Closed
Bug 1176649
Opened 10 years ago
Closed 1 year ago
Use an OLTP database for splice "business data"
Categories
(Content Services Graveyard :: Tiles: Administration Front-End, defect)
Content Services Graveyard
Tiles: Administration Front-End
Tracking
(Not tracked)
RESOLVED
INCOMPLETE
People
(Reporter: oyiptong, Unassigned)
References
Details
(Whiteboard: [story])
We are currently
Reporter | ||
Updated•10 years ago
|
Reporter | ||
Comment 1•10 years ago
|
||
We are currently using Redshift, an OLAP database as our primary datastore.
This poses a few problems:
* insertions are slow because of the expectations of transactions in Redshift. We currently have to lock the database to make every tile Test-And-Set operations upon tile ingestion
* insertions are slow because performance in Redshift is optimized for large aggregations and not single insertions and small queries. Every single tile insertion procedure does a join between the tiles and adgroups tables
* the problems will exacerbate with the additions of state-level targeting and other targeting criteria
* we will need to modify rows soon, which is again, not optimized in Redshift
The way to solve this is to use an OLTP database to store this data. The data will be the "business data", i.e. the data splice 1.2 will write.
This bug will allow us to make splice 1.2 a CMS for Tiles with acceptable performance and security.
Postgres is the preferred DB, due to driver compatibility with Redshift.
You need to log in
before you can comment on or make changes to this bug.
Description
•