Spark Performance Tuning - Part 3

Today's guest on the Data Exposed show is Pedro Lopes, a PM in the SQL Server TIGER team, and he's in the studio today to show off some of the things his team has been working on over the last year to help improve the tooling around SQL Server performance troubleshooting built right into SSMS.
[03:00] - Microsoft SQL Server Performance Dashboard
[06:00] - Activity Monitor Active Expensive Queries and Show Live Execution Plan
[09:20] - Show Query Text
[11:10] - Compare Showplan and Showplan Analysis
[22:20] - New Query Reports, including Queries with Forced Plans and Queries with High Variation.
[23:13] - Plan Comparison - the ability to compare the plans for a selected query.
https://blogs.msdn.microsoft.com/sql_server_team/tag/comparison-tool/
Be sure to follow the Data Exposed show on Twitter at @DataExposed!
Excellent
I want to this demo scenario for learning this features by myself, and sharing this features with friends
Regards,
Yoshihiro Kawabata
This great stuff by Tiger Team..I will try to use this as much as Possible and spread the word.
the QP is a descriptive result oriented language, what we want the result to look like you aim for.
we need ways to tell the QP how to effectively give us the best plan when it guesses wrong. Other
databases do a better job at a higher cost BUT they do not offer SELECTIVE VALUE with a general lower overall cost.
this is to help that and parameter sniffing: have the Query Processor take in table that references columns in other tables. (you can turn into a bit map). The fist table assigns main"big" table number 350000000 with special "sys.user.xxxname" name related to other normal tables with the same numbers in system tables
each table themselves by a company's design has columns left to right 0..n mapped to sys.columns internally, and tables used in the program that start with a number in system tables .
now one col may be the table number 350000000 + the table referenced the next may be region for example the country another may be illness, another may be age, another may be comments, medication, and user, each are assigned and the main columns are up to 3 columns
now the user of the code for the table results is interested in finding money, and life death
relationships fast and they may change. and the speed may depend on how many people are in what region or how many illness.. in a region and it differs across the nation
now the table 350000000 forward tells sql server which columns by order of importance to use statistics and distribution for the query to hurry the plan. this is like always encryption and
disclassification column vi ssms one select and the other acts on it.
the table info would be given with a keyword at the start of the program using a "using" keyword
followed by a created table (name = table number) and "insert into" statement with the info above
sql server when it finds the keyword "using ref" to that same table number before a query
would make its plan according to the selectivity of the columns and the density of the selected columns.
that would take care of the parameter sniffing in most cases for each user for medical, finance, government customers (please give me a job or reward if it works --- terry quillin unemployed DBA) it should give blazing speed to the queries and you lots of money and customers !
it would take a little longer for the developer but for the user big results it could win wars, make lots of money, save lives and hopefully cost little to implement..
in the comments they could enter in details of maybe most important numbers or something that you can charge more for like being in the top 1, or encryption types to change to behind the scene in enclave or different algorithms to use in an enclave or some complicated thing to set up that takes a lot of time that you could have set up and ready to run much faster than an agreement or negotiation or only when needed if needed badly