You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@hbase.apache.org by Eric Charles <er...@u-mangate.com> on 2011/04/04 08:49:52 UTC
Why HTableDescriptor DEFAULT_VERSIONS is 3?
Hi,
Is there a particular reason to have chosen 3 for
HTableDescriptor.DEFAULT_VERSIONS?
("not too low, not too big"? - I didn't find discussions about this).
Tks,
- Eric
Re: Why HTableDescriptor DEFAULT_VERSIONS is 3?
Posted by Joe Pallas <pa...@cs.stanford.edu>.
On Apr 4, 2011, at 10:48 AM, tsuna wrote:
> On Mon, Apr 4, 2011 at 10:40 AM, Stack <st...@duboce.net> wrote:
>> Want to make an issue to change it Joe? (As Ryan says, no
>> justification that I remember other than that is how its always been).
>
> Personally I think that 3 is a good reasonable default. Maybe most
> people don't really need 3 versions, but most of the time I'm sure
> they can pay for it, they won't even notice. It can be a life-saver
> after you screw up to be able to get back to older versions... If you
> truly have a "big data" problem (few people really do), then you
> probably will know what you're doing, and you'll tune the number of
> versions appropriately for your needs.
I've created HBASE-3730 so the pros and cons can be discussed in the JIRA.
joe
Re: Why HTableDescriptor DEFAULT_VERSIONS is 3?
Posted by tsuna <ts...@gmail.com>.
On Mon, Apr 4, 2011 at 10:40 AM, Stack <st...@duboce.net> wrote:
> Want to make an issue to change it Joe? (As Ryan says, no
> justification that I remember other than that is how its always been).
Personally I think that 3 is a good reasonable default. Maybe most
people don't really need 3 versions, but most of the time I'm sure
they can pay for it, they won't even notice. It can be a life-saver
after you screw up to be able to get back to older versions... If you
truly have a "big data" problem (few people really do), then you
probably will know what you're doing, and you'll tune the number of
versions appropriately for your needs.
--
Benoit "tsuna" Sigoure
Software Engineer @ www.StumbleUpon.com
Re: Why HTableDescriptor DEFAULT_VERSIONS is 3?
Posted by Stack <st...@duboce.net>.
Want to make an issue to change it Joe? (As Ryan says, no
justification that I remember other than that is how its always been).
St.Ack
On Mon, Apr 4, 2011 at 9:31 AM, Joe Pallas <pa...@cs.stanford.edu> wrote:
>
> On Apr 3, 2011, at 11:52 PM, Ryan Rawson wrote:
>
>> because it always has been? I think the original BT paper probably
>> had the number '3' in there somewhere...
>>
>> But yes, not too big, not too small. There probably isnt a reasonable
>> setting here, I'm guessing 1 isnt quite right either.
>
> Why? Why shouldn't 1 be the default? If you don't have an explicit use case for more than one version, which I would wager many people don't, what is the point of it?
>
> joe
>
> PS Yes, the Google paper has the number 3 in it somewhere: in the web-crawling example. Since it was just one example, it doesn't make a very good basis for a system-wide default.
>
>
Re: Why HTableDescriptor DEFAULT_VERSIONS is 3?
Posted by Joe Pallas <pa...@cs.stanford.edu>.
On Apr 3, 2011, at 11:52 PM, Ryan Rawson wrote:
> because it always has been? I think the original BT paper probably
> had the number '3' in there somewhere...
>
> But yes, not too big, not too small. There probably isnt a reasonable
> setting here, I'm guessing 1 isnt quite right either.
Why? Why shouldn't 1 be the default? If you don't have an explicit use case for more than one version, which I would wager many people don't, what is the point of it?
joe
PS Yes, the Google paper has the number 3 in it somewhere: in the web-crawling example. Since it was just one example, it doesn't make a very good basis for a system-wide default.
Re: Why HTableDescriptor DEFAULT_VERSIONS is 3?
Posted by Eric Charles <er...@u-mangate.com>.
Good to me as "de-facto" standard.
People should simply know that they will have 3 versions by default when
inserting data in hbase.
Tks,
- Eric
On 4/04/2011 08:52, Ryan Rawson wrote:
> because it always has been? I think the original BT paper probably
> had the number '3' in there somewhere...
>
> But yes, not too big, not too small. There probably isnt a reasonable
> setting here, I'm guessing 1 isnt quite right either.
>
> Its one aspect of your data modelling, so people should probably be
> setting this value on larger tables.
>
> -ryan
>
> On Sun, Apr 3, 2011 at 11:49 PM, Eric Charles
> <er...@u-mangate.com> wrote:
>> Hi,
>>
>> Is there a particular reason to have chosen 3 for
>> HTableDescriptor.DEFAULT_VERSIONS?
>> ("not too low, not too big"? - I didn't find discussions about this).
>>
>> Tks,
>> - Eric
>>
Re: Why HTableDescriptor DEFAULT_VERSIONS is 3?
Posted by Ryan Rawson <ry...@gmail.com>.
because it always has been? I think the original BT paper probably
had the number '3' in there somewhere...
But yes, not too big, not too small. There probably isnt a reasonable
setting here, I'm guessing 1 isnt quite right either.
Its one aspect of your data modelling, so people should probably be
setting this value on larger tables.
-ryan
On Sun, Apr 3, 2011 at 11:49 PM, Eric Charles
<er...@u-mangate.com> wrote:
> Hi,
>
> Is there a particular reason to have chosen 3 for
> HTableDescriptor.DEFAULT_VERSIONS?
> ("not too low, not too big"? - I didn't find discussions about this).
>
> Tks,
> - Eric
>