This is what I expected as an answer, that effort just fully destroying
any benefits.
All what I can hear is just better not touch stuff which works and find
improvements different way.
Change (algorithms/caching/...) where I agree, and seems like bottleneck
will be FS I/O or network delays.
J
That's why we move things off-heap.
On Thu, Dec 19, 2013 at 1:57 PM, Roman Vasilyev wrote:
> One of the core problems what I can see is garbage collection, it blocks
> whole server just to clean memory, same stuff with periods of it. On high
> load you constantly have to control does the heap cl
One of the core problems what I can see is garbage collection, it blocks
whole server just to clean memory, same stuff with periods of it. On
high load you constantly have to control does the heap close to the
limit specified on start.
-- Original Message --
From: "Oscar Bonilla" <6f6.
Never will say that Erlang will be faster and memory consumption will be
lower than native code. Rest of it totally agree with Brandon, in a few
words I'm talking about serious parallel project on improving Cassandra
basics.
-- Original Message --
From: "Andy Cobley"
To: "" ; "Roman
I'm not talking to throw away currently working code. Just port it to
C/C++, and have option to run Java based or "native" binary.
-- Original Message --
From: "Brandon Williams"
To: dev@cassandra.apache.org; "Roman Vasilyev"
Sent: 12/19/2013 11:26:21 AM
Subject: Re: C* engine
Let's
You seem to think something like this is akin to a weekend project. I would
recommend you actually read some of the Cassandra source code and better
understand how it is architected.
> On Dec 19, 2013, at 11:30 AM, "Roman Vasilyev" wrote:
>
> I'm not talking to throw away currently working cod
