These are my sketchnotes from last week’s ’24 Hours of Pass’ on-line event. All the talks were useful and interesting, my notes perhaps are neither….but they do serve to jog my memory.
’24 Hours of Pass’ is an online event. 24 speakers talk about 24 sqlserver related topics over, surprisingly enough 24 hours. It’s very good and the next one, at the time of writing, is in September.
The ’24 Hours of Pass’ points of presence are:
I’m going to put a bit of text around the works of art below, if only so that you can see where one starts and the other ends.
I’d like to emphasize that these are just my impressions. I may have misunderstood or written stuff down wrong or otherwise mucked up. The presenters slides should be available on the 24 Hours of PASS website – these will be much more authoritative….but will have fewer drawings of Terminators
SQL 2014: changes to the optimizer
First up was a great talk about changes to the optimizer in 2014 – specifically around the cardinality estimator. The way I understood it (which could be well wide of the mark – I can’t vouch for having recorded anything accurately….I may have been scanning or tweeting or wondering whether I could draw a penguin at a critical moment) was the the cardinality estimator has changed little or not at all since SQL 7.0.
Next, a talk about some new features in SQL 2014. The main bits I took from this were:
- buffer pool extension – I think I understand this to be a flash write cache
- the use of Azure for either backups or AlwaysOn
- the resource governor being extended to I/O
SQL 2014: Private and Hybrid Cloud Features
I had the vaguest idea of what Polybase was before the talk. My understanding now is that it allows Parallel Data Warehouse (and therefore the user) to talk to Hadoop seamlessly. I was particularly interested in the idea that you could have your Dimension tables in SQL and Fact table in Hadoop.
SQL 2014: Migration to SQL Azure
SQL 2014: Columnstore indexes
Columnstore indexes seem very cool. In principle data is stored on column-by-column (duh!) rather than row-by-row. This means that you can compress a gazillion times more effectively, and as long as people aren’t doing ‘select *’ from every table you won’t need to traverse as much data.