Enterprise open source Toolkit

Story: Row over report praising Windows patching

Posted by: David Wright (Friday 20 May 2005, 10:37 AM)

The pains of rebooting after a patch are the biggest problem with Windows at the moment - at least from a patching perspective.

My firewall (Linux based) has had 6 updates in the last 12 months. One was a Kernel update and required a warm re-boot.

My SuSE is updated regularly and hasn't needed a reboot since it was installed.

Yesterday, I got the Windows Installer update, it required a reboot... Erm, it is a separate program, it isn't the kernel, if the code is properly isolated there is no way it should require a reboot!

And patches like that mean arranging a time with my customers to schedule a reboot, typically after all the users have finished working, that's inconvinient for me and costs the customer more money for out of hours support, hardly makes it cheaper than an open source update, which typically doesn't require bringing the server down.

And if a reboot is necessary on one of the Linux boxes, it takes about 2 minutes to shutdown and reboot, my Windows Server 2003 - running on faster hardware - takes anything up to 8 minutes to boot-up, and then there is the couple of minutes it requires to shut down in the first place...

Rebooting during the day isn't really acceptable for a workstation, but it is fairly well accepted these days - mainly because of MS's dominance of the desktop. What they need to take into account when moving into the server room is that 0 down time is the only acceptable policy - what's the use of RAID array's, redundant power supplies, UPS, multiple network cards to try and ensure that the server runs 24/7, when the operating system keeps bleating for a reboot?

They really need to look closely at how the traditional server market works. Putting their software into the server room and saying that you have to reboot your servers regularly, because that's the way MS software works is not acceptable. They need to improve the quality and reliability of their server software to match what is already there if they want to compete over the long term.

I accept that sometimes the Kernel will need updating, but not every month or every second month, and if the update is to a subsidiary service or driver, then that service or driver should be re-initialised, it shouldn't require the whole machine to be restarted.

I am not anti-Microsfot per-se (I run a mixed Windows/Solaris/Linux environment), but when they come out with rubbish like this, it just makes me laugh. For a IS department using just Windows and thinking of going open source it might scare them enough to stay MS, for a department that uses a variety of platforms, it just makes for a great joke at coffee time with colleagues!

Microsoft can learn a lot by looking at the ease of management and reliability of traditional server OS's (and Linux). Telling people that Unix and Linux are difficult to maintain and don't have proper backup doesn't make it true. They should reduce the budget for the spin department and invest it in producing stable software that meets their customers requirements - and their customers real requirements, not what they think their customers want.