every day the importance, complexity and quantity of data is growing. Tools and Frameworks to deal with this use-cases also growing. To meet different needs to analyse the data, industry is understanding that monolithic server architecture may not be sufficient.
Systems requires flexibility in scaling. when we talk about scale, potentially we are talking about scale-Up and scale-out.
Scale-Up:
In this we will replace smaller capacity machines ( less CPU and less storage ) with higher capacity Machines.
Pros:
we can experience increase in performance immediately since all cpu, memory and disk ar attached together.
easy to maintain the system.
Cons:
we will experience longer downtime, because we need to setup the new server and migrate the data.
we will end up splitting the jobs to perform them on individual machines when the data grows.
Scale-Out:
In this we will be adding more approximately optimal capacity machines to address the data growth
Pros:
less or no down time in adding a node to the existing cluster of nodes.
we can have cheaper machines to accommodate more data in the database.
good for GET and SET operations since we would not require to handle data indexing
Cons:
since we can not have proper data indexing retrieval will be very slow for complex data joins.
we need orchestration to keep tab of data availability and data locality.
we will not have flexibility to optimize the query performance.
To address above mentioned issues in system scale another idea is Multi-Dimensional Scala
Multi-Dimensional scale address 3 key areas in Data Base
1) Query performance
2) Indexing
3) Storage/Data
In Multi-Dimensional scale we would be able to scale-Up and Scale-out any of the above mentioned components.
as described in the below diagram, each of the above components would be installed on different zone/cluster of machines as a service.
Internals of the How Multi-Dimensional scale is facilitated in apache couchdb and SnowFlake would be discussed later.
0 comments:
Post a Comment