February 14, 2017 at 10:24 am
After moving the Sql server to new hardware the CPU utilization dropped. However, the disk queue length went up and page life expectancy dropped little bit.Could you explain why this could happen? Please advise?
Thanks
February 14, 2017 at 10:46 am
Admingod - Tuesday, February 14, 2017 10:24 AMAfter moving the Sql server to new hardware the CPU utilization dropped. However, the disk queue length went up and page life expectancy dropped little bit.Could you explain why this could happen? Please advise?Thanks
Without knowing anything about the old and the new platforms, server versions etc. it would be purely guessing, can you elaborate further please?
😎
February 14, 2017 at 10:49 am
The SQL Server version is 2014. Its on same SAN before and After. Moved to new server with same configuration.
February 14, 2017 at 5:07 pm
Admingod - Tuesday, February 14, 2017 10:24 AMAfter moving the Sql server to new hardware the CPU utilization dropped. However, the disk queue length went up and page life expectancy dropped little bit.Could you explain why this could happen? Please advise?Thanks
This is perfectly reasonable. More powerful CPUs can process more data (i.e. need more stuff in RAM to work on) while showing less overall CPU usage. They are also more efficient at operations which adds to this.
I note that the two things you mention (DQL and PLE) are completely useless in modern performance tuning.
Best,
Kevin G. Boles
SQL Server Consultant
SQL MVP 2007-2012
TheSQLGuru on googles mail service
February 17, 2017 at 1:15 pm
TheSQLGuru - Tuesday, February 14, 2017 5:07 PMAdmingod - Tuesday, February 14, 2017 10:24 AMAfter moving the Sql server to new hardware the CPU utilization dropped. However, the disk queue length went up and page life expectancy dropped little bit.Could you explain why this could happen? Please advise?Thanks
This is perfectly reasonable. More powerful CPUs can process more data (i.e. need more stuff in RAM to work on) while showing less overall CPU usage. They are also more efficient at operations which adds to this.
I note that the two things you mention (DQL and PLE) are completely useless in modern performance tuning.
I'm not convinced that they're COMPLETELY useless... just less useful than before. After all, if you had a sudden spike in DQL or sudden drop in PLE, and nothing else in the way of symptoms, you might still be able to come to a useful conclusion. That aside, the DQL in this case could actually be expected to increase if the new server has more CPU capacity, as it will start doing more work per unit time, and the fixed rate of I/O associated with the SAN will not have changed, but more work can be thrown it's way in the same unit of time, and as it isn't going any faster, the work can now build up. It's something to at least keep an eye on, as eventually, you put enough load on it, and it's going to spike itself to the moon and then you've got trouble.
As the original poster has made no mention of SAN fabric speed, it's probably a good idea to mention it. While modern hard drives are capable of 6 GB/sec for I/O, and SANs in most places are just 1 Gbit/sec (and there are still some that aren't multi-pathed), that rather large difference is probably the largest I/O bottleneck around, and even a 10Gbit/sec network is woefully slow compared to a couple of modern hard drives. Just something to keep in mind when analyzing performance.
Steve (aka sgmunson) 🙂 🙂 🙂
Rent Servers for Income (picks and shovels strategy)
February 17, 2017 at 3:19 pm
sgmunson - Friday, February 17, 2017 1:15 PMTheSQLGuru - Tuesday, February 14, 2017 5:07 PMAdmingod - Tuesday, February 14, 2017 10:24 AMAfter moving the Sql server to new hardware the CPU utilization dropped. However, the disk queue length went up and page life expectancy dropped little bit.Could you explain why this could happen? Please advise?Thanks
This is perfectly reasonable. More powerful CPUs can process more data (i.e. need more stuff in RAM to work on) while showing less overall CPU usage. They are also more efficient at operations which adds to this.
I note that the two things you mention (DQL and PLE) are completely useless in modern performance tuning.
I'm not convinced that they're COMPLETELY useless... just less useful than before. After all, if you had a sudden spike in DQL or sudden drop in PLE, and nothing else in the way of symptoms, you might still be able to come to a useful conclusion.
Can you think of any scenario(s) where either of those things happen with nothing else in the way of symptoms?? I sure can't. Sorry, I stick by my statement. There are many other things to be watching that tell you the same thing and/or which give more details about WHY those two events you mentioned happened.
Best,
Kevin G. Boles
SQL Server Consultant
SQL MVP 2007-2012
TheSQLGuru on googles mail service
February 17, 2017 at 10:48 pm
Admingod - Tuesday, February 14, 2017 10:49 AMThe SQL Server version is 2014. Its on same SAN before and After. Moved to new server with same configuration.
Did you remember to turn Trace Flag 1118 on? Did you check to see what other trace flags were turned on on the old system and compare them to the new system? Is the system set to a "Balanced" or "Energy Saving" power profile? Is the NIC of the same quality and same speed as the old one?
--Jeff Moden
Change is inevitable... Change for the better is not.
February 18, 2017 at 11:02 am
Jeff Moden - Friday, February 17, 2017 10:48 PMAdmingod - Tuesday, February 14, 2017 10:49 AMThe SQL Server version is 2014. Its on same SAN before and After. Moved to new server with same configuration.Did you remember to turn Trace Flag 1118 on? Did you check to see what other trace flags were turned on on the old system and compare them to the new system? Is the system set to a "Balanced" or "Energy Saving" power profile? Is the NIC of the same quality and same speed as the old one?
Windows OS / edition, etc. same ? Any attempts to utilize NUMA ? Agree with Jeff Modem especially on BIOS and NIC power management,
February 18, 2017 at 12:41 pm
First, is the workload on the new server the same as what was typical on the old? Also, list and compare the server level settings between the new and old instance. For example, insure that are Min/Max memory are not at default and have been adjusted to reflect what memory capacity is installed on the new server.
SELECT *
FROM sys.configurations
ORDER BY name;
"Do not seek to follow in the footsteps of the wise. Instead, seek what they sought." - Matsuo Basho
Viewing 9 posts - 1 through 9 (of 9 total)
You must be logged in to reply to this topic. Login to reply