There was a per CF skip list and bloom filter in earlier C* releases(serialized
right before the actual value). Haven't checked lately but I believe this is no
longer the case, or it's configurable and by default it doesn't do it. There is
probably a ticket on JIRA that outlines that rationale.
us to stop looking at data pages if we know we have all
> selected fields for a given row only
>
>
>
>
> On Sun, Sep 7, 2014 at 11:30 PM, Mark Papadakis
> wrote:
>
>> Greetings,
>>
>> This heuristic helps us eliminate unnecessary I/O in certain workloads an
terleaved columns (as opposed to many all exact
columns req. of 1)
This is configurable on a per CF basis (We usually choose 2).
Maybe you could consider such a heuristic for C*, it should probably benefit
your users too.
Apologies if any of this doesn’t make sense in anyway, feel free to ignore:)
Mark Papadakis
@markpapadakis
For what it’s worth, our datastore is written in C++ and very similar to C* in
terms of functionality and semantics; whatever performance delta there is
there, it mostly comes down to reducing I/O and taking advantage of locality of
reference(both in actual I/O, but also in terms of accessing
m
releasing new features is a
higher priority item for the developers, for now. It will happen though.
Mark Papadakis
On Dec 19, 2013, at 9:22 PM, Roman Vasilyev wrote:
> Hello,
>
> Don't want to rise "holy war". Just let me share my crazy thoughts.
> I believe it co
