[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: help with site+database

On Thu, Jul 19, 2001 at 11:38:58AM +0200, Russell Coker wrote:
> Could be worse though.  CA software has install instructions that
> start with "type umask 0" because they want to create directories
> that are world writable for storing binaries (including binaries that
> are run as root from cron jobs).  CA programmers are stupid enough to
> want to do that, and stupid enough to not know how to change the umask
> themselves!

this is either extremely sad or extremely funny...i'm not sure which.

> The only potentially compelling feature of Oracle is multi-master
> replication.  According to the review published in the June 2001
> issue of Linux Magazine there is only one free database that supports
> replication, MySQL (which we agree has deficiencies).  Also MySQL only
> supports single-master replication.

postgresql has some sort of replication add on. my understanding is that
it will be included in the main postgres release in a few versions.

it's by Postgresql Inc, the company formed by several of the postgres
developers...see http://www.pgsql.com/

the version for pg7.1 is downloadable from

i haven't used it, so i have no idea how good it is.

hey, cool... i just noticed they have a mysql to postgresql conversion
tool at http://www.pgsql.com/download/mysql2pgsql.tar.gz

> So when things really went wrong you needed expertise on Sun
> clustering, Veritas, Oracle, and the mail server.  That requires at
> least three people!!!
> The situation of requiring three people from different countries
> working together to solve a problem is not a nice one.

hooray for commercial support! it's really worth paying a fortune for,
isn't it?

> > as far as i can tell, oracle gets recommended by consultants
> > because it's a very recognised brand name, but mostly because
> > of the enormous kickbacks provided by oracle (only "discounts"
> > aka kickbacks greater than 95% have to be approved by senior
> > management...anything less can be just agreed to by the sales
> > consultant).
> You call them kickbacks.  Does Oracle actually give cash to people who
> recommend it's use?

not as far as i know, and i doubt if that's what oracle calls them.
"discount", perhaps. or maybe "consultant's margin". technically
speaking, i guess they're not actually "kickbacks".

if the RRP of program "O" is $10,000 and the consultant buys it at 90%
discount (cost = $1000) but sells it to the customer at 10% discount
(price = $9000), then the consultant pockets $8000 and the customer
thinks he's got a good deal.

it may not actually BE a kickback, but it looks like one and smells like
one and the end-result is basically the same.

this seems to be fairly well known in the industry. just like when
buying cisco gear, anyone who pays anything close to retail price isn't
even trying.


ps: you remember that reiserfs vs xfs benchmark i was going to do when i
got my replacement IBM drive?  well, it arrived today.  

Version  1.01b      ------Sequential Output------ --Sequential Input- --Random-
                    -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
XFS              1G  6335  99 37139  47 10786  15  6062  92 35832  26 137.5   1
XFS raid0        1G  6362  99 75009  80 27923  41  6131  95 71328  60 173.8   2
reiserfs         1G  5726  98 36807  94 12050  20  5988  93 36898  35 129.3   1
reiser raid0     1G  5734  99 38379  98 26326  49  6021  96 66503  70 148.1   2
barracuda        1G  5509  96  7114  20  2763   4  4932  77  7001   7  96.1   1
                    ------Sequential Create------ --------Random Create--------
                    -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files:max:min        /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
XFS              16  2042  85 +++++ +++  2005  75  2081  89 +++++ +++   535  22
XFS raid0        16  2229  99 +++++ +++  2119  98  2181  99 +++++ +++  1024  55
reiserfs         16  5859  99 +++++ +++  6519  99  5776 100 +++++ +++  5735  97
reiser raid0     16  5364  98 +++++ +++  5815 100  5403  97 +++++ +++  5346  97
barracuda        16  4491  79 +++++ +++  6165  99  4843  88 +++++ +++  5637 100

the system is a dual P3-450 with 512MB RAM.

the barracuda lines are from a single ST19171W seagate barracuda scsi
drive (formatted with reiserfs), just for comparison purposes with my
current setup. they were good drives in their day, but i'll be glad to
get rid of the noisy old things.

the other lines are either a single IBM 40GB (60GXP IC35L040AVER07-0) or
a pair of them running with linux raid0, formatted as either reiserfs or

my interpretation of the results is that reiserfs wins on lots of little
files, but xfs is better for sustained throughput on large files, and
xfs uses less CPU.

if i was running a news spool or a large Maildir/ spool, i think i'd
stick with reiserfs but this is my workstation, where i have lots of
large files (incl. huge mbox files) so i think i'll be switching to XFS.

craig sanders <cas@taz.net.au>

Fabricati Diem, PVNC.
 -- motto of the Ankh-Morpork City Watch

Reply to: