NuoDB with Amazon S3

Question

I think about developing my new mobile and web app against NuoDB. I believe in their technology. The thing that I care the most is backup and restore in case of data corruption and read / write speed scalability.

From what I've read, you can use NuoDB with Amazon S3 as the archive node for storing the data. That looks very useful because you aren't limited on how much data you can store on a single bucket. Furthermore, It's easy to backup the data because you can detach the archive node and you get to have a backup of the data which you can restore back by attaching the archive node back and let the other node sync with it.

Because I expect a lot of data (terabytes), it seems like a good solution, because I will not need to export the data, compress it and store it on EBS volume(s).

1) How efficient is S3 as archive node with NuoDB for high read/write scenerios?

2) Is this is a good backup option to use with NuoDB (the one that I described above)?

3) Should I be better to use MySQL Load balancers? I've read that their are many solutions that you can use on Amazon RDS to scale the DB horizontally and without any code and DB modification (ScaleBase?).

Answer 1

Barry Morris here from NuoDB:

1. You are correct that NuoDB can use S3 as it's primary storage. In NuoDB terms you are configuring a Storage Manager for S3.

2. You can have as many other Storage Managers as you want, including others on S3, on a file system somewhere else, or on another supported Key/Value store.

3. All Storage Managers store the full database redundantly, so loss (or switching off) of Storage Managers will not cause data loss as long as one Storage Manager is still running somewhere in the system.

4. If you add an empty Storage Manager to a running database it will catch up to the current state of the database before taking part in database activity. For a very large database this will not be the most efficient way to bring on a new Storage Manager, but it will work.

5. As relates to backup, you can adopt one of a few different strategies:

   a) Do a conventional JDBC/ODBC/SQL level backup.
   b) Take a Storage Manager offline, and rely on the fact that we can restart directly from that Storage Manager if needed.
   c) Some people will feel that redundant storage in multiple storage managers is good enough and they don't need separate backups.

As for performance on S3 it will be load dependent. If you want us to give you some estimates then let us know what you are doing and we can estimate it or run some tests for you.