Quick Reference to Alternative data storages (themindstorms.blogspot.com)

18 points by alexpopescu ↗ HN
While it may probably not be exhaustive, my intention is to provide a quick reference to BASE systems (Basically Available, Soft State, Eventually consistent, as opposed to ACID: Atomicity, Consistency, Isolation, Durability) that would offer newcomers an overview of the existing projects in the field.

3 comments

[ 2.5 ms ] story [ 12.2 ms ] thread
This is still under work and input from everyone is more than welcome. Please help me fill in the gaps.
First of all, as much as I love Tokyo Tyrant and Tokyo Cabinet I have a hard time seeing it as an "alternative" data storage system. It is an incremental improvement over standard hash/b-tree storage libraries like BDB with a standardized network interface.

I think you are going to find it very difficult to create an accurate yardstick with which to measure performance of these various datastores. Once we leave the world of a monolithic RDBM server there are a lot of variables that come into play. If database X is network-bound when distributed across multiple hosts and database Y bottlenecks on disk i/o then you are going to get completely different results for each depending on the hardware and network layout/infrastructure of each test cluster. Each system probably also has a large batch of configuration options that can influence the results depending on how much information about the cluster can be provided to the storage system. Unless you have a dedicated cluster that you intend on running the tests on I would remove this section. A potentially better option would be to just have a table that lists the best-case and worst-case environments for each system. I would much rather know that system X works best when the entirety of the DB fits in RAM and the network link is fast than have some table of psuedo-measurements that do not really provide useful information.

This is a great idea! I'm not sure if there is enough info out there to fill this in, but it's definitely worth trying. I'll update the post with your idea and see what we get. Thanks.