Oracle Ready To (Continue) Linux Plunge

alphadogg writes "Rumors are swirling yet again that Oracle wants to get cozier with Linux and at least one financial analyst says customers can expect a tighter Linux-based appliance from the database and application vendor by the end of the month."

Comments

Competition from PostgreSQL and FreeBSD.

[Anonymous Coward]

In certain sectors, we're beginning to really see competition arising from PostgreSQL and FreeBSD. This is especially true where reliability is a serious concern.

With a proper data backup strategy, several Opteron-based servers running FreeBSD and using PostgreSQL as the database can often be used to replace hundreds of Sun servers running Oracle. Often times we see vast performance increases, as PostgreSQL is a leaner product in many respects. If you don't need some of the more advanced features of Oracle, then PostgreSQL is often a perfect alternative.

The BSD licensing of both FreeBSD and PostgreSQL is often seen as a major win, as it allows for licensing costs to be reduced to nothing.

The only downside is that IT administrators can't just go and blame Oracle when things go wrong. But then again, PostgreSQL is far simpler to administer than Oracle, and the training time for DBAs is much less. The potential for problems is significantly decreased.

Re:Competition from PostgreSQL and FreeBSD.

[Anonymous Coward]

Sorry, this post smells like bullshit to me. How about
substantiating at least one of your claims with some
evidence?

Not to say that either FreeBSD or PostgreSQL are poor,
but a "several" opterons replacing hundreds of Sun
servers is a joke, even if the Opteron chip can perform
as well as 10 sparc CPUs.

Either the owner of the installation is incompetent,
which doesn't seem likely given the money they'd have to
blow on the licenses... or you're talking about 50MHz
sparcs from 10 years ago, in which case you are using
"replace" in a misleading and dishonest context.

Re:Competition from PostgreSQL and FreeBSD.

[xilet]

Nah it could well be true. He just is not saying what generation Sparc they are using. I have seen a number of shops scrapping e250's left and right, so I am sure many places have large database setups running on much older sun equipment. My desktop [p4] runs circles around many of our old Sun boxes, granted those sun boxes have been working near flawlessly for well over half a decade. Many places did their 3-5 year server replacement cycle and used that to migrate off of Sun due to the higher reliability of some of the newer x86_64 platforms, I could see a 10-50x increase in performance on it. Also I have seen some really ugly replication setups for distributed databases so condensing it all on one higher powered box can give a nice speed boost. But I see it as less to do with omg bsd 0wnz solaris then it is with Moore's law. Also postgres is far more lightweight then oracle. Oracle can be tuned (and often is) really poorly, it is the same thing that we have seen with some of the microsoft TCO reports against Linux. A fully tuned Windows server with a series of add-ons and hundreds of man hours to tweak it, vs an out of the box install of linspire on a off the shelf desktop. Apples and Oranges. Oh and Oracle sucks.

Re:Competition from PostgreSQL and FreeBSD.

[Doctor Memory]

I can see your point about Moores law — a new Opteron box with a RAID controller and Gigabit ethernet should certainly outperform an older Sun server running 100base-T and U160 drives. However, the biggest obstacle to a wholesale move to a different DB & OS platform is the actual migration costs. Any shop that's so big it has dozens of Sun servers running Oracle is going to have Solaris admins and Oracle DBAs. These people are not going to shrug their shoulders and say "OK, I guess I'm going to do BSD/PostgreSQL now". These people are going to raise hell, provide Incontrovertible Proof that making the switch will lead to Inevitable Doom, and need to be replaced. So now you're going to have to hire new people who don't know your business, and the first thing they're going to have to do is migrate all your business-critical stuff over to the new platforms. Better hope those departing employees weren't too disgruntled to document everything thouroughly (or kept the docs up-to-date if they were concientious). And better hope your new PostgreSQL DBAs know Oracle real well (or be prepared to send your developers to PostgreSQL training), or they're not going to be able to port your thousands of lines of PL/SQL code. Unless, of course, you want to hire some consultants (who may know your industry, but probably not your business) to convert your legacy software to new, untested code and then turn it over to the staff who weren't involved in the conversion to support. Super happy fun time!

So while it seems like a good idea, unless you're working under a mandate from the CIO to replace Oracle or Solaris, it's probably cheaper and less disruptive to just stick with what you've got. Just ask any of those people who still run OpenVMS — they'll tell you!

Does no one remember "Raw Iron"?

[Ayanami Rei]

We've been down this path before. Oracle tried to put together an appliance solution for running 8i, but it never got off the ground. At the time, it was rumored that they were looking at Linux, BSD, and Solaris for the underlying OS.

There was significant pushback from hardware vendors and users for this sort of integration. From users because it was felt that Oracle would abandon the idea of a database that ran on whatever platforms it could, reducing choice in IT departments. From hardware vendors because it meant that only one provider would benefit, and everyone else was afraid that they'd lose the ability to sell Oracle certified configurations.

And Oracle had a hard time finding an easy platform to deploy it on. At the time, Linux and BSD were not as capable for scaling as they are now. And working with Sun would make integrating Solaris expensive.

Now conditions have changed. Solaris is open and modular. BSD and Linux scale more easily, and on more mature N-way platforms. So it might be a good time to revisit the issue.

However, one has to question the value of an Oracle appliance. Because while large companies are happy to dedicate machines to single tasks, smaller firms are more likely to want to have machines serving multiple roles, which may not come easily to an Oracle appliance (or may cost more if it is required to use Oracle-stack implementations of whatever the need is for).
Yet larger companies have budgets to test, configure, and roll out their own database servers anyway. And Oracle is looking at the small to medium sized IT market.

So I don't know if this is going to get much traction. They're going to, at least, have to create a generic server appliance that maybe comes tuned for Oracle, yet can be used for anything.
That might be a winner.