High Available web service focusing on link failure and recovery

[Jiannan Guo]

Hi,

Recently I started a project aims at decentralizing a Moodle e-learning web server to encounter link failure. Here's a detailed description:

Connection here (rural area in Africa) is fragile (60-70% uptime), which is the main problem in this project. And our goal is to enable students to access course content as much as possible.

Thus, I'm thinking to have a local server constantly caching web content and provide accessibility during down-time. Although, due to the interactive nature of online learning (discussion board, quiz etc), the synchronization should be bi-directional between master and slaves. Plus, slave server should be able to provide transparency to end users, record all interactions locally and update master server once link is recovered (race condition and conflicts need to be solved intelligently). These slave servers will be deployed in either Raspberry Pi or other low-power consumption platform powered by solar. Load balancing would be bonus.

In short, the system should share characteristics of web cluster and database replication, but emphasizing disconnected operation. Weak consistency is acceptable

I'm new to MySQL Cluster and I'm wondering whether it is a proper solution for my scenario.

Any advice would be highly appreciated!
Thanks in advance!