On Tue, May 12, 2009 at 5:47 AM, Sharad Agarwal <[email protected]> wrote:
> ok. I got a chance to try my hand at it. Ran the benchmark on my dev box. 
> Good to see the relative nos:
>
>                        ,   Object create,   Serialization, Deserialization,   
>    Total Time, Serialized Size
> avro-generic            ,      2453.88000,      5335.50000,      4526.00000,  
>    12315.38000,        211
> avro-specific           ,      1024.00000,      2912.50000,     10415.00000,  
>    14351.50000,        211
> protobuf                ,      1196.20000,      8483.00000,      5965.00000,  
>    15644.20000,        217

When you used protobuf, did you use:

option optimize_for = SPEED;

(See "Getting more speed" in
http://code.google.com/apis/protocolbuffers/docs/javatutorial.html .
I think the trade-off is similar to avro-generic and avro-specific.)

I've attached some hastily created graphs for easier visual
comparison; the second of which removes the Hessian row, since it
makes it hard to compare the others.

I'd be curious about a benchmark where the serialized data is larger:
maybe 1-2MB.

-- Philip

Reply via email to