SQL Server Desired Enhancements
The monthly Data Professionals blog party has come and gone. It happens the second Tuesday of every month – or at least is supposed to happen on that day. This month, the formidable Chris Yates (blog | twitter) has invited everybody to a birthday party – of sorts. As with many birthdays, there is always somebody that wishes you a happy belated birthday. For this party, it is my turn to offer up that belated birthday. It just so happens there was some coordination between Chris and myself for this belated birthday.
Read all about the invite from Chris’ blog. If you missed the link, here it is again – right here.
Plastic Surgery – Desired Enhancements
SQL Server 2016 has come with a ton of cool features, bells, whistles and well cool stuff (yes redundant). That aside, what are some of the really cool features that I would love to see in SQL Server? Let’s run through them (And yes, I will be a bit greedy. It is standard operating procedure when asking for gifts, right?).
I need some way of being able to reproduce a production database cleanly and efficiently in a different environment. Sure, I can script everything and develop an elaborate process to ensure I got an exact duplicate of the stats, stat steps, stats histogram, schema, procedures, indexes, etc etc etc. Being able to do all of that cleanly and efficiently is the key. This is a pretty big want from clients and could be extremely useful.
Microsoft has heard the pleas. Introduced with SP2 for SQL Server 2014 there is a new DBCC statement to do exactly that – DBCC CloneDatabase. Check out all the details here.
Gift #1, let’s check that off the list.
Instant File Initialization is fantastic and a huge time saver. Unfortunately this only applies to the data files. We need something like this implemented for the transaction log. Currently the transaction log must be “zeroed” or 0-stamped when new space is allocated. This mechanism can delay transactions and impact performance if there happens to be a required file growth or even when trying to manually grow the file or even restore the database.
Believe it or not, Microsoft has addressed this request as well. Microsoft has changed how the transaction log is stamped for a significant performance improvement. This is a part of SQL Server 2016. Bob Dorr explains it very well in his blog post on the topic. You can read his blog post here.
Wow, two for two. We can check gift #2 off the list.
Availability Groups seems to get bogged down under heavy load. The redo and log send seem to get backed up and can have a significant impact on production operations. We need the log transport to be faster. No, check that. Not just faster it needs to be 2-3x faster.
SQL Server 2016 comes to the rescue again. Amped up on SQL Steroids, Availability Groups has seen a significant improvement in log transport speeds to the secondaries. Some report it as at least twice as fast. The bottleneck has been moved out of the SQL Engine and it has really amped things up from a performance perspective. Here is a supporting article by Jimmy May on the topic – though it doesn’t go deep into the specifics.
Mark another one off the gift registry. Think we can maintain this pace?
Statistics seem to become stale for smaller tables which dramatically affects performance of certain queries. These tables will not see 20% of the rows updated in the leading edge any time before the turn of the year but they would likely change within the six months following the turn of the new year. We need to be able to force these stats to auto-update more regularly without extra intervention.
Fair enough, we already have a trace flag that can help with that (TF 2371). Maybe the environment or management is resistant to having trace flags implemented for something such as this. You never know what the political red tape may dictate.
SQL Server 2016 to the rescue again!!! SQL Server 2016 has this trace flag enabled automatically. You don’t need to do anything extra special. What this means is that those stats on the smaller tables may actually get updated without intervention despite the lack of change to the rows in the table.
That is four for four. Should we take this birthday party to Vegas? Don’t assume I have stacked the deck either! ;0)
I am getting very frustrated with the constant clearing of usage stats every time I rebuild an index. Just because I rebuild the index, it does not mean that I no longer need the usage stats from prior to the index rebuild. I need to be able to see the usage stats for a the time spanning before and after the rebuild without creating a custom process to capture that information. Sure it may not be an insanely difficult task to perform, but it is extra process I have to build out. It’s the principle of the matter.
SQL Server 2016 to the rescue yet again. This age old bug of usage stats being cleared is finally fixed. It is frustrating to say the least to have to deal with this kind of bug. It is a huge relief to have it fixed and be able to get a consistent clear picture of the usage information since the server has been up.
For more information, you could read this article by Kendra Little – here.
Cha-ching. We are now five for five.
Digging a little deeper on this one. I would really love to see an enhancement to Resource Governor. Not just any enhancement will do. I need it to be enhanced so it will also affect the reporting services engine and the integration services engine in addition to the database engine. I want to be able to use RG to prevent certain reports from over consuming resources within the SSRS engine. Or for that matter, I want to make sure certain SSIS packages do not consume too much memory. If I can implement constraints on resources for these two engines it would be a huge improvement.
We will have to wait for a while on this one. It is currently not scheduled for delivery
This one is going to be a little tougher. It’s not in place. It would be a fantastic gift in my opinion. I would like some tool such as Extended Events to be able to monitor the workload and determine best recommended trace flags to implement.There are many trace flags that are far from well known but could be extremely helpful to production environments based on the workload and internal workflow. Not all would trace flags are built for all environments. An analysis through some automated tool for best recommended flags to implement (again solely at your discretion) would be fantastic.
Get Profiler out of Management Studio finally. Enough said there. There really is no good solid reason in my opinion to keep it around. It is deprecated. It is hardly helpful with 2014 or 2016 and it is just dead weight. Extended Events really is the better way to go here.
Can we please fix the spelling of JSON? It really needs to be spelled correctly. That spelling is: JASON.