Apple is (or was) the Biggest User of Apache Cassandra
One thing I did not realize about Cassandra is that Apple is (or was) one of the biggest Cassandra users out there:
The article further states that FoundationDB can operate on cheaper hardware, less nodes, and faster. It states that Apple could reduce their cluster size by 5-10%.
5-10% off of a cluster that size is not something to be sneezed at. We are talking upwards of 7500 servers and millions of dollars in savings in hardware and even more devops costs.
Since RAM is the new disk, disk is the new tape an in-memory data store backed by a disk is going to support reads that are orders of magnitude faster than a data store like Cassandra that uses disk as a primary storage mechanism. For example, Redis has a data model that is similar to Cassandra but it is entirely in-memory.
Of course, it all depends on your requirements. If your needs are to accumulate massive amounts of information that is queried infrequently or in off-peak batches, then Cassandra is very appropriate. But if you require consistent performance for both reads and writes you should look elsewhere.
Word in Goldmacher's circles is that Apple will be “replacing” its huge Cassandra noSQL implementation with FoundationDB. Apple uses Cassandra for “iMessage, iTunes passwords, a bunch of stuff,” he says.
In fact, Apple is touted as having one of the largest production deployments of Cassandra of all, with over 75,000 nodes storing over 10 petabytes of data. Cassandra is a free and open source database with a commercial version offered by DataStax
The article further states that FoundationDB can operate on cheaper hardware, less nodes, and faster. It states that Apple could reduce their cluster size by 5-10%.
5-10% off of a cluster that size is not something to be sneezed at. We are talking upwards of 7500 servers and millions of dollars in savings in hardware and even more devops costs.
Since RAM is the new disk, disk is the new tape an in-memory data store backed by a disk is going to support reads that are orders of magnitude faster than a data store like Cassandra that uses disk as a primary storage mechanism. For example, Redis has a data model that is similar to Cassandra but it is entirely in-memory.
Of course, it all depends on your requirements. If your needs are to accumulate massive amounts of information that is queried infrequently or in off-peak batches, then Cassandra is very appropriate. But if you require consistent performance for both reads and writes you should look elsewhere.