They're faster - often by factors - than row based database if your workload is mainly based on aggregates. That's why they're so effective for datawarehousing.
On the other hand if your workload is update/delete heavy then they can be much slower than row-based databases.
Just imagine that you're updating a dozen rows at a single record. In a row database that's one search, in a column database it's 12 searches.
They're complimentary to each other - that's why you often see say banks using row-based databases for the day-to-day workload (lots of updates) and column databases for analytics (lots of data being aggregated).
On the other hand if your workload is update/delete heavy then they can be much slower than row-based databases.
Just imagine that you're updating a dozen rows at a single record. In a row database that's one search, in a column database it's 12 searches.
They're complimentary to each other - that's why you often see say banks using row-based databases for the day-to-day workload (lots of updates) and column databases for analytics (lots of data being aggregated).