From: Copeland, Daniel (Daniel.Copeland@alcoa.com)
Date: Wed Jan 29 2003 - 14:26:26 EST
Thanks to all that replied...
Beerse, Corne wrote:
Some differences that come to mind:
The license fee for the os (and oracle on that platform?)
The addressable database size does differ.
The OS features can restrict Oracle in one or an other way. Think stuff like
threading, can Oracle coop with (take advantage of) the threading scheme as
available in linux and HP-UX (and between HP-UX versions: 10.20 v.s. 11.x)
The hardware performance does differ. however I don't know how this relates
to Oracle and to your usage.
Having access to the source can be an advantage for you.
Since Linux is more popular, it is more popular among hackers...
In the end, for us, only the license fee and the clients knowledge is a
reason to pick the one or the other (or the M$ database...) since our
systems don't use state of the art oracle stuff and don't grow out of
bounds.
If it is between manageability of the 2 oracle variants, the difference is
in the system details, not in Oracle. If you don't tell the user, they don't
see the difference between HP-UX based or Linux based.
Bill Hassel wrote:
As long as you have *small* databases and *small* numbers of users,
it seems like a reasonable choice. Naturally, you will need people
trained in Linux that also know how to pick the PC hardware so it
will actually work. When I say small databases, terabytes are not
really the hallmark of Linux and PC platforms in general. Also,
ultrahighspeed fibre channel support may be iffy, along with support
for large scale disk arrays (EMC, Hitachi, HP, etc).
Similarly, backup of large databases will need state-of-the-art
tape drives like the Ultrium and this implies SCSI or fibre cards
that can keep up with these tape drives (and of course, drivers
are required to support these brand new tape drives.
As far as users, if you need hundreds to thousands of users, you'll
need to max out the processor count and Linux doesn't handle dozens
of processors yet. You'll also need high speed LAN connections, Gigabit
or better yet, LAN port aggregation which does not exist for Linux
as far as I know.
And there's the reliability issues...failover to a hot spare is not
really ready for production yet (like MC Service Guard).
SO, if you have small databases, a small user count and not much growth
ahead, Linux may be a solution.
About the only issues are with load balancing and failover support
in the Linux kernel. There are some pieces there, most are undergoing
fairly intense revisions. It's hard to argue against Linux is a large
production environment when Google is run with 10,000 Linux boxes.
Just a simple matter of software...
Rita Workman wrote:
I don't run Linux here..but a friend does. He runs HPUX and Linux and
loves them both. Fact is with this combination he has managed to
completely remove anything Windows !! As with many things make certain
what your needs are. Example: Do you need to keep your apps up and
running - or - can you afford downtime? Is there a way to cluster Linux
boxes the same as you can with MC/SG on HPUX boxes? Any potential
problems with any third party software?
Remember, mgmt job is to look at the cost....we need to remember to
point out what they may be sacrificing if they go to something
'smaller'. You may be able to think up a few more things.
Allan Marillier wrote:
I have been trying for a while to get Linux in some environments. Of course,
it is not necessarily appropriate for every environment and application, but
it is a good solution with a lot of cost savings potential. I have 2 servers
under evaluation at the moment, running RedHat Advanced Server. At a cost of
$2500 per year for 24x7 support from RedHat, and lower hardware cost, you
can't lose. The idea is to use the existing V2500 Oracle database servers
with Linux Oracle application (middle tier) servers. We will probably have a
number of Linux app servers with a Piranha load balancing server to handle
load distribution. I considered using cluster manager for HA, but I'm not
sure there is any point. If we have 4 app servers and one fails, the load
will be absorbed via Piranha over the remaining 3. If we cluster and it
fails, where will it fail to? One other server that now does double work, or
a development server that then stops development? Clustering the load
balancing server makes sense, as it then becomes critical to distribute
connections.
I also saw a presentation last year on Linux with Oracle 9i RAC (Real
Application Clusters) where 2 servers saw shared disk, and Oracle handled
failover internally. When one server was shutdown abnormally, the workload
just moved transparently to the other server. RAX requires Oracle enterprise
licensing, not standard, and that also becomes a bit expensive. I've also
heard from people who recently went to Oracle World in California that
Oracle is pushing Linux in a big way, they apparently do all development on
Linux and port to other OSs, and now have a large percentage of their own
production servers in house on Linux as well.
Of course, for ftp, DNS, web and even sendmail servers, I don't think you
can beat a Linux server.
Stefan Farrelly wrote:
We just spent ages working this out as were a big HP-UX and Oracle shop, and
the
verdict was the Oracle licenses cost the same (or very slightly less) on
Itanium+Linux than on HP-UX so why bother going through the massive exercise
of
changing ? Just to save a few pounds on HP-UX OS costs (as Linux is free) ?
HP-UX
license costs are very small now anyway...
Just a little more info...In my scenario, we would have multiple Intel based
Linux Oracle application servers instead of expensive and no more capable
hp-ux boxes (no more capable based on TPM-C ratings) load balanced by a
Cisco LBR. From everyone's responses, there doesn't seem to be any good
reason why this wouldn't work. There is no need for Cluster software (load
balanced replaces this). The only questions is whether the Linux support
will be competent.
Thanks,
Dan
--
---> Please post QUESTIONS and SUMMARIES only!! <---
To subscribe/unsubscribe to this list, contact majordomo@dutchworks.nl
Name: hpux-admin@dutchworks.nl Owner: owner-hpux-admin@dutchworks.nl
Archives: ftp.dutchworks.nl:/pub/digests/hpux-admin (FTP, browse only)
http://www.dutchworks.nl/htbin/hpsysadmin (Web, browse & search)
This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 11:02:25 EDT