Do you expect the data size to grow beyond 1MB ?
If that is the case, LDT types are the appropriate datastructure for it.
(BTW, pure in-memory namespaces do not have the 1M limit for record)
If you know that the record size is <1MB,
then you can use normal list/map datatypes and do a scan UDF (over a set if needed), and modify the record of a bunch of keys in one go. This functionality is not available for LDTs yet.
You have to write some logic around the list datatype in the UDF.
You can use the prepend & drop function of list and easily achieve what you want.
http://www.aerospike.com/docs/udf/api/list.html
BTW, we have a planned project where we are porting some of the redis libraries.
We have stack functionality in it. If you can wait, may be it will be simpler to use it.