Linked by Eugenia Loli on Wed 15th Mar 2006 16:35 UTC
Databases "Within the past two years, Oracle, IBM and Microsoft have all released freely available versions of their flagship database servers, a move that would have been unheard of just a few years ago. While their respective representatives would argue the move was made in order to better accommodate the needs of all users, it's fairly clear that continued pressure from open source alternatives such as MySQL and PostgreSQL have caused these database juggernauts to rethink their strategies within this increasingly competitive market."
Thread beginning with comment 104737
To read all comments associated with this story, please click here.
PostgreSQL performance scalability
by Don T. Bothers on Wed 15th Mar 2006 19:41 UTC
Don T. Bothers
Member since:

I really need an answer for this from someone who has strong experience with PostgreSQL. We are a company with about 200 concurrent users with very heavy database usage. Our current database is about 250 Gigabytes big and is on SQL Server 2000 running on quad cpu Dell attached to an EMC unit through fiber. We are currently planning for dramatic growth. We estimate that within two years, we will have around 1000 very heavy database users and a 500 Gigabyte database. Furtheremore, within five years, we will have around 2000 very heavy database users and probably a database approaching 1 Terabyte. We need to know how PostgreSQL has scaled for people and whether it can be used to plan our future infrastructure growth.

Reply Score: 1

Joe User Member since:

a database approaching 1 Terabyte

Wow, can I ask you what kind of data you have in this DB? What kind of company?

Reply Parent Score: 1

Don T. Bothers Member since:

It is an ERP database system. With only 200 users, the database already grows about 60 gigabytes per year. We plan to increase the number of employees, increase the number of transacations created per employee, and increase data-mining, analysis, and logging. This is our enterprise core database and it will need high availability and scalability. It will be an enterprise system. We will deploy a redundant fiber SAN, with either EMC or NetApp Storage Arrays. We will most likely deploy a RHEL (AMD64 -> Itanium or Power) solution or Solaris (AMD64 -> Sparc) solution since both provide a path to grow. Although these solutions are pricey, they are actually dwarfed by the cost of Oracle, DB2, or Sybase, especially as we start scaling more and more.

Reply Parent Score: 1

dmantione Member since:

The amount of data is no problem for PostgreSQL, however:
- PostgreSQL needs to analyse its data for statistics regularily. This can take a lot of time if you have a large database.
- Query optimization becomes very important for large data. The PostgreSQL query optimizer often gets it right, but sometimes gets it wrong. In such cases you must manually tweak queries so the optimizer gets it right. This occur s especially when you use a lot of views or subqueries.
- Aggregates in PostgreSQL can be slow, i.e. don't try count(*) on a result with 5 million records.

Reply Parent Score: 1

MaXX Member since:

I suggest you to subscribe to the postgresql-performance mailing list, you'll find a lot of knowledgeable people to help you.


Reply Parent Score: 1