Slashdot Mirror
Betting your farm on Linux?
l2b writes "The Standish Group has an interesting article which they talk about the hype surrounding Linux. It goes on to predict division in the Linux ranks. Worth a read, or perhaps entering the
I Bet the Farm on Linux Contest. "
Seems honest to me, linux is not ready for prime time, since it runs best on PCs, which are not designed for high relability, and low downtime. 3 minutes downtime won't take care of a reboot once a year on my machine (My bios takes nearly a minute beofre it even tries to load an OS) Add in that most reboots should be in response to hardware problems and you realise that linux isn't up to the task.
Try this: add a disk drive to a linux system, while the power is on, and you are running with full load. Can't be done easially, controllers that can do this are rare, getting linux to re-scan the hardware chain is a mess. (I think it is doable, but not nicely) Can SMP linux deal with the failure of a processor? How about hot swapping out that processor for anouther? Maybe we should add more processors next month too. What about more memory? Opps, one network controller broke. when you have as much hardware as a high end server has, even though you use the best, the failures happen all the time. You use RAID (support exists, but isn't well tested or robuse enough) and the ability to hot-swap anything. Now granted only a few systems support this hot swapability, a testiment to how difficult it is. (I should know, a major portion of my day job is makeing hot swap work) Linux may never have this, but mission critical requires it.
Where does this leave for linux? EVERYTHING! mission critical is important, but if 3 minutes of downtime (while you do the once a year reboot) will cost you 6 million (not an outragious figgure) Don't you think a company can afford to pay 3 million a year to get hardware and software to work. What about the rest. What is the cost of a small department server going down? Can't be much, as we all know that NT is selling many servers. Now offer to replace Nt with a cheaper system that crashes less? Not much of a savings, but it is there, and it saves headaches. Most importantly, this is the largest number of comptuers! Who cares if the computer on my secritaries desk reboots, she needs more coffee anyway. If the comptuer on my desk reboots (like it did today when I accidently pulled the plug) 4 people care (the ones on X terminals off my system). If the fileserver in the backroom reboots the entire company cares. The downtime is measurable. (10 minutes times 500 people is 2 weeks of downtime, or several thousand dollards.)
Don't forget that except for the Mainframe, most high end comptuers run unix. Linux looks like unix, so it is easy for admins to go between linux and Unix.
No, Linux is not ready for the enterprize, but don't be fooled, the enterprize doesn't consider NT ready either, and NT has the entire marketing muscle of M$ behind them. Of course NT is getting into the enterprize, but not in anything important.
There is one thing to be said for hype about Linux taking the workstation market away from MS in the next year, but at the rate it's going I really think it will at some point in the future .(3-4 years?)
I installed a copy of RedHat 6.0 on one of my development machines at work and demonstrated Gnome and E to a couple of co-workers. One of them is a big NT fanatic.. his first words were "Holy Sh*t!". I thing Linux 2.2.x and Gnome or Kde is really close NOW.. they are easy to use and look pretty. (don't laugh.. that's what most techies judge OSes on)
I think the only things that need work is clean up under the hood.. or at least some good GUI based tools to make it easy for the Windoze user to configure their system, yet leaving a nice console there for the rest of us.
And making sure that there is some common ground for app development between distributions. (IE: standard libs, mabey dir. structures.. etc)
-Ex-Nt-User
Well, this seems a reasonable enough editorial to me. They're right: Linux isn't ready to do the sort of things that Tandem Himalaya, IBM Sysplex and DEC VMS cluster systems can do. I don't think that this is a big deal; nobody wants to pay for a Sysplex system to be a departmental file server, either. Each type of system has its role, and I don't see Linux as going any further than Sun currently does with their systems.
But too, while the hype is annoying, I don't see it doing any real long term damage to Linux. Linux is just the IT industry buzzword du jour, as `client server' and `data warehousing' once were. Eventually people will figure out that it's not the silver bullet that's going to cause a miracle in the industry, and things will settle down. Linux will have reasonable success and lots of people will be happily using it.
cjs
The world's most portable OS: http://www.netbsd.org.
Hi John,
:)
I hope that the backlash from the more forward of my Linux-loving brethren
doesn't just put you into auto-delete-my-mail-mode.. A lot of people see
what they want to see, and read what they want to read, no matter what is
meant..
In any case, I agree with your article, but one part of it is a little
sticky for me, and it's something that I've just plain heard too much of
lately.
Linux is not ready for the enterprise. Correct. Now, explain to me again,
who is saying that Linux _is_ ready for the enterprise? Linus? Alan Cox?
No. For the most part, it's members of the various trade presses, people
who don't understand the rigors of high-transaction computing. It's easy
to tell where someone's coming from; if they mention NT, they don't really
MEAN enterprise computing.
On the other hand, when most people talk about Linux, they're referring to
i386 Linux. From what I understand, other ports do a much better job of
scaling. But, still, I don't think any of the people who actually design
and implement the kernel are fooling themselves. Saying in a press article
that Linux isn't ready for the enterprise is little more than preaching to
the choir.
Regards,
--
Blue Lang
AIX Systems Administrator
IBM Global Services Division
i browse at -1 because they're funnier than you are.
The DMS provides near real-time data to the Operations Centers, showing the current state of the utility's electrical distribution system. Operators can view constantly-updated schematics of the electrical system on linux-based workstations or on a twenty-four foot wide, eight-foot tall projection wall. The entire system was developed using the Linux operating system. Linux computers acting as Front-End Processors (FEP) collect data on the current states of electrical devices such as transformers and circuit breakers. This data is stored in an in-memory data base that is distributed to all DMS servers and workstations. Each computer has a current snapshot of the state of the distribution system.
Real-time data is averaged and archived onto linux-based database/web servers to allow access to historical data throughout the company.
A suite of X Window-based applications lets the operators monitor and control the power system. Linux was selected as the operating system for the DMS because it had proven itself stable in other applications developed by the development group. Most of these applications require high availability, as they monitor systems 24 hours a day. Linux has proven itself through use in the field, with some computers boasting continuous uptimes of two years.
Linux provided extremely easy remote administration of machines, which was particularly important as the DMS operates on machines spread across a wide geographic area. The ability to run a debugger on an X Window-based application that may be running on a machine a hundred miles away, but display its output on a local machine is a huge benefit to quickly pin-point and resolve problems.
The DMS uses Alpha-based machines for servers, and Intel-based machines for workstations.