Database Reference
In-Depth Information
Time series tables
With DynamoDB, when you are creating a table you are going to specify the throughput.
So from that DynamoDB will allocate resources to serve your requirements with low
latency as best as it can. The core concept is that you should identify the core pattern of
your tables that will be accessed via an application and based on that analysis you can re-
structure your application and tables.
Let's say, for example, you are designing a table to map the customers' search patterns on
your application on the Web. So you can design your DynamoDB table with hash and
range type primary keys with consumer ID as the hash attribute and date/time as the range
attribute. So within this application your customer data will grow indefinitely with time;
however, your application will show the uneven access pattern across items given in the
table. In this pattern, it is possible that the latest item will be more frequently accessed and
eventually the older items will be rarely accessed. So if you get this kind of known pattern
and table behavior from the application, you can justify this at the time of designing your
schema. Instead of storing each and every item on the same table, you can divide it into
multiple tables based on access behavior. You can create tables to store daily or weekly
data. On whatever table the data access rate is high, you can request higher throughput on it
and for tables that have unaccessed or older data, you can cut down some throughput and
save on resources.
So, storing frequently-accessed items called
hot
content with great throughput in one table
and rarely accessed items called
cold
items with reduced throughput in another table can
help you to save resources. You can even refuse old data items by simply deleting their re-
spective tables, or you can switch that content to another storage service such as S3 for fu-
ture reference. To delete the whole table will be more efficient than a one-by-one process
as it will double the write throughput as you will perform many put operations.
