What is the architecture solution for a synchronization file with distribution systems?

I understand the complexity of the question. This is for self-study. So if you can suggest which resources I can learn from. it also would help.

I want to architect for a system like dropbox. I have never work with distribution system before, So could you give me a  full solution for this problem.
How to sync when servers distribute around the world
How to handle Conflict and file version.
What should have in back-end side?

I am very appreciated if somebody can help.
Dao NguyenApplication DeveloperAsked:
Who is Participating?
btanConnect With a Mentor Exec ConsultantCommented:
i believe key factor for this focus is elasticity for scaling with client agent, and push notification cum real time messaging of distributed object. You can check out stacksync using ObjectMQ (as abstraction layer doing bulk of the work focus stated) as it compared to Dropbox and scheme of similarity. the former is treated as personal cloud in the study. You may want to consider looking at openstack and owncloud as well
Dao NguyenApplication DeveloperAuthor Commented:
I understand the complexity of the question. This is for self-study. So if you can suggest which resources I can learn from. it also would help.

Aaron TomoskyTechnology ConsultantCommented:
I'd suggest you look at some open source solutions and see how they do it. Start with owncloud: https://github.com/owncloud
Upgrade your Question Security!

Your question, your audience. Choose who sees your identity—and your question—with question security.

Gerald ConnollyCommented:
@btan - you might want to leave questions open a bit longer before awarding points, this one has only been open barely 24 hours, and the extra zones only a few hours, the Experts on this forum do come from all around the globe, and live in different times zones
Keith AlabasterEnterprise ArchitectCommented:
Dao Nguyen was the question asker, Btan gave the accepted solution so good luck to him. Just because the points have been awarded already doesn't mean you can't post additional comments to make the answer more rounded or more beneficial :)

I would have been interested to understand which aspect of the architecture you were pursuing - was it as a reference architecture? Solution architecture?

Enterprise Architect
A truly distributed system is inherently problematic, for reasons you seem to be aware of.

Timing, stamping, requires synchronization. This often done by each one getting clock update to atomic clock, UTC/GMT. Then they can further synchronize clock (for assessing most recent update) by comparing clocks with each other, adjusting towards average, such as for the two with the closest value. (Those over an hour or two off can be eliminated from averaging algorithm). Once their time is synchronized, then they can be trusted again.

For data receipt, of more than a single packet, packets are numbered, such as "#1 of 10". Now they can be received out of order until all are present and ready for processing. If one or more are missing after time-out period, request for retransmission of (only the missing) packet is made.

Updates are made via queuing principle (FIFO - first in first out, like line at bank or store checkout lane) as opposed to stacking (LIFO - last in last out - like for food trays). Method to insert into queue used to manage updates arriving out-of-sync. Separate queues for incorporating priority such as first control, then update/writes, then reads (including printing requests).

End result though, is that need usually remains for having a single source being the sole trusted repository for backup and restore.
(thought it was open, got 'neglected' alert', but type slow, missed some earlier comments, got interrupted..
Asked by: Dao Nguyen (not btan)

It would be nice to have closing remark (datestamped) by asker

re: > you might want to leave questions open a bit longer

- Alternatively, close here for part of some questions, open new questions to address other question. Can leave pointers in each to connect the parts.
Dao NguyenApplication DeveloperAuthor Commented:
Sorry for late response. I am really appreciated all help from experts here.
My question is to solve How is the architecture work to implement the system in real life with lastest technology we having?
There are a number of question will be a part of big question.
What distribution framework can be use for this case or we have to manually implement a framework?
How to handle conflict when there are multiple distribution servers there?
It's pretty problematic. I guess I need time to come up more questions.
Keith AlabasterEnterprise ArchitectCommented:
Not a problem  - most often its just a terminology thing. Sounds like you are after a solution architecture.
All Courses

From novice to tech pro — start learning today.