Linux: Now 400 Distributions Strong

Page 5 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
palladin9479, is that why Linux powers 60% of the server market? Because it's unreliable? Is that why more than 90% of today's supercomputers run it? Because it crashes unexpectedly at any given time?

Hmph.

In case it does crash, there are automated scripts that restore its previous state in a matter of seconds (about the time it takes to boot + a few more miliseconds for the services to start). If it's really REALLY bad and it needs a downgrade (which doesn't happen that often in a corporate environment because they're usually tested before it goes live), it can even be done remotely via SSH.

This liability issue is a bunch of crap. Do you think Apple runs on Windows Server? No. They run on FreeBSD. Well, "Mac OS X Server", to be more precise (which according to Wikipedia, "is based largely on the FreeBSD distribution and includes the latest advances from this development community.") This liability you speak of is the thinking of a corporate executive that constantly needs to blame somebody for his own incompetencies. Like I said, test before you go trough with it. Downgrade as a last resort. Linux is a very stable environment. It's very popular in the server sector probably because of the lack of constant rebooting (which would translate as down times, which could mean potential profit losses).
 


I don't believe he said Linux was unreliable, he was talking about Red Hat being unreliable. The person signing off on a big purchase is going to want to know that if something goes wrong it's someone else's problem and not theirs. It doesn't matter how many supercomputers are running Linux.
 
[citation][nom]palladin9479[/nom]And before you get all snarky, you haven't actually provided a product.Let me tell you Linux guys something you might be missing, either on purpose or through a deliberate mental blockade. When you purchase a product and sign a licensing agreement the provider is under a legal obligation to meet your needs. If there is severe glitch with one of the Windows Server 2008 installs we contact MS and within minutes they have someone working on the issue. Same with Oracle and our Solaris systems. When we looked into RHEL Linux they were very shaky on providing us the service we needed, we had the money but there were legal issues of liability should something go really bad. It all boiled down to "we'll help you if we can, but your mostly on your own". This is due to Linux being code from many different providers and thus most of it is licensed on a "as is" basis.Enterprise's are not in the business to "home build" software solutions, there comes with ~zero~ legal liability for the providers. If a kernel crash results in severe service outage, Torvalds isn't going to be on the hook for fixing it. I know none of this matters to home hobbyists or even small business's, you guys don't purchase $100K+ a year support contracts, but to big business it's very important. Senior executives (the guys who sign the purchase requests) can't sleep well at night without knowing there is no chance of something extremely bad happening the next day.Now this isn't to say products based on Linux are this way. RPA's and VMWare ESX are both examples of products who are Linux based. But in both case's the product manufacturers enter into a legal agreement to support their products and take on that liability. Should one of our ESX servers go bat sh!t crazy on us, one phone call to VMWare and help is on the way, they will provide a fix for whatever is ailing us. Same for the recovery point or IDS or one of the many other appliance type devices we have. The manufactures are responsible for testing / implementing security patch's and we hold them to it. In essence this is where the Linux OS fits in the best, the ability for manufactures to not have to develop their own OS from scratch.[/citation]

Snarky? I was just pointing out that a competent IT department could download the entire Debian repo if they desire and host it on a local server. Not really any building to do. Not even a bunch of work. Just an option...
 
Debian IS the product. It can be deployed across your network. You don't have to buy it. (you can purchase pre-made cd's - cheap if that's more your style) You don't have to sign off on a EULA. I ain't telling you what to do, just saying it can be done, inexpensively.
 
[citation][nom]LuckyDucky7[/nom]Why can't they just make a concatenated Linux OS?Call it something like, I don't know, Grand Unified Linux or something.Which can install any package you throw at it.Which works with most plugins and software out there.If you can do that you can get some more adoption since developers know what they're developing for.[/citation]

No no no. If that happens everybody would be using it. The 1% current users wouldn't feel special anymore.
 


Shows you didn't actually read my posts, else you would of seen the several sentences where I discuss RHEL. Their paid support isn't that great, they dress it up nice but it boils down to "if we can help you we will, but we can't always help you with things not written by Red Hat", basically any code or software not designed and made my Red Hat they don't support, which is most of Linux. Makes them good for appliances and packaged systems, not so good for core enterprise systems.

Also whatever you do, you do NOT have your enterprise "build" a solution for anything that will be revenue producing or that may put the company at liability for something. That's just asking to have something bad happen and be the fall guy. So the idea of having a bunch of Linux hobbyists "throw together" a Debian build is complete and utterly ridiculous. You can do this for proof of concept, but not for anything critical.

Of course none of this matters as you guys aren't even reading past the first three sentences. I already said we use Solaris for our core systems, NT is just for the client infrastructure. This is why Linux isn't making much head room in corporate IT. All those numbers you guys dig up are almost entirely appliance application servers, pre-built by a third party and certified for specific functions. I know this because we use them too, everything from Recovery Point Appliances (RPA) to ESXi servers to Sidewinders. We have tons of appliances that do all sorts of interesting things, but none of them are built "in house". We didn't "buy" into Linux, instead we bought a service from a company who happens to use Linux in it's product.

As an example is this entire thread. Here I am, a person who doesn't sign the PO's but instead makes the requests and recommendations for those same POs. And what I recommend tends to get bought. I don't build stuff based on some fan based loyalty complex, no room for that around here. We build based on needs, we have a need for specific function and we determine if we're going to do it in house, from a third party vender, or contract out to another entity. If we're doing it in house then we typically use Solaris on SPARC or NT on Dell x86. Although the NT stuff is done by a separate set of guys (they look as us Unix guys as weird).

Ultimately the Linux fanboys in this thread have failed to convince me to take another look at the various distro's, even though I use CentOS at my own home. If you want to make way into the Corporate IT world and further your cause, stop the crusader attitude and start looking for business solutions.
 



And when something horrible goes wrong? When millions of USD are suddenly on the line your answer to the CIO would be "well at least it wasn't expensive!". We pay 100K+ a year in support contracts, tens of millions of USD per site, billions of USD total for the entire program (world wide). You think the purchase price of an OS is even on our radar? We don't care how much the OS costs, we're interested into the support and guarantee options. We want a prepackaged system that's guaranteed to do what we want when we want it and in exactly the manor we want it. This is why we use Solaris + Oracle + BEAWLS and FF for our client side. And although I hate BEAWLS with the very deepest parts of my soul, it does exactly what it's advertised to do and integrates well with Oracle and Solaris. And with Oracle owning Sun now, I can see Solaris 11 will be even more tightly integrated with this ecosystem.

That is the difference between a hobbyist mentality and an Enterprise IT mentality.
 
[citation][nom]palladin9479[/nom]And when something horrible goes wrong? When millions of USD are suddenly on the line your answer to the CIO would be "well at least it was expensive!". We pay 100K+ a year in support contracts, tens of millions of USD per site, billions of USD total for the entire program (world wide). You think the purchase price of an OS is even on our radar? We don't care how much the OS costs, we're interested into the support and guarantee options. We want a prepackaged system that's guaranteed to do what we want when we want it and in exactly the manor we want it. This is why we use Solaris + Oracle + BEAWLS and FF for our client side. And although I hate BEAWLS with the very deepest parts of my soul, it does exactly what it's advertised to do and integrates well with Oracle and Solaris. And with Oracle owning Sun now, I can see Solaris 11 will be even more tightly integrated with this ecosystem.That is the difference between a hobbyist mentality and an Enterprise IT mentality.[/citation]

Sounds like you have what you need then. Look no further 😉
 



For now yes, but with every major revision we look at multiple options. We thought about migrating to RHEL for hosting but they didn't provide us what we were looking for. And right now one of our central administrative costs is maintaining a single security baseline for everything. Patching isn't just "ohh hey look, there's a patch, lets download it real quick". First you download it and apply it to your lab systems, do your testing to ensure it doesn't break something important, then come up with a patching plan, schedule ASI's, migrate core systems to alternative environments (we maintain two hot sites), patch first site, do checks, then migrate back and patch alternate site. It's becoming extremely manhour intensive to do this, and time = money in this world. The man-hour costs for our IT staff to do a single evening of ASI maintenance exceeds the cost of the W2K8 site license (regular time + overtime). At least Oracle Solaris provides a patch roll-up (ORA CPU) every few months, it's similar to the NT Service Pack idea. Makes patching a one-shot deal that we only have to do every month (Unix), every other week (NT). Still expensive but as long as we can get it in a single salvo it's manageable. What we're really looking for is something similar to WSUS but for Linux / Unix. "Build your own Repo" is not a solution, we have too many disparate systems and that would actually be more expensive then doing it manually.

Anyhow back to the topic at hand. It's been my experience that RHEL is probably the best "Enterprise" class Linux OS available, with CentOS being it's twin brother for the mass's. There are only a few items they need to shore up, one being their service options / guarantees, and the second being their Enterprise Management tools. We don't manage one box, we manage over a hundred with several thousand clients across a rather large geographic area.
 
Status
Not open for further replies.