Deletion Based On Timestamp

[Samarth Gahire <Sa...@talentica.com>]

I have one suggestion regarding Deletion
Why don't We provide deletion on the basis of timestamp.
It will be more useful as well as time efficient if there provision  to delete the data which is inserted in particular range of time on the basis of timestamp.
Example:-

1)      Lets say I will ask Cassandra to delete data which is inserted (whose TimeStamp is less than the given timestamp) before certain date(TimeStamp).

2)      Or another way to delete data inserted between certain time span .like delete the data inserted between two input timestamps.

Thanks and Regards,
Samarth Gahire.

Re: Deletion Based On Timestamp

[Nick Telford <ni...@gmail.com>]

This would require a full table scan (or some kind of secondary index on
timestamp built-in). You can have Cassandra discard columns after a
particular time by writing them with a TTL, but this isn't quite what you're
looking for.

Unlike BigTable/HBase, timestamps in Cassandra aren't really something that
your data model should make use of. They're used primarily for conflict
detection/resolution and compaction/TTL discarding of old data. Given that
this request goes beyond the scope of Cassandra's design, you're better off
taking this to the cassandra-users list.

If you have your heart set on deleting columns by timestamp, your best bet
is to build a Hadoop MapReduce job to do it for you. This will obviously be
quite a heavy-handed approach, so it's something you should not be running
very frequently.

Regards,

Nick Telford

On 28 September 2011 15:21, Samarth Gahire <Sa...@talentica.com>wrote:

> I have one suggestion regarding Deletion
> Why don't We provide deletion on the basis of timestamp.
> It will be more useful as well as time efficient if there provision  to
> delete the data which is inserted in particular range of time on the basis
> of timestamp.
> Example:-
>
> 1)      Lets say I will ask Cassandra to delete data which is inserted
> (whose TimeStamp is less than the given timestamp) before certain
> date(TimeStamp).
>
> 2)      Or another way to delete data inserted between certain time span
> .like delete the data inserted between two input timestamps.
>
> Thanks and Regards,
> Samarth Gahire.
>
>