Amazon EC2（EBS）上的集群数据存储的可能架构是什么？

Question

I am running a Django website (on Apache) deployed on an EC2 instance. Currently it uses a single MySQL instance but ideally I would like to have a cluster. Please advise on the pros and cons on the following possibilities as well as list others if any:

• MySQL Cluster: If I deploy MySQL on several EBS instances, I still need a partitioner to have a distributed cluster. Any solutions for this?

• Cassandra: How does setting Cassandra up go with EC2? Also, if I use Cassandra currently, do I use the ability to use Django Models, or can I write a driver plugin for Django?

• RDS: Does RDS support clusters? So far I only found documentation on Master-Slave architectures.

Thanks,

-Sid

Answer 1

MariaDB 10 will work perfectly for this set up.

We have also been using Percona MySQL cluster on 4 nodes within EC2 for 2 years now, and it has performed flawlessly.

Answer 2

• MySQL Cluster: not sure what you mean by partitioner. You will need at least 3 instances: 1 management and 2 data nodes. MySQL cluster is in-memory database so you will need to large-memory instances for data nodes. Check MySQL Cluster docs for capacity estimation. Also, MySQL Cluster needs high-speed low-latency interconnects to achieve the best performance. EC2 is not the best environment for it.

• Cassandra: should work just fine - http://wiki.apache.org/cassandra/Cassandra%20EC2

• RDS: RDS does not support clusters. High availability does not come easy with EC2 since it does not provide full access to networking features like broadcast/multicast, ARP takeover, etc. Some people are making progress though: http://www.mysqlperformanceblog.com/2010/06/17/high-availability-for-mysql-on-amazon-ec2-part-1-intro/

Just to note: MySQL Cluster and Cassandra are different solutions from MySQL. You will need to rethink how your app stores and accesses the data if you decide to make the switch.