Technically there could be data in an sstable with a later time stamp than what exists in the memtable. Consider the use case of issuing a delete in the future to avoid race conditions. On Sat, Sep 5, 2015 at 10:42 AM Ray Sutton <ray.sut...@gmail.com> wrote:
> This documentation from Datastax may be helpful to understand the purpose > of memtables and sstables. > > http://docs.datastax.com/en/cassandra/2.0/cassandra/dml/dml_write_path_c.html > > Ray > > On Sat, Sep 5, 2015 at 10:36 AM Anuj Wadehra <anujw_2...@yahoo.co.in> > wrote: > >> Memtables are for storing writes in memory till they are flushed to disk >> as sstables and once flushed, space gets released from commit logs too.. If >> your are updating some columns only that data would be there in memtables >> not entire row..Dont think of memtables as row cache.. >> >> This is my understanding..Anyone is free to add/correct info provided... >> >> >> Anuj >> >> Sent from Yahoo Mail on Android >> <https://overview.mail.yahoo.com/mobile/?.src=Android> >> ------------------------------ >> *From*:"Jim Shi" <hj...@yahoo.com> >> *Date*:Sat, 5 Sep, 2015 at 10:11 am >> *Subject*:memtable and sstables >> >> Hi, I have a question: >> >> in a read operation, why data in memtable needs merges with data in >> sstables? >> >> I thought if a key exists in memtable, it has all the data related to the >> key. Is this assumption not correct? >> >
