4 ms·

I'm curious what results we will see when people start benchmarking it running on the same hardware but Windows in one case and Linux in the other. Also, what

I'm curious what results we will see when people start benchmarking it running on the same hardware but Windows in one case and Linux in the other. Also, what would happen (if anything) if the Linux results crushed the Windows ones? Would that be embarrassing to Microsoft? And to take the thought to the next level of paranoia, let's say Microsoft already ran those benchmarks in-house and found the Linux version vastly faster. To avoid embarrassment, would they slow it down to be more in line with the Windows version?

eknkc 10y ago
I'd guess that it would be slower on Linux by design, but not in a malicious way. They must be using various win32 apis and might have needed to shim them on Linux. Maybe they had assumptions on NTFS performance, paging behavior etc. All would contribute to a performance difference.
- taspeotis 10y ago
  > They must be using various win32 apis and might have needed to shim them on Linux Here's a high level overview: http://arstechnica.com/information-technology/2016/12/how-an-old-drawbridge-helped-microsoft-bring-sql-server-to-linux/ http://arstechnica.com/information-technology/2016/12/how-an...
- manigandham 10y ago
  Official blog post from today that shows more details: https://blogs.technet.microsoft.com/dataplatforminsider/2016/12/16/sql-server-on-linux-how-introduction/ https://blogs.technet.microsoft.com/dataplatforminsider/2016...
- NicoJuicy 10y ago
  If I'm not mistaking, SQL is an OS on itselve, so it was designed crossplatform from the start
  - Someone 10y ago
    Even if it was, it likely was tuned for, or even built for, only one OS for years. That will have made an impact. For example, if thread creation is faster on an OS, starting threads for smaller tasks becomes a win; if locking primitives are slower, it may be better to take a full-table lock more often.
smnc 10y ago
I'm not sure if current releases of SQL Server include the infamous DeWitt clause (http://stackoverflow.com/questions/12115397/is-it-against-license-to-publish-oracle-and-sql-server-performance-test http://stackoverflow.com/questions/12115397/is-it-against-li...).
easuter 10y ago
Keep in mind that SQL Server has had a role in driving certain windows features which I'm not sure exist natively in the Linux kernel. IIRC fibers were added to the NT kernel specifically for SQLS.
- simtel20 10y ago
  I have heard this too. However when working with SQL server it was generally recommended (iirc) that fibers not be used except, and were only goof for very specific niches.
userbinator 10y ago
It reminds me of when the Windows version of Firefox was faster running in WINE than the native Linux one... which then begs the question of whether the same could be true of SQL Server. https://appdb.winehq.org/objectManager.php?sClass=application&iId=11312 https://appdb.winehq.org/objectManager.php?sClass=applicatio... ...but it's unfortunately not even installable.

eknkc 10y ago

I'd guess that it would be slower on Linux by design, but not in a malicious way. They must be using various win32 apis and might have needed to shim them on Linux. Maybe they had assumptions on NTFS performance, paging behavior etc. All would contribute to a performance difference.

taspeotis 10y ago
> They must be using various win32 apis and might have needed to shim them on Linux Here's a high level overview: http://arstechnica.com/information-technology/2016/12/how-an-old-drawbridge-helped-microsoft-bring-sql-server-to-linux/ http://arstechnica.com/information-technology/2016/12/how-an...
manigandham 10y ago
Official blog post from today that shows more details: https://blogs.technet.microsoft.com/dataplatforminsider/2016/12/16/sql-server-on-linux-how-introduction/ https://blogs.technet.microsoft.com/dataplatforminsider/2016...
NicoJuicy 10y ago
If I'm not mistaking, SQL is an OS on itselve, so it was designed crossplatform from the start
- Someone 10y ago
  Even if it was, it likely was tuned for, or even built for, only one OS for years. That will have made an impact. For example, if thread creation is faster on an OS, starting threads for smaller tasks becomes a win; if locking primitives are slower, it may be better to take a full-table lock more often.

smnc 10y ago

I'm not sure if current releases of SQL Server include the infamous DeWitt clause (http://stackoverflow.com/questions/12115397/is-it-against-license-to-publish-oracle-and-sql-server-performance-test http://stackoverflow.com/questions/12115397/is-it-against-li...).

easuter 10y ago

Keep in mind that SQL Server has had a role in driving certain windows features which I'm not sure exist natively in the Linux kernel. IIRC fibers were added to the NT kernel specifically for SQLS.

simtel20 10y ago
I have heard this too. However when working with SQL server it was generally recommended (iirc) that fibers not be used except, and were only goof for very specific niches.

userbinator 10y ago

It reminds me of when the Windows version of Firefox was faster running in WINE than the native Linux one... which then begs the question of whether the same could be true of SQL Server. https://appdb.winehq.org/objectManager.php?sClass=application&iId=11312 https://appdb.winehq.org/objectManager.php?sClass=applicatio... ...but it's unfortunately not even installable.