Although, it is dependent on the ERP above, it converts Unidata Pick Style procedures to SQL compliant, then imports that data into a MS SQL server and finally serves it to handhelds via their server app which is windows only...

Everything else, email, webservers, non-ERP sql servers, back up servers, monitoring, VPN, etc runs on Linux...

It is virtually impossible to live in a Windows Free world in my company... I have tried...

The absolute ONLY use for EVER having MS, whether on a server or on a workstation, is the absolute and unavoidable need to use some specific software that only exists for IT and nothing else. That means, generally, niche software for a specific field of work.

Not exchange. That is not a reason, because it doesn't do anything that you can't do with other/better software.

To the OP - others have some great points here. Wish I could have a bit more Linux here, and it might happen someday. The current environment is mostly Windows.

BosstonesOwn -
Man, that's nice.
Wish we could go with some massive PCIe SSDs for some of the databases that we have coming online soon... Instead, we have FC cards, only 2 for each box at that. I have a funny feeling that they will be saturated more quickly than we think, as the SAN, and storage behind it is already fairly loaded. Unfortunately, on-board drive capacity isn't that great on these hosts. Might be able to squeeze in 6 drives though..

It might have to come down to at least a couple of mid-size IODrive2's in each host (couple massive, beastly systems - similar to what you posted) for Data / log / tmp volumes for SQL. I'm just not sure the SAN can handle what we're going to throw at it.