Greenplum Database Basics

[jk2012]

Hi,

I'm trying to understand the Greenplum Database for R&D purposes and was hoping to get some clarification on a few points.

As I understand it, a segment host seems to be a physical server and then you get a segment instance per cpu core. Is that correct?

Assuming I've got the above correct, what's the typical configuration? Is it lots of physical servers for the segments, or is it just a few servers with a lot of cores, therefore giving you lots of segment instances?

Finally does anyone know any good resources for learning these sort of basics? I've been reading the admin guide/install guide and looking on greenplum.com, but it all seems to gloss over this sort of architecture detail.

Thank you

[carlo.ascani@2ndquadrant.it]

Hello,

keep in mind that there is not a "typical" or a "correct" configuration, because it depends strongly on different needs.

As I understand it, a segment host seems to be a physical server and then you get a segment instance per cpu core. Is that correct?

Correct. Every node (phisical server) could have as many segment as its cores.

Assuming I've got the above correct, what's the typical configuration? Is it lots of physical servers for the segments, or is it just a few servers with a lot of cores, therefore giving you lots of segment instances?

That is hard to answer, actually. It really depends on your needs and budget and many other things.

I really advice you to contact some specialist to get help.


Regards,

[jk2012]

Thanks for the reply Carlo,

Can anyone recommend any good resources for learning the basics? As mentioned, I've been reading the admin guide/install guide and looking on greenplum.com, but it all seems to gloss over this sort of architecture detail.

Thank you

[carlo.ascani@2ndquadrant.it]

My advice is to begin with the concept of "shared-nothing" architecture.
Starting from wikipedia: http://en.wikipedia.org/wiki/Shared_...g_architecture

Cheers,

[naikdeval]

Since, I've to do analysis on large data and expected to do real time, want to learn Greenplum. Please guide me how can I start for the same.

[gabriele@2ndQuadrant.com]

Hi,

there are several ways you can learn how to use Greenplum. For instance, through a course (my company for instance regularly publishes Data warehousing courses from our catalogue) or through books. It much depends on how familiar you are with the data warehousing discipline. I must confess that having a PostgreSQL background really helps with Greenplum, as the shared nothing architecture can be seen like 'an extension' to PostgreSQL.

Learning Greenplum indeed involves learning several technologies and topics, including SQL, Data modelling, Data warehousing, Dimensional Modelling, ETL, Business Intelligence, scalability, systems administration with Linux, etc. One thing is how to use Greenplum, another thing is also understand why you need Greenplum, what kind of data you want to store in Greenplum and what usage you intend to do with your data.

Please do not hesitate to ask for more advice here, preferably more specific.

Cheers,
Gabriele

[ads]

It's about time we write a "Data Warehousing with Greenplum" book ;-)