Picking up right where we left off yesterday, we continue down through the not-so-highly-upvoted questions from https://pollgab.com/room/brento.

Here’s what we covered:

• 00:00 Introductions
• 00:22 Andrew from Canada: To Null or not to Null? That is the question. On tables with 100M to 1B rows and with several columns that contain data 20% of the time, should we still allow nulls or set them to an empty/zero value instead? If it’s just retuned but not queried, does that matter?
• 02:02 FARAN: Hi Brent, is separating MDF and LDF on different physical drive still needed when we use SSD?
• 04:34 Maciej: Hi Brent, Recently Amazon made available Babelfish for Aurora PostgreSQL. Could it seriously threaten Microsoft business model and force them to ship more improvements to SQL Server engine (e.g. like Parameter-sensitive plan optimization)?
• 06:59 That’s All Folks – already answered
• 07:17 Michael Devor: Brent. My friend found a plan w/missing index recommendation. For testing he created it but the plan does not show it used. Could the engine be using some part of the index without reporting it? Could there be a reason to keep the index? (nothing else would seem to need it)
• 09:10 MancDBA: Hi Brent, what are your thoughts on DBA Job postings that ask for a “Expert” on various topics eg “Expert at AlwaysOn” or “Expert at TSQL”. I have 10 years experience, should I be an “Expert” by now?
• 11:16 Faruk Ami: Hi Brent, i want to create report for top consuming cpu queries on our sqls but some of them are running only Entity FW queries 🙁 how would you approach this? tried using query store but we are running in multi tenancy (schema level) which makes it a bit hard to investigate
• 12:20 Maciej: Hi Brent, What was the highest value for missing index (returned by sp_BlitzIndex) that you have ever seen?
• 13:13 DrDrakken: Hi Brent! Learned a lot from your fundamentals training and started my own experiments on cardinality estimation. What I don’t understand (and my friend neither): Event simple queries with a sort request way to much memory even if returned row estimates are accurate (sql2019)
• 14:45 Steve E: I know it’s maybe too early to say but do you think SQL Server 2022 will finally fix parameter sniffing for good?
• 16:30 MikeO: I have heard you mention the cost of Sql Server, at what point should processing be done within the application instead of SQL Server ?
• 18:30 FrustratedAnalyst: Hi Brent, many thanks for hosting these office hours sessions, I get a lot out of them! As an analyst who wants to move over to data engineering or administration, how would I impress potential employers when I don’t have much working experience in those areas?
• 20:19 Mike: Brent, can you change the default SQL Server port 1433 years after the server has been configured? What steps or links can I read before doing this task? Need to do this for IRS Audit purposes.
• 21:33 Jim: Hi Brent. Thanks for doing these! We have a vendor app with a ~3 TB DB, mostly one table (docs in DB…grrr). The DB is stored on a Netapp LUN (flash storage). Would you have any recommendations here, as to whether it would be good to add a database file?
• 23:53 Todd Chittenden: You always state that there are enough poorly configured SQL Server to keep us all employed. I’m having a hard time convincing bosses, clients, etc. that SQL performance tuning is needed. How do I get this off the ground? Any suggestions?
• 26:22 SayItAintSo: Do I understand from your black Friday sale “The final time I’m selling it!” that you might be retiring??? (: Asking for many friends.
• 28:41 Oleg: Hi Brent, you often mentioned you wish developers do this or that in your trainings, what would be your ultimate checklist or wish list that you would like all database developers to keep in mind when developing new application?
• 30:48 Torben Iisager: Hi. During rollback and database recovery we ask for the status, but the percent count fluctuating up and down, we don’t know why (and don’t know the look for doc on this issue), the recovery/rollback completes despite that, but I would like to know why the number is fluctuating.
• 32:03 LookingAtProcsAllDay: Hi Brent, love your office hours. A friend just upgraded to 2019 at the same time as the hardware. He is seeing reduction in CPU and Duration of some of the most used procs. But query store metric on memory grant has gone up 8x. New hardware has 3x the memory. Is this OK
• 35:04 DataWriter: My friend’s table has 25 indexes on it. dm_db_index_usage_Stats shows many are unused/barely used: 12 indexes are .002% of the total index usage, but 53% of the total user_updates. His manager isn’t a SQL guy. Is there a way to quantify the performance gain from removing them?
• 36:33 Wrap-up