- DynamoDB:
- DynamoDB is
created by Amazon
- Is used by Amazon to store
shopping carts of
their customers:
- The DynamoDB can process millions of shopping cart "updates" in a day
- DynamoDB is
created by Amazon
- Key principles taken into
account when
designing
DynamoDB:
- Incremental scalability:
- Dynamo must be able to
scale out (=
increase) by
one (1) storage node
at a time
- When an new node is added to Dynamo, the change must have minimal impcat on the on going operation
- Dynamo must be able to
scale out (=
increase) by
one (1) storage node
at a time
- Symmetry:
- Every node in
Dynamo have the
same set of
responsibilities
I.e.:, Dynamo is a "peer-to-peer" system (not master-slave which is asymmetric)
- Every node in
Dynamo have the
same set of
responsibilities
- Decentralization:
- This is an extension of the
symmetric principle.
The design of Dynamo is decentralized peer-to-peer
- This is an extension of the
symmetric principle.
- Heterogeneity:
- Dynamo must be able to
exploit heterogeneity in
the infrastructure:
- Work distribution must be proportional to the capacilities of each data server (node) !!!
This is essential when a new node is added that has a higher processing capability
- If this more capable node is given the same amount of load, this new node will be under utilized !!!
- Dynamo must be able to
exploit heterogeneity in
the infrastructure:
- Incremental scalability:
- Dynamo is a
distributed file storage system
that provides
high performance through:
- Load distribution through
sharding
and
data replication
!!!
- The data is
sharded using a
"range-based"-like
sharding method:
But the method uses ranges on the Hash(key) value !!!
-
The sharding method is called:
- Consistent hashing
which can:
- Accommodate heterogeneity
- Accomodate data replication
- Load distribution through
sharding
and
data replication
!!!
|
|
We will look at the consistent hashing based sharding method next....
- SOSP paper on Dynamo: click here