So
what is HTAP? It's Hybrid Transactional/Analytical Processing. It's tipped
(hyped??) to be the future of business applications. The story goes like this-
For
the last 30 years or so, a very well tried-and-tested data approach remained
unquestioned by anyone- "we need separate databases for OLTP workloads and
OLAP workloads". While OLTP workloads needed the relational, normalized
data models, OLAP needed star-schema or the dimensional data model. Enterprises
heavily adopted and deployed separate models for their operational and
analytical systems because it was a known devil. There are at least 2 big
problems with the approach
- Data will have to move from operational data stores to analytical data stores at a time-lag (means your dashboards, KPIs, business analytics reports etc. don't reflect the 'current truth' about your business)
- There will be data duplication since the same crux of the data is stored in multiple places (production DBs, staging, operational data stores, data warehouse, data marts etc. etc.)
This
approach was challenged by a research paper published in 2009 (the research was
sponsored by SAP)
The
research argued that there can be one common database for your OLTP and OLAP
needs provided you have an in-memory, column-store database. With the
advancements in the data center technologies, faster CPUs, better memory
related infra, it's possible to have one
data model for both the needs. SAP HANA is the most popular HTAP database right
now. Other big DB vendors like Oracle and Microsoft are heavily investing on
HTAP since many see potential and SAP HANA has started performing in the
enterprises. Microsoft calls this approach as ‘tabular model’
If
this lives up to the hype, then we will see big changes in both the application
development and business intelligence landscapes. So better keep your ears open
No comments:
Post a Comment