Hybrid Storage Model

[malargopal]

Hi All,

1)Why vertica is called Hybrid Storage Model.

All because it has Wos and Ros. Is it so ?

2) We have K safe levels in Vertica.

K-level
Number of Nodes Required
0
1+

1
3+

2
5+

K
(K+1)/2

What is K here? Does it mean the levels(0,1,2,3...k).
if subsitute 2 to k then
(2+1)/2 , how is it possible it should have 5+.

Please explain.

Thanks,
Malar

[nnani]

Hello malargopal,

1)Why vertica is called Hybrid Storage Model.

Vertica Storage model has a WOS store and a ROS store,
WOS store is a memory based store, The data resided in a row format and unsorted format, all this data resides in the memory of the Node
ROS stores is Disk based store, the data resided in the columnar format, sorted format. all the data stored on the Disk of a node.
Now lets say you need to insert data into Vertica
You fire a insert query, the data first gets into the WOS and then automatically is pushed into ROS after some time.
that means your data first resided in Vertica in a row format and then automaticall is converted to Columnar format.
How does this happen? this happend due to the Tuple mover operations running between the WOS and ROS.
The Tuple mover operations are of two type, Mergeeout and Moveout, The tuple mover operations run in the background and can be customized according to user requirements.
Same applies to COPY, UPDATE, DELETE.
Vertica stores the data in a row format and also in the Columnar format hence it is called the Hybrid storage model.
I have attached a pictoral representation of Hybrid storage model which can help you understand.

We have K safe levels in Vertica.

K safety level is the safety level that you need for your database, this feature counts when it comes fault tolerance.
This means how much safe your database is to any issues(server down, maintanance)
basically k safety means that your data is copied and you have multiple copies of database in your cluster, so that If your any node goes down, the Vertica cluster is up and running flawlessly.

If you have set your database to K safe level 1 for a cluster of 3 nodes then every projections is copied/mirrored to offset node, so that you have multiple copies of each projection, so supposedly, one of you node goes down, you still have all thedata on that node avialable on other nodes and your database continues to run, but if two nodes go down then your data is missing and database cannot run.

Vertica supoports K level of 0,1,2
For k level to be 1 - 3+ nodes
For k level to be 2 - 5+ nodes

You can read the Concepts guide for more information on this

Hope this helps you.

[malargopal]

Thank you