Computer systems that don’t provide enough working time are worthless at best and disastrous at worst. But the fine distinction between “available enough” and “99.999% available” can end up hitting the bottom line pretty severely. Rick Cook explains the major options for designing a system with an appropriate amount of uptime.
Designing High-Availability Windows Systems
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
as it stood 10 years ago.
How, in this day and age, can you talk about high availability and not even mention how the servers connect to the network?
or coping with DoS attacks?
oh, and there was nothing in the article about uptime versus cost design tradeoffs, or even an explanation of why the costs go up nonlinerly with the uptime.
or anything about maintenance downtime or system upgrading.
hell of a critic aren’t ya? Are you a college professor? :-p
Nope. Just someone who spent a lot of time designing highly available systems and wonders why OSnews bothers to link to such superficial summaries.
How, in this day and age, can you talk about high availability and not even mention how the servers connect to the network?
Damn, i was going to say that.
or anything about maintenance downtime or system upgrading.
Not to harsh he mentioned RAID5 and data-loss:-)
Dont use windows!
Goto:
http://h20223.www2.hp.com/nonstopcomputing/cache/76385-0-0-0-121.as…
for high availability systems.
Hmmm… my job is in HACMP (High Availability Cluster Multi-Processing) development at IBM, and I thought I knew about all of our competitors. I guess it never occurred to me that HP would have an HA solution, nor has anyone I work with mentioned anything about this.
Taking a quick glance at the website, it seems like they are selling standalone servers with built-in redundancy and local resource swapping at the kernel level. There’s supposed to be an HP ServiceGaurd HAC for HP-UX and Linux, but I can only find literature on the Linux version.
You’d be surprised what you can do with Linux-HA, and there’s Sun Cluster, the only HAC design that uses kernel hooks instead of IPC, but at the end of the day, the only viable competitor to IBM HACMP on AIX or Linux is Veritas Cluster Server. Analysts and CIOs are split on whether HACMP or VCS is the king of HAC. VCS has a more modern implementation and better cross-platform support, while HACMP has the polished integration with AIX and more advanced recovery features (especially across remote sites).
Honestly, VCS is the better choice on Linux for the time being, as the HACMP for Linux port is relatively immature, and it’s probably one of the best options for Windows HAC as well (although I’m not an expert in that area). However, if you want a UNIX HAC solution, HACMP for AIX on System P is what my technical lead calls “the Cadillac” of HAC.
The technology HP is selling comes from former Tandem Computers (if you remember that company). They are selling 7×24 systems since the early ’80. They are running their own OS (Guardian) and not Linux. And sorry I have to say that (I have got experience with IBMs HACMP technology), but what HP (former Tandem) is selling is much more advanced than HACMP.
Edited 2006-06-26 07:13
Am i the only one to which this title sounds like an oxymoron?
I’m not the availability that I’d be concerned about, it’s what happens when you want to really use that “highly available” system continuously with 99.999%. And I wouldn’t say you can’t do that with Windows, because there are some examples, still, in my book, on a 99.999 system the use of Windows is just another unnnecessary risk factor.
two or three Linux servers running Heartbeat & DRBD are a very nice and *cheap* option imho. I’ve set up such a HA and “load balancing” system very easily with two Debian systems, one running Samba, DCHP & DNS, the other one running the internal mailserver + groupware for a small 20 people office. It works flawlessly – if one of the node fails, the other takes over immediately.
Perhaps not an enterprise-grade solution, but for small or midsize companies this works perfectly.
Tom
this article is trying to achieve something you just can’t. High Availability with Windows?
It’s just like writing: How to keep a hungry wild lion from eating my little bunny.
Perhaps the person wants high availability for playing solitaire? 😀
I read this interesting article on Wikipedia the other day:
http://en.wikipedia.org/wiki/Myth_of_the_nines
It’s about reliability and uptime. Quite relevant in this case. Sorry if I’m not too coherant, I’m slightly under the influece of alcohol
My takes on this:
– That wasn’t much of an article, but more of a summary
– a poor excuse of a summary…
– Windows and high availability really IS an oxymoron.
– windows clusters: YAO (Yet Another Oxymoron).
Now maybe i’m beeing too harsh, but IMHO there’s nothing more to it that this.
“wonders why OSnews bothers to link to such superficial summaries.”
I would guess because it’s an informit article.
I think the general idea is that some people have to use Windows and while they are at it they might was well make a HA too. These Admins are very GUI based and most of them can’t write a script much less edit a text file. Now on the oher hand I am an iSeries Admin and I have a few Linux boxes and I can edit a text file and write a shell script and would have a need for a Linux HA or an iSeries HA solution, but then again I see Widnows as the Fisher Prices OS for the Admins who can’t read and need Pictures to help them tell a story. In the end some people need Windows as much as Widnows needs them and that is OK. Just keep it away from me!