« first day (1358 days earlier)      last day (3504 days later) » 

1:05 AM
Glenn did update the post with a new test.
 
1:22 AM
@AaronBertrand Yes, saw that. Much better, but then with a VM that size, would we be using BPE?
Perhaps we would. I don't know.
 
1:58 AM
Nope, really doubt it. BPE is for local machines that are craptacular and don't support adding memory but do support some hobbled-together local SSD.
Behold the developers of your future enterprise applications
0
Q: Install JDBC driver 4.0 for SQL Server

etho201How can I confirm that JDBC is running and properly configured/registered in the system? I followed the installation instructions but I believe there has to be more to it than what the instructions show. It just says to extract sqljdbc__enu.exe (it's an executable zip file) into "C:\Program Files...

Somebody shoot me in the face
 
 
2 hours later…
3:39 AM
@AaronBertrand I admit I don't know much about this, but would BPE not be potentially useful for Azure VMs running SQL 2014 Standard, limited to 128GB Buffer Pool?
 
 
1 hour later…
5:00 AM
hi
Create table DEPT('DEPT NO', Number (2),'DNAME',varchar2(14),'LOC',varchar2(13));
gave ORA-00904: : invalid identifier
 
 
2 hours later…
6:40 AM
@user285oo6 Single quotes are for string literals, double quotes are for quoting identifiers. And there must be no comma between a column identifier and its type in CREATE TABLE.
 
tq
for insert into emp_company values('SHANKAR','TATA',2000,90-JUL-10); it gave an error column not allowed
 
7:10 AM
Fair enough. You could do 90 - 10 (which would yield 80) in VALUES, but you can't do 90 - JUL - 10 there. JUL looks like a column name to Oracle and you can't reference any columns in this context.
You probably meant to specify a date there. If I were you, I would google for something like oracle date literals.
 
 
2 hours later…
8:51 AM
@AaronBertrand Having seen the output from the past and present of enterprise application developers I see nothing much has changed ...
Nothing like working Data Warehousing to see the nasty scabby bits of the enterprise application world. I've pretty much concluded that most enterprise applications are developed by incompetent people.
 
9:05 AM
The "bug" in H2SQL was not a bug after all.
will work just fine - if you use standard SQL syntax for the alias. nlsGuid = CASE is invalid (standard) SQL. If you change that to case .. end as nlsGuid your query works just fine (and will still work in SQL Server). When I do that, your example query and data returns exactly the same result in h2 and SQL Server — a_horse_with_no_name 3 hours ago
The guy was using the + for concatenating and the SQL-Server's syntax alias = ....
(and blamed h2sql for the "bug" in outer joins)
I wonder what are these guys drinking. Another one:
So, one guy, 5 years ago, said that his tables were "collapsing" with 2 million rows and that's it? You did not test? (By the way, I've added -1 to that non-answer unreasonable claim.) — ypercube 48 secs ago
 
@ypercube Tempted to flag that "answer". Breaking the reference is the only thing stopping me.
 
@AndriyM The 5-years old one? I think downvoting is enough.
I wonder how he found that question and answers.
 
9:23 AM
@ypercube It's just not an answer technically. It confirms someone else's idea, adds a tangential point about indexing and asks a question.
 
You are right in that.
 
@AndriyM ya the documnetation says it should be given in a dd-monnth-yyyy format
The ANSI date literal contains no time portion, and must be specified in exactly this format ('YYYY-MM-DD'). Alternatively you can specify an Oracle date value, as in the following example:

TO_DATE('98-DEC-25 17:30','YY-MON-DD HH24:MI')
 
@user285oo6: See the SQLFiddle
Various ways to do this. TO_DATE() also, if you want a different than the default format.
 
in my column i want it as dd-mm-yyyy so better to go for TO_date i guess
 
That's (the output) irrelevant.
 
9:38 AM
super site , then what should be the command for that particular dsired output
 
Dates are stored in an internal format - no matter how you send them to or how you retrive them from the database.
You can store the dates using any of the ways above, TO_DATE, date literals, etc.
For output, you can use the TO_CHAR() function.
 
so first the normal numberic way and then i need the TO_CHAR()
for the output
what are these errors i googled them and it said something about the max time of idle connection
and for the closed connection error i click on re-connect it and shows the same error again
how can i avoid this connection idle
 
 
3 hours later…
12:25 PM
Sums up football pretty well
 
JNK
12:45 PM
morning all
 
1:21 PM
morning
sql fiddle is down
 
@PaulWhite I'm not sure how big you can configure the SSDs on those VMs - do you buy your own and ship it to Microsoft? It may help in cases where your total data+index size is somewhere between 120GB and (spare space on SSD + 120GB). I had high hopes for this feature but it seems more practical to just opt for the edition + VMs with more memory. It's not quite like jumping from a standard boxed product to enterprise.
@PaulWhite this post suggests that you need at least 512GB SSD in that config (4x-10x mem) and that your expectations should be low if your physical memory is above 64GB.
I have a mail in to Evgeny to (1) update the post to get rid of the "Enterprise required" but and (2) address the questions about why > 64GB will yield little gain - I think he's explained a bit of that to us privately but he should address it there too.
 
1:42 PM
I love me from a year ago
3
> In the pantheon of horrible ideas, the following is near the top
 
tataratata ...............
6
A: SSIS Sequential Processing

billinkcDon't have independent data flows in the same task. I know the Import/Export wizard will do that but just because a team at Microsoft does something, doesn't make it a best practice. The Data Flow gets its power and performance through "free" parallelization. If you don't want that, please, for t...

 
Si, and this reminded me of it
-1
Q: SSIS Delay the second process

SQLMikeProcess Flow: We have a Data flow task inside a For each container. Inside the Data flow task we have 3 ole-db source connection. Each source connection executes a stored procedure and their results will be passed to the flat file designation. right now all the 3 ole-db executes at same time. ...

 
@billinkc Close as duplicate?
Please say yes, I haven't used the hammer for a while.
 
If anything, I would VtC as unclear
Plus, I thought the dupe had to have an accepted answer
 
1:58 PM
No, it can be closed. The question is considered duplicate, not the answer.
But I forgot, I got no gold sql-server.
347
A: When did I get close-vote superpowers?

Shog9Yes, you are now a Superhero, able to wield the mighty Mjölnir. The rules are: You can instantly close as a duplicate any question that was originally asked with a tag you have a gold badge for. You can instantly reopen any question closed as a duplicate that was originally asked with a tag ...

 
@ypercube really?
better get cranking
 
I had closed some questions before - they had the sql tag.
@bluefeet I rarely answer in SO now. Like 2-3 questions a month.
 
@ypercube sounds familiar. I'd like to answer more but I don't have time
 
2:19 PM
@bluefeet can you close or send it here and we can merge it? stackoverflow.com/questions/25994684/…
It's a rather awful question but better not be in 2 sites.
 
done - here is the dba version to close as a dup
0
Q: SqlServer: can one table's PK be the FK of two table's?

user3408973I have three tables project, users and proj_user. The PK of proj_user table is the FK of project and users table's pk e.g.(project_id,user_id) My question is, is it possible if the project_id and user_id's data is different like project_id = '12345' and user_id = 'abc987' Please reply kind re...

then @AaronBertrand or @JNK can you merge? ^^
 
Thnx
 
JNK
2:46 PM
done
 
@JNK thx
 
ruh roh
 
At minimum, please include an example using this function. — Kermit 22 secs ago
 
Huh, so the firewall issues I have at work appear to be due to my usage of not-IE for my web browsing
That's umm an interesting and very ahem secure choice for this organization to make
 
3:02 PM
mmhmm
 
Couldn't use an FTP protocol from opera or chrome but the help desk guy said I had permissions to download. "He doesn't mean use IE on the public internet, did he?" DOWNLOADING
 
@Kermit they fixed the answer
 
@bluefeet removed my down vote
 
@Kermit as did I
 
3:14 PM
@bluefeet the system works... it would be nice to be notified if i down voted that they made an edit.. i'm sure @AaronBertrand has proposed this
 
@Kermit if he didn't I know it's been proposed
 
The Heap™ has beautified this answer
2
A: Date in SELECT statement is always the same

Jonathan MThe SYSDATETIME() function is evaluated once in your query, because it is considered a runtime constant. You can try a windowing function such as ROW_NUMBER() INSERT INTO table1 (invoice, detailline, somedata1, somedata2) SELECT ROW_NUMBER...

 
@Kermit And not only. @billinkc spotted an error and removed the unmatched parenthesis.
 
@bluefeet how many edits before it becomes a community answer?
 
Why ask, let's just keep editing
 
3:21 PM
@Kermit a zillion.
 
@billinkc I think if 5 unique users edit it, it becomes a community answer... we need a 5th
Someone add a : after ROW_NUMBER() ... @PaulWhite
 
Please don't use varchar without length and please don't include date/time data as part of a unique key. — Aaron Bertrand 1 min ago
 
@AndriyM yeah.
 
@Kermit not automatic any longer
 
22
A: Can we disable automatic community wiki conversion for answer edits?

Grace NoteWe have disabled all forms of community wiki automatic conversions, not just for answers but for questions as well. The feature was never really about encouraging substantive edits as much as it was about prevent abuse. And in fact it punished people who made not just substantive edits, but lots ...

 
3:22 PM
My question is why the "please don't include date/time data as part of a unique key" part is not a link
 
@Kermit or someone should delimit the objects
 
@PaulWhite way to poop on my party
 
Sorry to barge in with facts and stuff
 
you'd be a terrible politician
3
 
Thank you!
 
3:24 PM
you are most welcome
 
@AndriyM I don't think I have a blog post on that one
 
@AaronBertrand That's what I was afraid of.
 
@AndriyM It can be very useful as part of a clustered index, often in the lead... but to presume that two things can't happen at the same time is very risky.
 
@AaronBertrand There's a question where I'm thinking part of a solution could be a foreign key referencing a PK or UQ that includes a date column. Not a datetime, though, so maybe that's okay.
 
Yes, for a date, probably ok. Probably some isolated cases where it's ok for datetime too (the same customer can't place the same order from the same machine at the same millisecond, for example). But I don't see the purpose in including that in the key there.
@PaulWhite writing on NOLOCK. I could swear I remember Sch-S still being taken out when NOLOCK queries are issued. Now I can't even get Sch-S to show up under read committed.
What kind of foolish thing could my half-awake brain be thinking?
 
3:31 PM
@AaronBertrand It does! How are you checking for Sch-S?
 
sys.dm_tran_locks, sp_lock - they both show shared db-level and that's it
I fear this is going to become a habit - lack of sleep makes me a moron
 
JNK
Yeah it has to do a sch-s lock for pretty much any operation right?
 
@AaronBertrand Well Sch-S won't be held for very long. Were you expecting it to be held to end of transaction or something?
 
@PaulWhite that is probably my mistake - I need to run NOLOCK against a much bigger table
 
JNK
Or do a select in a loop?
 
3:33 PM
@JNK IIRC snapshot isolation doesn't take Sch-S.
Profiler, extended events, TF 1200
 
What?
0
A: MySQL Function only once with group by

Arjen T.Try Date(), doesn't give the right format.

 
JNK
> All queries, including transactions running under row versioning-based isolation levels, acquire Sch-S (schema stability) locks during compilation and execution. Because of this, queries are blocked when a concurrent transaction holds a Sch-M (schema modification) lock on the table.
that's 2008r2 so maybe it's changed since?
 
No more likely my IIRC was faulty :)
 
JNK
It makes sense though since you would not expect to be able to run a query if the structure of the table is changing.
 
3:36 PM
@JNK Yeah I was just thinking SI had a different mechanism to detect schema changes.
 
Well more like you would not expect to allow the schema to change while a query is running. Potato, potato?
Like I said, I'm tired...
 
@JNK Well I was wrong. SI does take Sch-S. (just tested to make sure!)
Shucks.
 
JNK
@AaronBertrand Yep same thing
 
i'm getting my own physical env with 7TB of SSD. Time for BPE testing that's #fasterThanAzureSSD!!
 
JNK
@swasheck that was cracking me up yesterday
 
3:43 PM
@JNK the elderly coitus one was ... something
 
@swasheck wait, what?
 
JNK
Paul's was good too :)
 
@swasheck ah, I see
and I wish I hadn't
 
@JNK paul white?
 
JNK
yeah
 
3:47 PM
@swasheck he is a very knowledgeable SQL Server dude
and he lives in the future
 
i didnt know if randal posted a good one
 
JNK
@BrentO http://i.imgur.com/OdZBu4U.jpg #FasterThanAzureSSDs
 
0
Q: How should long queries relevant to a question be handled?

AaroninusWhere a question concerns query performance, how should the query be included in the question if it is very long (tens or even hundred of lines)? On SO, code posted in a question should be whittled down to just the relevant lines. However, with an SQL query, subdividing it will change the behavi...

 
4:11 PM
his formatting is just weird
1
A: Sql Subquery result

Gordon LinoffYou can do this with an update using join: update a set code = b.code from a join c on c.latlong = a.latlong join b on b.code = c.code;

 
JNK
4:33 PM
yeah that's odd
 
i fixed it
1
A: Sql Subquery result

Gordon LinoffYou can do this with an update using join: update a set code = b.code from a join c on c.latlong = a.latlong join b on b.code = c.code;

 
JNK
4:53 PM
snark edits always end well!
2
 
@Kermit ugh
based on this lovely comment, I'm vtc as too broad
Now I'm trying to create a pice of procedure. — Alienware 8 mins ago
 
i'm going to make an applewine with a friend so i'm sending him a text
"Montrachet ranges between 59 and 89 so it's a perfect basement yeast" and the next suggested word is "infection."


Thanks, Apple.
 
5:14 PM
@AaronBertrand, @PaulWhite: Why SQL-Server allows different syntax inside subqueries vs. simple queries, regarding UNION and ORDER BY?
Example. In simple query:
SELECT * FROM a
UNION ALL
SELECT * FROM b
ORDER BY x;
the ORDER BY is ordering after the UNION is done.
 
JNK
@swasheck another reason I don't use any autocomplete thingies on my phone
just too frustrating
 
@ypercube I'm struggling to relate your question to your example.
 
But in subqueries, you are even allowed to have:
SELECT *
FROM
    (SELECT * FROM a
    ORDER BY y
    UNION ALL                 ---- ???
    SELECT * FROM b
    ORDER BY x
    ) AS c ;
 
JNK
@PaulWhite I think there's a part b
 
@JNK Yeah. Must be using a slow keyboard ;)
 
JNK
5:19 PM
@ypercube that's invalid for me
 
I understand it can be handy for TOP subqueries but why is it allowed only in subquerie?
 
@ypercube Yes it is confusing. I usually try to avoid the counter-intuive first example by writing the query more explicitly. I assume its SQL standard.
 
JNK
I get an error on the ORDER BY in the subquery
Is that valid in 2012 maybe?
 
@JNK I had it working in SQL-Fiddle but it's down now.
 
JNK
> The ORDER BY clause is invalid in views, inline functions, derived tables, subqueries, and common table expressions, unless TOP or FOR XML is also specified.
 
5:21 PM
I had TOP (n) as well, maybe that's why is allowed
 
yo
 
JNK
yeah that would be why
 
@Zane sup?
 
Not much.
 
JNK
it works if I put top on both statements
 
5:22 PM
@JNK OK. But if you remove the SELECT * FROM ( wrapper, it is invalid.
 
JNK
yeah it is
diff exec plans too
easy repro:

SELECT *
FROM
	(SELECT TOP (100) OBJECT_ID FROM sys.objects AS o
	--ORDER BY object_id
	UNION ALL
	SELECT TOP (100) OBJECT_ID FROM sys.indexes AS i
	ORDER BY object_id
	) x


SELECT TOP (100) OBJECT_ID FROM sys.objects AS o
--ORDER BY name
UNION ALL
SELECT TOP (100) OBJECT_ID FROM sys.indexes AS i
ORDER BY object_id
the second one has a merge join
I'm presuming from the ORDER BY on the set?
yep
If I put the order by in the outer select in the top query I get the same plan
 
@ypercube yes, ORDER BY with TOP in a union, subquery etc. only dictates the rows to include. ORDER BY without TOP is not valid in a union except for at the end, because order by can only apply to the whole concatenated set.
 
@Zane ok?
 
@ypercube stop making sql server do your dirty work anyway
 
And the result is that these two queries may give different results:
SELECT * FROM a
UNION ALL
SELECT TOP (1) * FROM b
ORDER BY x ;

SELECT * FROM
(
    SELECT * FROM a
    UNION ALL
    SELECT TOP (1) * FROM b
    ORDER BY x
) c ;
 
5:26 PM
:17815400 yes exactly - since there is no outer ORDER BY, SQL Server only has to use the order by to determine which 100 rows to include, it does not need to sort the output of the subquery. Outside of a subquery, that's not true.
@ypercube give different results, or give the same results but potentially in a different order?
 
Different.
 
Oh yes I see
<---- again, I blame two kids < 2
 
JNK
@AaronBertrand Welcome to my fun!
Don't worry it gets worse
 
but then it gets better
 
JNK
I'm not there yet but I'm getting close to better
4.5 and 2.5 year olds
that does look better
This is what we deal with:
 
5:30 PM
<3
 
@bluefeet yeah. I'm doing alright now. I got incredibly drunk for a few days and feel a bit better.
 
i still deal with that too ... its great
 
JNK
yeah it's not terrible but it can be headache-inducing
 
hahaha. yeah. it's fun while it's fun - then when it's time to do something productive they can't let go of the fun and it turns into a problem
@Zane sorry that happened. glad you're doing better but also hope that you feel free to "chat" in here. processing can be therapeutic.
 
5:45 PM
> I was told by a person in charge of managing our data that they "could not imagine" using SQL Server for this
0
Q: 700 million row mixed text/numeric data - read-only in a single table

MisterDataOur data has some small character fields, and a few floating point and integer columns. About 36 columns in all. It is about 700 million rows. Is it feasible to use MS SQL Server to host this data? It is basically read-mostly, and writes happen once per day in batch. I was told by a person in ch...

 
JNK
wth
This isn't huge at all for a reporting dataset with bulk loads. Assuming your average column is about 6 bytes, you're talking about 140GB total data. I can't imagine someone in charge of managing data who can't imagine that SQL Server would have an issue handling this. — JNK ♦ 29 secs ago
 
VtC as too broad/opinion based or do we link them msdn documentation about what SQL Server supports
 
@swasheck are those ... uniforms?
 
@Kermit yeah
 
ewww
 
5:53 PM
@Kermit I had a uniform in HS, what's wrong with it?
 
i originally felt the same way --- but man they are great (especially from an administrative perspective)
@bluefeet a personal bias that leans against any sort of conformity, perceived or otherwise? :)
 
@bluefeet conformity
 
Had a college gf point out that my color selection was very consistent for clothing. Upon further inspection, it matched my grade school uniform
 
@bluefeet when you're as religious about nonconformity as he is, then you have to conform to his nonconformity in order to be acceptable
 
It makes life a lot easier with uniforms
 
5:55 PM
@Kermit As a counter, it reduces the opportunity for children to be brand whores
 
@billinkc this. and other things.
 
@billinkc As a counter, it increases the likelihood for rebellious behavior.
 
citation needed
3
 
@Kermit ????????????????????????????????
 
@billinkc yup
@Kermit what are you talking about?
 
5:57 PM
@bluefeet jesus
 
Worked out well for him, everyone knows his name
 
ok. i'm leaving. we've fed the troll and it's just going to become 30 minutes of inconsistencies and contradictions.
 
SQL Server will have no problems with this amount of data, assuming that you are not talking about Express, you use adequate hardware, you design the tables and indexes intelligently, etc. You can prototype and benchmark to prove "the person in charge" wrong, or you can just believe them and move on to something else. Would be interested in knowing what they think could handle this data if they don't think SQL Server could. Are they one of these "MySQL is webscale!" sheep? — Aaron Bertrand ♦ 10 secs ago
 
@Kermit In Greece, they abandoned uniforms in primary schools about 35 years ago (high schools 5-10 years before that). 10 years later, the conformity was to dress according to the latest fashion trends.
 
@swasheck :)
 
6:01 PM
@bluefeet @AaronBertrand can someone delete the picture of my kids, please?
goodbye, for now ...
 
6:24 PM
@AaronBertrand There are people who think 'MySQL is webscale'?
I talked to a Hadoop consultant once who told me something like "There's no way SQL Server could handle this data. It's over 40 gigabytes."
 
@James oh I'm sure
 
Design anything poorly enough, you'll need to throw a whole server farm at it just to get an eventually correct answer
 
Yeah that's why I added this
And I closed as "too broad" but it's actually "too narrow" - the only possible answer, really, is that yes, SQL Server can handle this (but you, on the other hand, could design and use it in such a way that it can't - but you could do that with any platform). — Aaron Bertrand ♦ 29 mins ago
 
When I worked in hosting, we'd get some customers who needed 40 web servers and 2 database clusters to handle the same traffic that other customers with similar businesses could manage with 1 web server and 1 db server.
(Due to horrible design or bizarre platform choices like Microsoft Commerce Server)
 
6:48 PM
Y'all want to punt stackoverflow.com/questions/25971803/… over here and then close as duplicate of dba.stackexchange.com/questions/77247/…
Merge, whatever
 
@billinkc I have voted to migrate the question dear chap
 
can you say sql injection?
0
A: How to use variable table name in CTE

alrocYou can't use variables directly as table names; you need to use dynamic SQL to construct the whole query & concatenate the variable with the table name. DECLARE @tblName VARCHAR(30); SET @tblName = '...'; EXEC (';with tbl1_CTE as (select * from ' + @tblName + ') select * from tbl1_CTE');

 
@Kermit You, kind sir are, a beacon for all of humanity
 
@billinkc right'o
 
they didn't get your comment @billinkc
That's what I named my first child! — alroc 30 secs ago
 
7:01 PM
@bluefeet not too chatty?
 
They did, played along with little bobby tables. Also, they updated their answer to call out that it's vulnerable to injection
 
@Kermit I'm hoping they will understand the point of the original comment - then I'll clean up
 
@bluefeet need one more magic vote
0
Q: not able to create SSISDB catlog in Sql server 2014

Radhigetting below error while trying to create catlog in slq server 2014 integration services. any idea what i missed in installation or anywhere else? The catalog backup file 'C:\Program Files\Microsoft SQL Server\120\DTS\Binn\SSISDBBackup.bak' could not be accessed. Make sure the database file exi...

 
@JNK something weird happened here: dba.stackexchange.com/questions/77368/…
The q is marked as duplicate of the other q (that was merged with it)?
 
Any replication people - would you describe the "currency" of the copy as real-time, near-real-time, or something else?
And of course I mean during the time when replication isn't broken
 
7:21 PM
@Kermit cast
I find it funny when people upvote just to offset a DV
+1. I'm puzzled as to why this answer received a downvote. I added an upvote, I don't see anything "wrong" about this answer. — spencer7593 16 mins ago
and comment about it
 
7:36 PM
This guy thinks that rows have names
-1
A: 1 to Many Sql insert

AugusteIn your insert statement, it looks like you forgot to add the name of the row: It should look like: INSERT INTO Bucket VALUES (Thenameyouwanttogivetherow)

 
JNK
@ypercube closing as a dupe is a pre-req for merging
 
@JNK So this is normal ???
 
JNK
yeah
 
One shows that is merged.
The other is closed as duplicate.
 
JNK
oh thats odd
 
7:45 PM
yeah that seems backward
 
Someone had flagged one of them perhaps?
 
8:01 PM
WOW! I'm not sure why you downvote me. The table has Bucket table has only 2 columns, of which the first column is a primary key which is automatically added. I mistakenly type Thenameyouwanttogivetherow, because technically he will add a new row to the table, with the primary key automatically added. — Auguste 9 mins ago
How can you mistype "Thenameyouwanttogivetherow"?
 
I have a hot key to type that
 
Oh, he corrected.
Thenameyouwanttogivetherow -> ThenameyouwanttogivetheField -> Valueofthefield
And then added the obligatory link to w3schools.
At least we know where he learned all that stuff.
 
"learned"
 
@ypercube Language issue?
 
@PaulWhite Could be.
 
8:05 PM
A quick look at the answers there reminded me why I don't often click SO links.
 
But still the answer does not address any issue of the question.
 
Right.
 
@PaulWhite If I were to follow spencer's advise, I should downvote all answers - in compensation of the poor guy. -5 is rather harsh.
 
What is worse providing a w3schools link or wikapedia
 
@ypercube That was the conclusion I was heading toward before sense prevailed and I closed the tab. I don't need no SO in my life.
2
 
8:08 PM
Me too. Need food and maybe beer.
 
So he has some understanding for my down-vote, at least - nothing to do with the language.
 
But if he had the column list, his query would the same as the OP's ...
 
@ypercube the answer serves no purpose in either case.
 
8:31 PM
Can you say "mislead"?
Hi Max, this is on my DBA box that I use for offloading DBCC checks, so I'm using a good backup. The growth settings for these databases are controlled by best practices issued by the 3rd party vendor. I'm not allowed to touch them. They're set to percentages, which I disagree with, but it's out of my control. Management isn't open to changing because we charge clients by space used for hosting. I do monthly maintenance on VLFs but they have been this high before and I've run the restore with no issues. Thanks. — sqldriver 31 mins ago
 
soooooooo ... Buffer Pool Extension ... not sure i understood it how i thought i did
i'm reading that it wouldnt be good for read-heavy (or DW/BI) workloads
 
Right, big scans will constantly force other data back to disk again (assume "big" > memory + BPE size) so it doesn't really matter if they were in physical memory or BPE, you're still paying the cost of moving data from the real disk into either physical or fake memory
 
Is it real or is it BPE
 
I think the important key is understanding that DW means queries will exhaust physical memory.
Whereas OLTP queries should not.
(And perhaps very read-heavy OLTP workloads should be classified as DW for these purposes.)
 
8:49 PM
PS: How was the Othello thing? Did you win?
 
@AaronBertrand i think that makes sense to me now
 
9:58 PM
@MartinSmith thnx.
@MartinSmith As expected. Lost all my games except 3 against kids and 1 against a 74-years old guy.
 
@AaronBertrand near real time - as always I've seen it take something around 2 to 5-6 seconds (not same instance)
 
10:20 PM
@marian thanks, that's what I expected. Do you use the copy for read workloads? Any contention issues?
 
in our env replications are used for reports mostly, so yeah, most of it is read operations
@AaronBertrand didn't experience much contention, but from time to time there are rogue sessions blocking the replication jobs (nothing too serious, though). But in a busy environment I'd believe a separate distributor is needed, as it can get pretty busy and big if the transaction retention is long (by default the Distribution database gets all rows from a transaction as separate rows, so this can get fluffy).
 

« first day (1358 days earlier)      last day (3504 days later) »