Comments on: Amazon announces simpleDB, Momentum Builds for Simple Databases http://userprimary.net/user/2007/12/14/amazon-announces-simpledb-momentum-builds-for-simple-databases/ Seth Falcon blogs about programming, technology, and life in Seattle Fri, 21 Nov 2008 16:58:22 +0000 http://wordpress.org/?v=2.6.2 By: Jake L http://userprimary.net/user/2007/12/14/amazon-announces-simpledb-momentum-builds-for-simple-databases/#comment-4793 Jake L Tue, 18 Dec 2007 03:59:27 +0000 http://userprimary.net/user/2007/12/14/amazon-announces-simpledb-momentum-builds-for-simple-databases/#comment-4793 Hi Seth, I agree with your point about using thrift. Actually it's a lot more flexible than you think. It was built to handle schema changes with backward computability to older objects. ThruDB, that you mentioned in your post, is built on thrift and I've recently added a number of examples showing how to store search thrift structures in a number of languages. -Jake Hi Seth,

I agree with your point about using thrift. Actually it’s a lot more flexible than you think. It was built to handle schema changes with backward computability to older objects. ThruDB, that you mentioned in your post, is built on thrift and I’ve recently added a number of examples showing how to store search thrift structures in a number of languages.

-Jake

]]> By: Seth http://userprimary.net/user/2007/12/14/amazon-announces-simpledb-momentum-builds-for-simple-databases/#comment-4780 Seth Sat, 15 Dec 2007 01:24:56 +0000 http://userprimary.net/user/2007/12/14/amazon-announces-simpledb-momentum-builds-for-simple-databases/#comment-4780 I agree on 1-4. memcached is certainly a reasonable system to build on top of. I have a feeling that in a memcache-based model you would have seriously limited query capability -- could you do anything other than query by key? To extend on item 2: we want a system that makes it easy to replicate and partition data across machines so that each node maintains all data in memory (while also persisting to disk). On item 3: it is important that the system provide flexibility in terms of the durability guarantees -- there has to be a trade off in order to attain a performant system. I agree on 1-4. memcached is certainly a reasonable system to build on top of. I have a feeling that in a memcache-based model you would have seriously limited query capability — could you do anything other than query by key?

To extend on item 2: we want a system that makes it easy to replicate and partition data across machines so that each node maintains all data in memory (while also persisting to disk).

On item 3: it is important that the system provide flexibility in terms of the durability guarantees — there has to be a trade off in order to attain a performant system.

]]> By: Alex V. http://userprimary.net/user/2007/12/14/amazon-announces-simpledb-momentum-builds-for-simple-databases/#comment-4779 Alex V. Fri, 14 Dec 2007 23:32:44 +0000 http://userprimary.net/user/2007/12/14/amazon-announces-simpledb-momentum-builds-for-simple-databases/#comment-4779 There seem to be a lot of contenders in this space, but no one that's really running away with it. I think we can debate the various nuanced differences of this solution vs. the other, but I'm starting to think that what we're looking for is: 1) A schema-less datastore with single-key access 2) Distributed 3) Persistent 4) Some (but maybe not lots) of query capability. Maybe what we really want is a disk-backed for of memcached? There seem to be a lot of contenders in this space, but no one that’s really running away with it. I think we can debate the various nuanced differences of this solution vs. the other, but I’m starting to think that what we’re looking for is:

1) A schema-less datastore with single-key access
2) Distributed
3) Persistent
4) Some (but maybe not lots) of query capability.

Maybe what we really want is a disk-backed for of memcached?

]]>