• From: "Peter B. Volk" <PeterB.Volk@xxxxxxx>
  • To: <mysql-dde@xxxxxxxxxxxxx>
  • Date: Thu, 5 Jan 2006 22:22:20 +0100

how about this:

Dear Mr. something,

we (Fabricio Mota and I) are enthusiastic MySql users and would like to tribute 
a very interesting Idea to MySql. This Idea (in our oppinion) is a could be a 
feature that could become a key feature of future MySql releases. The 
suggestion is on a new storage engine. In todays industrie data is distributet 
among many and widly seperated manufacuring sites to increase performance and 
to decrease costs. Typically this type of situation is not the easiest for the 
IT department. E.g. a reporting over all sites would involve that the date must 
be collected into 1 central repository for querring. The process of collecting 
all of the data into a central repository is usally a not trivial process and 
can consume many resources in terms of hardware and developer costs. 

The proposed storage engine would solve the problem of not beeing able to query 
all sites at once. Our proposed solution is called Distributed Database Engine. 
The purpose of this engine is to distribute the data in a row granularity with 
based on logistical statistics. This means the following. There are serveral 
Databases in the cluster. On one site (with the id 1) on a table called jobs 
the site of course needs the dataset where siteId=1 the most.  So exaclty these 
datasets are physically stored on the sites MySql server and all of the other 
rows are located on other servers. the Advantage of DDE is that from site 1 we 
can still query data sets with siteId=2 or a different number. Waht the 
DD-Engine does in the backgroud is to query all other servers in the cluster. 
So for the reporting people at site 1 there is no difference where the data set 
is located. During the whole process there are allway statisics beeing taken 
from the whole process. So if i would be cheaper to store a dataset on one 
computer because the cost equation is showed that it is cheaper than the row is 
shipped to the apropiate server. This way you can always have the dataset in te 
right location. 

Analyis have shown that there are many research projects that have analysied 
exaclty this type of query system with mutch success. But there has only been 
Informix to acctually have taken this type of system to production. Neither DB2 
nor Oracle proviade anything like this system.

Of course the complete system could not be described in a short email. So I 
hope that I have awoken your interest in our proposal. Fabricio Mota and I have 
been discussion a design  document, with many details, that we are willing to 
share. Plaese feel free to ask any question if you are intersted. What we would 
like to know if MySql is interested in this type of system and if so what type 
of resources MySql could tribute to the system. Of course we already have and 
will have even more question during the implementation phase of the project. 


Peter Benjamin Volk
MySql-DDE discussion list

Other related posts: