Re: Questions about Postgres and Oracle

From: Gints Plivna <gints.plivna_at_gmail.com>
Date: Mon, 10 Dec 2012 22:07:08 +0200
Message-ID: <CAN2wOq2-0FqmG4EyWQ0gkB2V++egTEcOhNZjQcj2mrZE3FXH1A_at_mail.gmail.com>

So looking at these sentences:

1) Relaxing the availability and reliability is definitely not an option
2) The current oracle database is a mission critical database for our customers
3) Oracle DBs are an outdated, monolithic way to handle data and not

at all scalable
And then choosing PostgreSQL is somehow beyond me. (3) looks like a typical statement from a NoSQL zealot, but then why PostgreSQL - the same old, outdated, non-scalable and non-performant technique?? ;)

So the question is - are there any sensible arguments at all? Licence costs? Application doesn't work well? Oracle is too complex? There are new friends saying PosgreSQL is the best DBMS ever?

I always like to compare databases with cars. If someone uses Ferrari for transporting potatoes, and Ferrari performs badly, it is not Ferraris' fault. If someone tries to force a bus perform as fast as Ferrari then definitely he will fail.

Also sentence "rewriting the current convoluted, complex application" seems like rewriting application from scratch or at least major project per se unless all the logic is in the application and Oracle is more like stupid data store.

Speaking about small databases - there is Oracle Standard edition or SE One. If someone is writing a new app, then I fully understand that there are various options and Oracle is just one of them, but for existing apps (and here I mean written not just for Oracle but for any DBMS) I would question any financial advantage to change DBMS: - how many man-months or probably years we will need to make new app and perform migration
- do we really have people skilled in new DBMSes, i.e. is there any real reason (apart from pure hope for the best) why next app will be better than the previous one?
- do we really understand the new DBMS and know what impact it will have on our app, personnel, administrative procedures etc. - cannot we just throw out some unnecessary paid options (for EE), downgrade edition or do something like that, which most probably won't need much changes in application? If we really really need EE options - do we really really have them in new DBMS, and does it work more or less the same way (just pure option titles don't count)

Of course all that sometimes is worth nothing compared to __management decision__.

Gints

2012/12/10 Sandra Becker <sbecker6925_at_gmail.com>:
> Relaxing the availability and reliability is definitely not an option.
> Also reducing the volume of data to only 30-to-90 days will not meet our
> requirements. Our customer's auditors, as well as the customer's
> themselves would leave us in droves. The current oracle database is a
> mission critical database for our customers.
> Sandy
>
> On Thu, Dec 6, 2012 at 10:57 PM, Paresh Yadav <yparesh_at_gmail.com> wrote:
>
>> In addition to excellent points mentioned by many posters above:
>>
>> Recently one of my client decided to move to PostgreSQL after getting
>> frustrated with Oracle`s licensing costs and ..... The migration is in
>> progress so we will know the results in few months. We had done some
>> back-off tests for PostgreSQL against Oracle and what I had come to the
>> conclusion is that for a 2 TB ODS database that will grow by about 700 GB
>> per year up to total of 5 TB PostgreSQL won't satisfy the performance
>> and manageability requirements. We decided to relax the database storage /
>> size requirements by storing only 30 or 90 days partial data as needed by
>> various application specific databases instead of a unified single massive
>> Oracle database. We also relaxed some of the availability and reliability
>> requirements because we think PostgreSQL can't meet them without complex
>> architecture involving clustering etc, would like to know your feedback
>> about our decision above.
>>
>>
>> - PostgreSQL documentation recommends maximum of about 100 partition
>> per table per instance. Beyond which you are expected to use clustering.
>> - No partition wise join etc.
>> - As someone mentioned limited support for partition which my manager
>> put nicely as `PostgreSQL lets you manage your own partitions!`. This is
>> similar to how it was done in Oracle 7.3, circa 1997.
>> - Extremely primitive Query optimizer
>> - If you are spoiled by AWR, OEM, Hints, SQL Profiles, intelligent
>> optimizer in Oracle, you will find PostgreSQL lacking many of those features
>>
>>
>> IMHO PostgreSQL will be excellent replacement for Oracle for small low
>> volume / load databases. I wish to see something rise and challenge Oracle
>> so as to have a healthy competition in the market.
>>
>> Cheers,
>> Paresh
>>
>>
>> On Thu, Dec 6, 2012 at 4:02 PM, Jared Still <jkstill_at_gmail.com> wrote:
>>
>>> On Thu, Dec 6, 2012 at 8:33 AM, Sandra Becker <sbecker6925_at_gmail.com>
>>> wrote:
>>> > "Oracle DBs are an outdated, monolithic way to handle data
>>> > and not at all scalable".
>>> >
>>> >
>>> That is pure FUD.
>>>
>>>
>>> Jared Still
>>> Certifiable Oracle DBA and Part Time Perl Evangelist
>>> Oracle Blog: http://jkstill.blogspot.com
>>> Home Page: http://jaredstill.com
>>>
>>>
>>> --
>>> http://www.freelists.org/webpage/oracle-l
>>>
>>>
>>>
>>
>>
>> --
>> Thanks
>> Paresh
>> 416-688-1003
>>
>>
>
>
> --
> Sandy
> Transzap, Inc.
>
>
> --
> http://www.freelists.org/webpage/oracle-l
>
>

--
http://www.freelists.org/webpage/oracle-l

Received on Mon Dec 10 2012 - 21:07:08 CET