Netcraft has produced their monthly rating of reliability of hosting companies. Out of 41 listed, only 5 run that other OS while 24 run GNU/Linux. That shows what happens when people actually have a choice in supplier of software. The best tool for the job gets chosen. Further, IDC reports that M$ makes way more revenue than anyone else for server software so the folks getting the third-best software on servers are actually paying a premium for it… Sad.
Writing
3435 articles30692 comments
Comments
platforms
17519
12812
206
3
2browsers
23981 
11884 
11736 
4664 
4286 
1643 
198 
14 
2 
0 
0 
0 
“M$ makes way more revenue than anyone else for server software so the folks getting the third-best software on servers”
You know… it could be that the server failure have nothing to do with the OS, but with the hardware. The people who choose FreeBSD or Linux, instead of paying for an OS, used that money for better hardware.
So what your saying is:
People who make good decisions tend to make sound choices; and people who make bad decisions tend to make bad choices?
I can’t really argue…
face-palms aside.
it may also be safe to assume that people who rank in the top 50 on this list, all are purchase very nice kit with good redundancy. (heck I got a raid-5 in my sub-$1000 desktop…)
given that all server hardware has redundancies at all levels up-time boils down to software:
redundancies are easy:
-OS lives on a cluster
-each node of the cluster has hardware redundancy
-all dual/triple tapped to the network
-with multiple network services
-all fed from UPS
-etc…
only thing NOT redundant in the list is the OS.
“That shows what happens when people actually have a choice in supplier of software”
I do not think that the information in the cites supports that conclusion. For one thing, Linux is used for some 60% or more of web servers and so accounting for 60% of the top 41 vendors by any selection process is not remarkable. I would wager that it is found on 60% of the worst performing hosting services as well.
The list shows nothing about choice in selecting software.
Well Robert, it’s obvious that someone sprinkled magic pixie dust over the servers in those top 41 listed hosting companies. How else would they be the top 41? Oh, and by the way, aliens with their flying saucers built the pyramids.
Yep. People tell me that server software is simple and easy so GNU/Linux can run it but that real men need desktops where only M$ can hack it. My mind boggles. If server software is so simple, M$ should be 100% available and snappy. Any servers I have seen running 2K or 2003 have been sluggish and stubborn. That other OS always seems to be doing something other than what the user wants.
“. If server software is so simple, M$ should be 100% available and snappy. Any servers I have seen running 2K or 2003 have been sluggish and stubborn.”
2000? 2003? Old News Pog.
My shop is running some 200+ windows servers out of a total of nearly 500 servers. many of these running components of enterprise line of business applications. They are not only quite responsive, but reliable as well.
But perhaps we are just unique, eh?
Ah, but generally you run one service on a Windows server only, while on a Unix/Linux/OSX server you’ll be running several services, so you’ll save on the number of servers needed. Every time I’ve run the costs, Windows is uneconomical.
Those places that are using it, usually are using it to run things like Exchange or Sharepoint. Exchange and Sharepoint are excessively expensive as well.
And of course you have the problem of dealing with a company that is projected to be bankrupt by the fall of 2014. Yes, I know. That’s my projection, and no one else is agreeing with me. But no one else has managed to provide any proof that I’m wrong, and I am using Microsoft’s own numbers…
Wayne
“Ah, but generally you run one service on a Windows server only, while on a Unix/Linux/OSX server you’ll be running several services, so you’ll save on the number of servers needed.”
Nope. Most of our Enterprise linux instances are dedicated function running single services as well. furthermore single service/function per instance is also mandated for certain applications for conformance (PCI for instance).
As far as you predictions of bankruptcy are concerned, Meh…
Single service makes some sense when a server is working hard with that service but the majority of servers in the world were idling. That’s why they use virtualization now to keep server counts down and maximize ROI. Small shops and individuals can get great service from a small number of servers running multiple services. I did work in one school that had 7 servers but they were idiots… Several different folks were in charge of different services and there was no trust at all etc. Heating was a problem in that server room, so they left the door open to a student-filled hallway. One server running GNU/Linux could run the whole place. I installed a virtual machine with GNU/Linux and put more services on it than the whole rest of the system and my instance was faster than all those 2K and 2K3 instances.
“Single Service?”
To put this into a personal perspective, I work for Microsoft and I’m asking for a slice of a machine. Even a virtualised slice; I don’t care.
I’m probably going to end up with a dedicated server (for something that even I think is a waste of time), and that is silly. I might even get redundancy with other machines, which is less silly, and possibly worth it at say a thousand a pop.
But seriously. Any *nix server I have ever come across runs at least twenty services (aka daemons, and I can write one for you if you ask) at one time. They even include stupid stuff like time synchronisers.
And, you know what? Every single Windoze server I have ever come across does exactly the same thing.
It’s just a dozy little bit of well-understood kernel and sysadmin.
There is genuinely no observable difference in terms of “can I do this with that?”.
Beyond that, it’s down to reliability and TCOS and the cost of administration. Linux on the desktop was never there. Linux on the server is going backwards.
Dr Loser wrote, “There is genuinely no observable difference in terms of “can I do this with that?””
You jest.
Until Server 2008, that other OS burdened its servers with a GUI… Most people I know are still running 2003 with a GUI. That’s a huge load of resources. One graphical session slows the whole mess down.
The EULA for Server requires a fee to be paid for each client. That means if you have x clients and x CALs and you want to add 10 more clients, you are legally stopped until you buy more clients. What kind of GNU/Linux server does that?
Then there is performance. I was using 2003 to authenticate at one place and we could watch the packets on the wire. There were delays up to 30s for a server that was idling. I have also seen 2003 crash when I logged in remotely with RDP from XP. When I did the same login from GNU/Linux with rdesktop, 2003 kept going. What’s with that? That other OS makes no sense to me on a server upon which many clients depend. I could go on and on but it’s a waste of time, I am sure.
In terms of multiple services, I give you some terminal servers I have set up with GNU/Linux. I run all the applications and desktop processes for a large number of simultaneous users, say 24, and a web server and database for other resources students can use from the browser. I can do all of that in less than 2gB, and it is more responsive than the typical desktop of that other OS because lots of stuff is cached and shared between processes. Instead of having 24 clients giving poor service, I can have 24 giving great service on the same hardware, sometimes ancient stuff. Where I worked a few years ago, folks had seven server for about 100 clients and I could easily do everything with 1 or 2 GNU/Linux servers.
Perhaps you could stop using Windows Server 2003 and move up to Windows Server 2008?
Assuming that you tried 2003 in the first place.
“Sometimes ancient stuff?”
Suck it up, mate. Buy recent hardware. Moore’s Law and all, plus busted caps.
Here’s a set of anecdotes for you.
When I started with computers, back in around 1987 or so, we worked on exactly the same principles you claim: thin clients (3270 terminals, typically) with a single Multics server (VOS, a proper OS, accidentally fault-tolerant).
We served, for example, every single Travel Agent in Australasia. On three or four six-processor servers.
I’d absolutely love to see GNU/Linux (and why the heck do you still insist on the GNU bit?) servers do this for Ohio alone.
It is a wretched bit of failed nonsense that simply leads idiot managers astray. And you are one of the idiots putting accidental money in their pockets.
Good for you!
I don’t know how many travel agents that is but you probably could not do it with that other OS and GNU/Linux loves thin clients. Largo, FL, runs about 400 users on a server with GNU/Linux. I don’t know what the limit is except RAM, CPU and network. Machines that I build rarely need to serve more than 30 users but almost any COTS PC with some extra drives can do it handily. Cost is about $25 worth of server per client, typically.
Back in 1990.
Linux wasn’t there at that point.
GNU/Linux loves “thin clients?”
Maybe. Got an example of that?
And of course 3000 clients (1988) is just 30 clients (2011) multiplied by a hundred.
You’ve got a bit of learning to force yourself through, Robert. It really doesn’t work that way.
And, were they running full desktops or merely using local clients and transmitting a transaction back and forth?
Largo, FL, Munich DE, my schools, etc.
Thin clients have had had slow but steady growth for years. Now they are exploding. Even M$ is selling software for thin clients.
“And of course you have the problem of dealing with a company that is projected to be bankrupt by the fall of 2014. ”
Welcome to the club!
http://penguinday.wordpress.com/2010/08/11/the-world-according-to-freetards/
http://penguinday.wordpress.com/2010/08/11/ten-years-later-super-loon-edition/
Well, I am a fan of FLOSS and I don’t think M$ will be bankrupt by 2014, but their margins will be much reduced. The world just doesn’t need their cash-cows.
While their current share of running PCs looks good. The share of PCs sold with that other OS keeps declining. That is below 80%. There is no sign of a slowdown and with the rise of tablets and smart phones share will plunge while the rest of the world prospers.
Realizing that my job is not IT support, I have been tasked to set up servers in the past at different times for groups I have been in (some internal Web, some print and file, some dial-in pool, some ftp, some mail server). Why was it nearly everytime, despite all the fancy manuals I would try that other OS in its flavor of the day NT4, 2000Server, 2003 and after much ahir pulling and swearing give up. No wonder you need all those specialized training classes certificate programs. And everytime, the Penguin rode in to the rescue, usually in one of those “Learn Linux” books with the CD of the distro of the day glued on it; though distros are LOADED will well commented setup files, and “how-tos” especially Slackware. All that was needed was some basic reasoning and success came. I never understood how MS grabbed even a part of the server market at least for the simple UNIX lower case-letter socket basics (telnet, ftp, http, smtp, ntp, samba, etc.)
The POS that is the monopoly rests on convincing the world they need that other OS, making it so complex that using a server helped to manage machines and overcharging for everything. M$ doesn’t make it on web servers because there is no need for that other OS there.