Hello,
I need to design an object-oriented clustering technique which need to be implemented into the mobile devices. Some concerns need to consider. The relationships between objects such as inheritance, aggregation and association need to cluster close to each other and re-clustering is not in the consideration in the design. I have think for quite a long time and do not have any idea yet. Could anyone proposed me some idea?Thank you.

Hello, actuallt this is quite urgent for me.
Am I posting the question in the wrong section?
If i'm posting it into the wrong section, can anybody suggests to me the right section to post?
Your kindly help will be appreciated the most.
Thank you

um.. maybe I am not clear on what you are asking, but I don't see how this is related to database design.

Do you have more information possibly? Are you talking about a distributed system or something specific to databases?

sorry 'object-oriented clustering technique' doesn't mean anything to me.

Actually I'm trying to develop an object-oriented database management system for PDAs. One of the issue to design the database is clustering technique. Now I'm having some problems in designing the clustering technique.

Clustering is the technique to cluster objects into partitions to improve the performance of the database. Thus, I urgently need help in designing an appropriate clustering technique for limited resources PDAs.

you are trying to split a DB across PDAs? that seems... difficult and confusing

well.. I will leave now. I cannot help you. I haven't messed with clustering yet (even at a regular level), but some of the more experienced DB people around here have. good luck!

Wow... You're starting from scratch?

Have you resolved the Lemma/Thunk issues yet? Which communication model are you using? Have you decided on a hardware prototype, or are you using a collaborative participation model (such as a java quorum) that doesn't address the hardware?

Thinking on a slightly more abstract level, what do you envision using the database to do? How robust does it need to be? Do you want a single authoritative source, or will you settle for autonomous response? How much redundancy do you need to provide the desired level of reliable response, and does the network topography matter for reliable response (do you need to serve the whole cloud, or only specific access points)?

That brings me back to "wow". Do you have any formal specification, or are you just "winging it" as you go?

-PatP

Wouldn't it be better to just get the PDA's to access a central database? The issues with your approach seem to be:

• Performance - to access all the data you'd need to spread your query around each PDA and wait for all the result to come back. Even ignoring all the data transport issues (queuing, redundancy etc) this seems slower than accessing a central server and getting the data in one hit.
  
• Reliability - if a PDA is turned off then it can't give access to it's part of the database, even with redundancy you'd still have a risk of some data not being available. This doesn't happen with traditional databases.
  
• Battery life - is a big issue with most PDA's, I only turn my ones on if I want to do something with it. If I held a part of a database on my PDA then that data data would not be available for others if I have it turned off or I'd have to be permanently recharging it as it's allways connected to the WIFI.
  
• Cost - I can hire a huge and fast MySQL database server on the internet for very little money but PDA memory is expensive and their processors are very slow. It just seems like the most expensive solution to a problem that doesn't really exist.
  
• Complexity - the most succesfull systems are those that do something useful and are easy to implement and maintain. It's not at all clear what the advantages of your system are over a traditional approach and it looks very complex to implement.

Could you explain what benefits your system could provide? On the positive side it certainly looks more interesting that the work I'm currently being paid to do

Mike

Well, this is actually my research topic and it is the topic where decided by my supervisor. I don't have much understanding yet and to be honest, ya, all will be started from scratch.

Can you point out some important points which i should consider in designing its algorithm or any suitable algorithm which already purposed by other people can be used?

Thank you

It would help us a bunch if we could see your assignment. Can you either scan the assignment paper and post it here, or give us a URL so that we can see exactly what you are being requested to accomplish?

A simple network design from scratch takes more man-months than you could possibly have to dedicate to a research topic. There must be something in the specification that you are missing, and we NEED to know what that is in order to avoid sending you on wild goose chases.

-PatP