Windows Server and Linux are growing up fast, but they still aren't a match for the solid, stable Unix systems on the high-end of the server scale.
That's the conclusion of a report from The Yankee Group, which released it 2006 Global Server Reliability Survey on Wednesday. The report compared a number of server operating systems in areas of reliability, down time and recovery.
It found that Windows Server 2003 showed the highest reliability gains, surpassed only by mature Unix-based server operating systems like HP-UX from Hewlett-Packard and Sun Solaris 10 from Sun Microsystems .
Windows Server 2003 had nearly 20 percent more annual uptime in similar deployment scenarios over Red Hat Enterprise Linux.
The Yankee Group found that corporate Linux, Windows and Unix servers experience on average three to five failures per server per year, resulting in 10.0 to 19.5 hours of annual downtime for each server. The down time for Linux systems was longer not due to a software failure, but because Linux often isn't as well-known or that well documented.
"One of the reasons for extended down time often had nothing to do with performance and reliability of the OS," said Laura DiDio, research fellow for application infrastructure and software platforms at The Yankee Group.
"The one random element I can't emphasize enough is I think some of the disparity we see between Red Hat Linux down time and Windows and Unix comes not so much from any inherent flaws in the Linux core kernel, but the unfamiliarity of some of thenetwork administrators with Linux."
When a Linux system fails, it can sometime send a Linux administrator, who likely has less years of experience than a Sun or IBM Unix veteran, scouring the Internet for documentation or a fix.
Overall, DiDio said, all of the server operating system environments have shown markedly improved reliability in recent years, both in hardware and software. The improvements in hardware from Dell, HP and other vendors has given the operating systems a better base on which to run.
This in turn helps the software, which has also improved. With each new release of the Windows server since Windows 2000 was released in 1999, Microsoft has showed a 20 to 30 percent improvement in reliability, said DiDio.
The poorest performing operating system was Debian GNU/Linux, while SUSE Linux from Novell had the best performance, even topping some of the big Unixes. But it's used so little as to be statistically irrelevant, said DiDio.
Another reason Linux may prove shakier is that many shops, more than 50 percent in this survey, were making some level of customization to their Linux environment. Making changes to the kernel and underlying system increases the chances for breakage somewhere in the system.
But don't unplug the mainframe just yet. In terms of power, performance and reliability, DiDio said the only thing that could come close to mainframe performance is the high-end Unix systems from Hewlett-Packard, Sun and IBM. "And even that is debatable."
Author : Andy Patrizio
news source
That's the conclusion of a report from The Yankee Group, which released it 2006 Global Server Reliability Survey on Wednesday. The report compared a number of server operating systems in areas of reliability, down time and recovery.
It found that Windows Server 2003 showed the highest reliability gains, surpassed only by mature Unix-based server operating systems like HP-UX from Hewlett-Packard and Sun Solaris 10 from Sun Microsystems .
Windows Server 2003 had nearly 20 percent more annual uptime in similar deployment scenarios over Red Hat Enterprise Linux.
The Yankee Group found that corporate Linux, Windows and Unix servers experience on average three to five failures per server per year, resulting in 10.0 to 19.5 hours of annual downtime for each server. The down time for Linux systems was longer not due to a software failure, but because Linux often isn't as well-known or that well documented.
"One of the reasons for extended down time often had nothing to do with performance and reliability of the OS," said Laura DiDio, research fellow for application infrastructure and software platforms at The Yankee Group.
"The one random element I can't emphasize enough is I think some of the disparity we see between Red Hat Linux down time and Windows and Unix comes not so much from any inherent flaws in the Linux core kernel, but the unfamiliarity of some of thenetwork administrators with Linux."
When a Linux system fails, it can sometime send a Linux administrator, who likely has less years of experience than a Sun or IBM Unix veteran, scouring the Internet for documentation or a fix.
Overall, DiDio said, all of the server operating system environments have shown markedly improved reliability in recent years, both in hardware and software. The improvements in hardware from Dell, HP and other vendors has given the operating systems a better base on which to run.
This in turn helps the software, which has also improved. With each new release of the Windows server since Windows 2000 was released in 1999, Microsoft has showed a 20 to 30 percent improvement in reliability, said DiDio.
The poorest performing operating system was Debian GNU/Linux, while SUSE Linux from Novell had the best performance, even topping some of the big Unixes. But it's used so little as to be statistically irrelevant, said DiDio.
Another reason Linux may prove shakier is that many shops, more than 50 percent in this survey, were making some level of customization to their Linux environment. Making changes to the kernel and underlying system increases the chances for breakage somewhere in the system.
But don't unplug the mainframe just yet. In terms of power, performance and reliability, DiDio said the only thing that could come close to mainframe performance is the high-end Unix systems from Hewlett-Packard, Sun and IBM. "And even that is debatable."
Author : Andy Patrizio
news source
