Unix high availability and scalability survey

we're in the process of reviewing of unix infrastructure main objective is to consolidate on the less versions possible
key decision factors are scalability and high availability options given our multi-datacenter infrastructure, features like HP's continental cluster are top on our wish list based on the forum experiences, what are the relative strength of the various brands? I'll try a little list first:

RedHat Linux:

1. + GFS for active-active clusters
2. + runs on commodity hw
3. - no geographic cluster

HP-UX:

1. + great scalability from blade servers to superdome
2. + continental cluster
3. - itanium

Solaris:

1. + runs even on x86 (vmware consolidation with windows/linux farms)
2. + scalability up to M-series

does it have some form of geographically dispersed cluster?

AIX:

1. + does it have any pluses?
2. - IBM

is it scalable enough? how well it behaves on clustering?

One of the biggest mistakes most organizations make is to base a decision purely on technology or technical details.

What is generally the most important factor is this:

• What is the current skill base of your employees?
• What systems are they expert in?
• What systems are they trained it?
• What systems do they use now?

Many companies have failed when they try to make a major technological shift that does match the skill base of their employees.

The most important factor is what the current technical team is most skilled and comfortable with, not technology.

this is sound but out of scope right now
we're currently upgrading our skill set, so we can steer one direction or another based on the best technology

How big is your organization?

How large is your IT staff or department?

The correct "technology path" always resides in the culture of the organization, not the best technology.

while these could be classified information, i can say we have multiple datacenters at more than SAN extension allowed distance, and we can count on a dozen of various skilled sysadmins

I doubt it... most companies publish the size of their organization and this is not considered "classified" or "confidential".

So, if you are not willing to share, we can't help you because you refuse to address the most important part of the discussion, the O&M side.

In fact, you have shared that the company is very large. This means that the most important decision is certainly not based on technology, but organization.

Furthmore, generally the "overly technical discussions" are not helpful because they distract from the most important discussion, which are the O&M discussions of existing skillsets, licenses, etc.

If are unwilling to discuss the important facts, we will close this thread, sorry. It is not possible to correctly advise you without the details, and we are not here to support "one technology is better than other" type of techie discussions (Rule #8).

This is because we have a lot of experience, and we know what is important, and what is not helpful.

May I just add that it also depends on the minimum service levels that are acceptable to your business and your customers. Not every technology may enable those service levels. If you could identify requirements that are crucial, then this may allow you to ask a more specific technical question as to which solutions meet those requirements (would that be allowed Neo?)

Yes, we certainly need more information. Budget is also important, which often related to how much "risk" an organization can afford.

Without addressing the organizational (cultural), budget and risk issues, it would be wrong to recommend one technology over the other.

As a side editorial story, there was once a publicly traded company (Broadvision, as I recall) that had a web portal product based on C++. The analysts strongly criticized that company for not being more open (Java-based). The company basically tried to force their C++ developers to become Java experts because Java was the "best technology". The company rapidly when downhill to bankruptcy.

One more story.

As a consultant many years ago, I was brought into a very large government organization that was Solaris based and was told management made the decision to change the entire infrastructure to HP-UX. They had tons and tons of HP-UX servers sitting on the docks; and all the Solaris employees were told to install them. It was a total disaster. It was nearly criminal sabotage going on, because these long time dedicated Solaris employees were not just going to become HP-UX lovers and leave behind Solaris, which they loved.

Now, I'm not posting here to say if HP-UX or Solaris was better for that organization. What I am saying is that if an organization has a certain culture, destroying that culture for a different technology that is not a part of the culture will fail, and generally always goes.

So, most reports, like the one requested by the original poster, are useless, technically. In fact, they generally are both destructive and counterproductive, especially in large organizations. Yes, of course if you are a sales person from "YYY Corporation" and your competition is "ZZZ", you really don't care about culture, because you want to see ZZZ replace YYY. Or, if you are a die hard techie, you really don't care about organization and culture, because you are blinded by technology.

So, what is useful, is to first have an understand of the organizational culture. Advising without these details just puts ammunition into the hands of reckless people who will advocate a decision based on the wrong focus.

I cannot stress this strongly enough. Culture is much more important than technology, very much so. As the old saying goes:

All Politics are Local

Only foolish people make IT decisions without considering the business and social impact of their actions.

let's say this could be an unlimited resources scenario:
if you had to choose a platform for maximun HA and possibly Active-Active configuration over multiple sites, based on you experiences would you choose HP-UX over Solaris? or AIX over Linux?
real numbers are not intersting at this stage
i would be interested in some reports from the trenches like "HP clusters are too slow to converge" or "stay out of AIX because support sucks" or "solaris clusters saved our jobs a couple of times" (just hypothetical examples, they don't really are that way)

---------- Post updated at 12:43 PM ---------- Previous update was at 12:39 PM ----------

as for the "cultural" side right now we have a strong Win background so every unix-like choice would be rather revolutionary
i'm just trying to address this revolution the most proficient way

And this is a rule violation here in our forums, clearly a Rule (8) violation.

Sorry, we are not going to do your analysis for you or get into the middle of your internal conflicts / revolutions. We don't even permit Windows versus Linux version Unix threads.. much less HP v. Sun v. Linux... etc. If you have "unlimited resources" (as you asked us to assume), go hire a professional services firm to help you.

Thread closed.