I think CQL has a problem, it doesn't deal with binary correctly. I have very successfully used C* to store big amounts of small pictures and CQL just cannot handle that efficiently.
I've looked quickly but I didn't see anywhere in the tests on CASSANDRA-3634 biggish column values tested (say in the order of 2-3 digits of KB). I'm willing to bet that with that, the difference is not 10%, because there is the time to serialize/deserialize to string, but also the exploded size of binary represented as hex strings. I think prepared statement are a very good candidate to solve that 'handling binary correctly' problem, but only if we use binary for it. To me that is *very* big argument against Strings. Sure we could add yet another feature (or some hack) to handle binary, but why bother when prepared statement with binary gives us that *and* we get 10% faster writes even for small values. -- Sylvain On Fri, Jan 13, 2012 at 5:01 PM, Jonathan Ellis <jbel...@gmail.com> wrote: > On Fri, Jan 13, 2012 at 9:45 AM, Tyler Hobbs <ty...@datastax.com> wrote: >> On Fri, Jan 13, 2012 at 8:11 AM, Gary Dusbabek <gdusba...@gmail.com> wrote: >> >>> >>> Not all languages <cough>Javascript</cough> make it easy to do binary. >> >> >> PHP also goes in this boat, which leads me to agree with Gary. > > I don't get it. Don't you already have the binary encoding done for > the PHPCassa? > > -- > Jonathan Ellis > Project Chair, Apache Cassandra > co-founder of DataStax, the source for professional Cassandra support > http://www.datastax.com
