Module wise database Designing

[usamakhalil]

i am working on an ERP. there is a core database. and for reporting data we have replicated databases. i want to ask if any could share knowledge of how large database user applications like Yahoo mail or Orkut are Designed.

more specifically, how DataBase is desinged for application like Yahoo mail or FaceBook.
do they have Distributed, Replicated Database and how?
Thanking in anitcipation

[tesuji]

Great task!
You should also have a look into the ERP branch too. For example SAP R/3, the world largest and most famous ERP system consists of about 18.000 tables (yes, eighteen thousands tables, compared with just 400 tables of MS Dynamics AX 4), 16.000 programs, 60.000 window forms. To install it about 25 GByte are needed only to storing the system's metadata including IDES. Microsoft, a heavy user of R/3 is managing about 4 TBytes of enterprise data with a single global production R/3 environment. Data management system of the 3-tier R/3 usually will be logically installed on one database, no matter the size is. Sure, database is physically spread over many disks, and multifold data mirroring is quit usual for safeguarding against failure. R/3 has its own distributed transaction processing which fully ensures acid no matter database servers are physically located.

What are the metrics of the ERP you are designing on?

(True pieces of advice: You need to have an exact data model (ERM) where almost every essential table derived from must fulfil 3NF without restrictions of any kind. And never try to develop serious ERP based on MySQL)

krs,
tesu

[usamakhalil]

Thanks tesuji.

we have decided to use sql server 2005 as our database. moreover we deliberately are voilating Normalization in Report Data. Data Tables would be designed so that have user load would not hamper performance.

i still wanted to know how yahoo, orkut handle their users data. any of their ER diagram would be a great help to me in understanding distributed databases.

Usama Khalil,