Hardware Configuration and other Stuff

From: Vivekanand Joshi <vjoshi(at)zetainteractive(dot)com>
To: pgsql-performance(at)postgresql(dot)org, pgsql-admin(at)postgresql(dot)org
Subject: Hardware Configuration and other Stuff
Date: 2015-03-18 19:07:49
Message-ID: 82780c1beb8bc319692c2455c5c3ab38@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-admin pgsql-performance

Hi Team,

I don't know under which section does this question comes, so I am posting
this question to both Admin and performance mailing list. Apologies in
advance.

Objective:

We are planning to use PostgreSQL instead of Netezza for our data warehouse
as well as database solutions. Right now, we have all our clients in one
Netezza box. What we are thinking of migrating our clients to dedicated
PostgreSQL for each of them. We will start with one of the client. If it
works successfully, we will be migrating all the clients one by one. The
objective is to get a better performance than our existing solution. We are
hopeful of that mainly because of two reasons. Firstly, we will have a
dedicated server for each of the client with good hardware instead of
having one server with all the clients on that. Secondly, we can spend on
hardware much easily than spending on a proprietary appliance.

I am hoping this community can help us to know that what would be the good
infrastructure/hardware that can help us in achieving our goal.

Here are few of the statistics which might act as a starting point.

Availability: High (24*7).

User Data : 700 GB which will increase to 1.5 TB in next 2-3 years.

Number of User Databases : 2 (One is the main database, other is used only
for working tables where tables gets deleted in every 48 hours)

Number of tables : 200 (in the main database), (2000-3000 in working
database)

Size of top 5 biggest tables : 20-40 GB

No of users concurrently accessing the system : 5-6 with write access. 10
with read access.

No of User Queries running on the system in a day : ~80K

Read-only Queries (Select): ~60K

Write queries: ~20K

Data Import Queries: ~1K

Typical Business Day : 18-20 hours.

I can pass on few complex queries to let you guys know what are we doing.

Here are few questions:

1.) I don't need a load balancing solution. It must be high availability
server and I can work with asynchronous replication. The most important
thing here would be recovery should be as fast as possible.

What approach would you recommend?

2.) Recommendations on indexes, WAL, table spaces. I am not asking about on
which key I need to make indexes, but an high level approach about how to
keep them? This might come out as a weird question to many but please
excuse me for being a novice.

*Most Important Question:*

3.) What would be the ideal hardware configuration for this requirement? I
know there is not a one-stop answer for this, but let's take it is a
starting point. We can come to a proper conclusion after a discussion.

What are the best on-line resources/books which can tell us about the
hardware requirements?

Warm Regards,

Vivekanand Joshi
+919654227927

[image: Zeta Interactive]

185 Madison Ave. New York, NY 10016

www.zetainteractive.com

Responses

Browse pgsql-admin by date

  From Date Subject
Next Message Joshua D. Drake 2015-03-18 19:37:40 Re: [ADMIN] Hardware Configuration and other Stuff
Previous Message Devrim Gündüz 2015-03-18 18:13:26 Re: libs, compat and upgrades (oh my)

Browse pgsql-performance by date

  From Date Subject
Next Message Tomas Vondra 2015-03-18 19:23:27 Re: Performance issues
Previous Message Vitalii Tymchyshyn 2015-03-18 19:07:31 Re: Performance issues