2 ms·
A pretty glaring flaw or omission in the analysis is using a table that's just 6 columns wide. Tables used in data analytics workloads are much wider, 50-100 c
by tedmiston 9y ago
A pretty glaring flaw or omission in the analysis is using a table that's just 6 columns wide. Tables used in data analytics workloads are much wider, 50-100 columns or more is common. That number of columns means scanning significantly more data for the row-oriented storage.
- hyperpape 9y agoDepending on the details, I'd wonder if you'd just end up pulling in a single cache line from the row, regardless of how wide the row is. Even in his benchmark, you're wasting a good 5/6ths of the memory bandwidth you're using (pulling a 24 byte row and only examining 4 bytes of it).