%Disk Time value over 100

Hello!
I have collected counters using perfmon from our Production SQL Server. I
have noticed that %Disk Time counter averaging 111 with maximum value being
2249 for particular drive. I was wondering how I detect whether this value
indicates a bottleneck. Drive in question is RAID 10. Is there conversion
formula I should use?
I know that value over 60 indicates a potential problem. Average Disk Queue
lenght is 1 which means to me that disk is OK. I am confused by high %Disk
Time value.
Any advice is appreciated,
Igor
Used 100 - %disk idle time to get your utilization.
"imarchenko" wrote:

> Hello!
> I have collected counters using perfmon from our Production SQL Server. I
> have noticed that %Disk Time counter averaging 111 with maximum value being
> 2249 for particular drive. I was wondering how I detect whether this value
> indicates a bottleneck. Drive in question is RAID 10. Is there conversion
> formula I should use?
> I know that value over 60 indicates a potential problem. Average Disk Queue
> lenght is 1 which means to me that disk is OK. I am confused by high %Disk
> Time value.
>
> Any advice is appreciated,
> Igor
>
>
|||%Disk time is a useless counter on high-end IO systems. It was designed for
sequential command IO systems, not current SCSI systems that support Command
Tag Queuing or the new SATA Native Command Queuing options. These are the
device-level commands that support scatter-gather IO. Short description is
that they full-duplex command and response for IO. The device can queue a
large number of IO requests, sort them optimally, and respond to them
asychronously and asequentially. RAID subsystems further reduced the
effectiveness of this counter buy abstracting a large number of physical
devices into one logical device presented to the OS..
I use Physical disk | Disk Read Bytes/sec, Disk Write Bytes/sec, Disk
Reads/sec, Disk Writes/sec, and Disk Queue Length to determine if I am IO
bound. Of course, you need to know the maximum capabilities of your IO
subsystem to use these numbers effectively. But since you benchmarked the
IO system with IOMeter when you built the server that is an easy comparison.

Geoff N. Hiten
Senior Database Administrator
Microsoft SQL Server MVP
"imarchenko" <igormarchenko@.hotmail.com> wrote in message
news:eKaIHNaqFHA.4044@.TK2MSFTNGP09.phx.gbl...
> Hello!
> I have collected counters using perfmon from our Production SQL Server.
> I have noticed that %Disk Time counter averaging 111 with maximum value
> being 2249 for particular drive. I was wondering how I detect whether this
> value indicates a bottleneck. Drive in question is RAID 10. Is there
> conversion formula I should use?
> I know that value over 60 indicates a potential problem. Average Disk
> Queue lenght is 1 which means to me that disk is OK. I am confused by high
> %Disk Time value.
>
> Any advice is appreciated,
> Igor
>
|||Jeffrey,
Thanks a lot!
Igor
"Jeffrey K. Ericson" <JeffreyKEricson@.discussions.microsoft.com> wrote in
message news:22B02C35-4D78-47A1-B273-88ED2F5EDACA@.microsoft.com...[vbcol=seagreen]
> Used 100 - %disk idle time to get your utilization.
> "imarchenko" wrote:
|||Thanks, Geoff. I really appreciate your elaborate reply.
Igor
"Geoff N. Hiten" <sqlcraftsman@.gmail.com> wrote in message
news:eSP65saqFHA.272@.TK2MSFTNGP15.phx.gbl...
> %Disk time is a useless counter on high-end IO systems. It was designed
> for sequential command IO systems, not current SCSI systems that support
> Command Tag Queuing or the new SATA Native Command Queuing options. These
> are the device-level commands that support scatter-gather IO. Short
> description is that they full-duplex command and response for IO. The
> device can queue a large number of IO requests, sort them optimally, and
> respond to them asychronously and asequentially. RAID subsystems further
> reduced the effectiveness of this counter buy abstracting a large number
> of physical devices into one logical device presented to the OS..
> I use Physical disk | Disk Read Bytes/sec, Disk Write Bytes/sec, Disk
> Reads/sec, Disk Writes/sec, and Disk Queue Length to determine if I am IO
> bound. Of course, you need to know the maximum capabilities of your IO
> subsystem to use these numbers effectively. But since you benchmarked the
> IO system with IOMeter when you built the server that is an easy
> comparison.
> --
> Geoff N. Hiten
> Senior Database Administrator
> Microsoft SQL Server MVP
>
> "imarchenko" <igormarchenko@.hotmail.com> wrote in message
> news:eKaIHNaqFHA.4044@.TK2MSFTNGP09.phx.gbl...
>

%Disk Time value over 100

Hello!
I have collected counters using perfmon from our Production SQL Server. I
have noticed that %Disk Time counter averaging 111 with maximum value being
2249 for particular drive. I was wondering how I detect whether this value
indicates a bottleneck. Drive in question is RAID 10. Is there conversion
formula I should use?
I know that value over 60 indicates a potential problem. Average Disk Queue
lenght is 1 which means to me that disk is OK. I am confused by high %Disk
Time value.
Any advice is appreciated,
IgorUsed 100 - %disk idle time to get your utilization.
"imarchenko" wrote:

> Hello!
> I have collected counters using perfmon from our Production SQL Server.
I
> have noticed that %Disk Time counter averaging 111 with maximum value bein
g
> 2249 for particular drive. I was wondering how I detect whether this value
> indicates a bottleneck. Drive in question is RAID 10. Is there conversion
> formula I should use?
> I know that value over 60 indicates a potential problem. Average Disk Queu
e
> lenght is 1 which means to me that disk is OK. I am confused by high %Disk
> Time value.
>
> Any advice is appreciated,
> Igor
>
>|||%Disk time is a useless counter on high-end IO systems. It was designed for
sequential command IO systems, not current SCSI systems that support Command
Tag Queuing or the new SATA Native Command Queuing options. These are the
device-level commands that support scatter-gather IO. Short description is
that they full-duplex command and response for IO. The device can queue a
large number of IO requests, sort them optimally, and respond to them
asychronously and asequentially. RAID subsystems further reduced the
effectiveness of this counter buy abstracting a large number of physical
devices into one logical device presented to the OS..
I use Physical disk | Disk Read Bytes/sec, Disk Write Bytes/sec, Disk
Reads/sec, Disk Writes/sec, and Disk Queue Length to determine if I am IO
bound. Of course, you need to know the maximum capabilities of your IO
subsystem to use these numbers effectively. But since you benchmarked the
IO system with IOMeter when you built the server that is an easy comparison.

Geoff N. Hiten
Senior Database Administrator
Microsoft SQL Server MVP
"imarchenko" <igormarchenko@.hotmail.com> wrote in message
news:eKaIHNaqFHA.4044@.TK2MSFTNGP09.phx.gbl...
> Hello!
> I have collected counters using perfmon from our Production SQL Server.
> I have noticed that %Disk Time counter averaging 111 with maximum value
> being 2249 for particular drive. I was wondering how I detect whether this
> value indicates a bottleneck. Drive in question is RAID 10. Is there
> conversion formula I should use?
> I know that value over 60 indicates a potential problem. Average Disk
> Queue lenght is 1 which means to me that disk is OK. I am confused by high
> %Disk Time value.
>
> Any advice is appreciated,
> Igor
>

%Disk Time value over 100

Hello!
I have collected counters using perfmon from our Production SQL Server. I
have noticed that %Disk Time counter averaging 111 with maximum value being
2249 for particular drive. I was wondering how I detect whether this value
indicates a bottleneck. Drive in question is RAID 10. Is there conversion
formula I should use?
I know that value over 60 indicates a potential problem. Average Disk Queue
lenght is 1 which means to me that disk is OK. I am confused by high %Disk
Time value.
Any advice is appreciated,
IgorUsed 100 - %disk idle time to get your utilization.
"imarchenko" wrote:
> Hello!
> I have collected counters using perfmon from our Production SQL Server. I
> have noticed that %Disk Time counter averaging 111 with maximum value being
> 2249 for particular drive. I was wondering how I detect whether this value
> indicates a bottleneck. Drive in question is RAID 10. Is there conversion
> formula I should use?
> I know that value over 60 indicates a potential problem. Average Disk Queue
> lenght is 1 which means to me that disk is OK. I am confused by high %Disk
> Time value.
>
> Any advice is appreciated,
> Igor
>
>|||%Disk time is a useless counter on high-end IO systems. It was designed for
sequential command IO systems, not current SCSI systems that support Command
Tag Queuing or the new SATA Native Command Queuing options. These are the
device-level commands that support scatter-gather IO. Short description is
that they full-duplex command and response for IO. The device can queue a
large number of IO requests, sort them optimally, and respond to them
asychronously and asequentially. RAID subsystems further reduced the
effectiveness of this counter buy abstracting a large number of physical
devices into one logical device presented to the OS..
I use Physical disk | Disk Read Bytes/sec, Disk Write Bytes/sec, Disk
Reads/sec, Disk Writes/sec, and Disk Queue Length to determine if I am IO
bound. Of course, you need to know the maximum capabilities of your IO
subsystem to use these numbers effectively. But since you benchmarked the
IO system with IOMeter when you built the server that is an easy comparison.
:)
--
Geoff N. Hiten
Senior Database Administrator
Microsoft SQL Server MVP
"imarchenko" <igormarchenko@.hotmail.com> wrote in message
news:eKaIHNaqFHA.4044@.TK2MSFTNGP09.phx.gbl...
> Hello!
> I have collected counters using perfmon from our Production SQL Server.
> I have noticed that %Disk Time counter averaging 111 with maximum value
> being 2249 for particular drive. I was wondering how I detect whether this
> value indicates a bottleneck. Drive in question is RAID 10. Is there
> conversion formula I should use?
> I know that value over 60 indicates a potential problem. Average Disk
> Queue lenght is 1 which means to me that disk is OK. I am confused by high
> %Disk Time value.
>
> Any advice is appreciated,
> Igor
>|||Jeffrey,
Thanks a lot!
Igor
"Jeffrey K. Ericson" <JeffreyKEricson@.discussions.microsoft.com> wrote in
message news:22B02C35-4D78-47A1-B273-88ED2F5EDACA@.microsoft.com...
> Used 100 - %disk idle time to get your utilization.
> "imarchenko" wrote:
>> Hello!
>> I have collected counters using perfmon from our Production SQL
>> Server. I
>> have noticed that %Disk Time counter averaging 111 with maximum value
>> being
>> 2249 for particular drive. I was wondering how I detect whether this
>> value
>> indicates a bottleneck. Drive in question is RAID 10. Is there conversion
>> formula I should use?
>> I know that value over 60 indicates a potential problem. Average Disk
>> Queue
>> lenght is 1 which means to me that disk is OK. I am confused by high
>> %Disk
>> Time value.
>>
>> Any advice is appreciated,
>> Igor
>>|||Thanks, Geoff. I really appreciate your elaborate reply.
Igor
"Geoff N. Hiten" <sqlcraftsman@.gmail.com> wrote in message
news:eSP65saqFHA.272@.TK2MSFTNGP15.phx.gbl...
> %Disk time is a useless counter on high-end IO systems. It was designed
> for sequential command IO systems, not current SCSI systems that support
> Command Tag Queuing or the new SATA Native Command Queuing options. These
> are the device-level commands that support scatter-gather IO. Short
> description is that they full-duplex command and response for IO. The
> device can queue a large number of IO requests, sort them optimally, and
> respond to them asychronously and asequentially. RAID subsystems further
> reduced the effectiveness of this counter buy abstracting a large number
> of physical devices into one logical device presented to the OS..
> I use Physical disk | Disk Read Bytes/sec, Disk Write Bytes/sec, Disk
> Reads/sec, Disk Writes/sec, and Disk Queue Length to determine if I am IO
> bound. Of course, you need to know the maximum capabilities of your IO
> subsystem to use these numbers effectively. But since you benchmarked the
> IO system with IOMeter when you built the server that is an easy
> comparison. :)
> --
> Geoff N. Hiten
> Senior Database Administrator
> Microsoft SQL Server MVP
>
> "imarchenko" <igormarchenko@.hotmail.com> wrote in message
> news:eKaIHNaqFHA.4044@.TK2MSFTNGP09.phx.gbl...
>> Hello!
>> I have collected counters using perfmon from our Production SQL Server.
>> I have noticed that %Disk Time counter averaging 111 with maximum value
>> being 2249 for particular drive. I was wondering how I detect whether
>> this value indicates a bottleneck. Drive in question is RAID 10. Is there
>> conversion formula I should use?
>> I know that value over 60 indicates a potential problem. Average Disk
>> Queue lenght is 1 which means to me that disk is OK. I am confused by
>> high %Disk Time value.
>>
>> Any advice is appreciated,
>> Igor
>>
>

% Disk Time (Perfmon)

I have a Windows 2000 Advanced Server with SQL Server 2000
Enterprise Edition on a SAN. What units are the %Disk
Time measured in from Perfmon?
Is the (% Disk Time E) = (Disk Time E) / (Total % Disk
Time) ?
Please help me with these questions.
Thanks,
MikeMike
This counter measured how busy a physical array is. In general it should be
less 55% otherwise you probably jave IO bottleneck.
"Mike" <anonymous@.discussions.microsoft.com> wrote in message
news:c13201c47a1b$4db045c0$a401280a@.phx.gbl...
> I have a Windows 2000 Advanced Server with SQL Server 2000
> Enterprise Edition on a SAN. What units are the %Disk
> Time measured in from Perfmon?
> Is the (% Disk Time E) = (Disk Time E) / (Total % Disk
> Time) ?
> Please help me with these questions.
> Thanks,
> Mike|||I don't tend to agree with that advice.
On it's own
<<
'%Disk Time
Is not a very reliable indicator of an IO bottleneck. You could easily have
a bottleneck if this counter is much lower than 55%. You might NOT have a
bottleneck if this counter is 55% or higher.
There are many, many more counters which you should take a look. You need to
take a look at queue lenght, wait times for disk transfers, time it takes
per read and write. In addition, there are a host of other counters
specific to your SAN that you should take a look at.
Unfortunately, I'm running a bit late right now and I don't have time to
write a long message. Tom Davidson from MS has a nice article in SQL Server
Magazine that dicusses some of these counters. I believe it might also be on
MSDN. It shouldn't be too hard to track down if you search by his name on
each site.
--
Brian Moran
"Uri Dimant" <urid@.iscar.co.il> wrote in message
news:u1rJ0yheEHA.3520@.TK2MSFTNGP10.phx.gbl...
> Mike
> This counter measured how busy a physical array is. In general it should
be
> less 55% otherwise you probably jave IO bottleneck.
>
>
> "Mike" <anonymous@.discussions.microsoft.com> wrote in message
> news:c13201c47a1b$4db045c0$a401280a@.phx.gbl...
> >
> > I have a Windows 2000 Advanced Server with SQL Server 2000
> > Enterprise Edition on a SAN. What units are the %Disk
> > Time measured in from Perfmon?
> >
> > Is the (% Disk Time E) = (Disk Time E) / (Total % Disk
> > Time) ?
> >
> > Please help me with these questions.
> >
> > Thanks,
> >
> > Mike
>|||Brian
First of all i have said 'probably'
Secondly if you have this counter higher than 55% for continuous periods
(let me say 15 min)then your SQL Server
may be experiencing an I/O bottleneck.
"Brian Moran" <brian@.solidqualitylearning.com> wrote in message
news:%23r7TWIieEHA.2848@.TK2MSFTNGP10.phx.gbl...
> I don't tend to agree with that advice.
> On it's own
> <<
> '%Disk Time
> >>
> Is not a very reliable indicator of an IO bottleneck. You could easily
have
> a bottleneck if this counter is much lower than 55%. You might NOT have a
> bottleneck if this counter is 55% or higher.
> There are many, many more counters which you should take a look. You need
to
> take a look at queue lenght, wait times for disk transfers, time it takes
> per read and write. In addition, there are a host of other counters
> specific to your SAN that you should take a look at.
>
> Unfortunately, I'm running a bit late right now and I don't have time to
> write a long message. Tom Davidson from MS has a nice article in SQL
Server
> Magazine that dicusses some of these counters. I believe it might also be
on
> MSDN. It shouldn't be too hard to track down if you search by his name on
> each site.
> --
> Brian Moran
> "Uri Dimant" <urid@.iscar.co.il> wrote in message
> news:u1rJ0yheEHA.3520@.TK2MSFTNGP10.phx.gbl...
> > Mike
> > This counter measured how busy a physical array is. In general it should
> be
> > less 55% otherwise you probably jave IO bottleneck.
> >
> >
> >
> >
> >
> > "Mike" <anonymous@.discussions.microsoft.com> wrote in message
> > news:c13201c47a1b$4db045c0$a401280a@.phx.gbl...
> > >
> > > I have a Windows 2000 Advanced Server with SQL Server 2000
> > > Enterprise Edition on a SAN. What units are the %Disk
> > > Time measured in from Perfmon?
> > >
> > > Is the (% Disk Time E) = (Disk Time E) / (Total % Disk
> > > Time) ?
> > >
> > > Please help me with these questions.
> > >
> > > Thanks,
> > >
> > > Mike
> >
> >
>

% Disk Time (Perfmon)

I have a Windows 2000 Advanced Server with SQL Server 2000
Enterprise Edition on a SAN. What units are the %Disk
Time measured in from Perfmon?
Is the (% Disk Time E) = (Disk Time E) / (Total % Disk
Time) ?
Please help me with these questions.
Thanks,
Mike
Mike
This counter measured how busy a physical array is. In general it should be
less 55% otherwise you probably jave IO bottleneck.
"Mike" <anonymous@.discussions.microsoft.com> wrote in message
news:c13201c47a1b$4db045c0$a401280a@.phx.gbl...
> I have a Windows 2000 Advanced Server with SQL Server 2000
> Enterprise Edition on a SAN. What units are the %Disk
> Time measured in from Perfmon?
> Is the (% Disk Time E) = (Disk Time E) / (Total % Disk
> Time) ?
> Please help me with these questions.
> Thanks,
> Mike
|||I don't tend to agree with that advice.
On it's own
<<
'%Disk Time[vbcol=seagreen]
Is not a very reliable indicator of an IO bottleneck. You could easily have
a bottleneck if this counter is much lower than 55%. You might NOT have a
bottleneck if this counter is 55% or higher.
There are many, many more counters which you should take a look. You need to
take a look at queue lenght, wait times for disk transfers, time it takes
per read and write. In addition, there are a host of other counters
specific to your SAN that you should take a look at.
Unfortunately, I'm running a bit late right now and I don't have time to
write a long message. Tom Davidson from MS has a nice article in SQL Server
Magazine that dicusses some of these counters. I believe it might also be on
MSDN. It shouldn't be too hard to track down if you search by his name on
each site.
Brian Moran
"Uri Dimant" <urid@.iscar.co.il> wrote in message
news:u1rJ0yheEHA.3520@.TK2MSFTNGP10.phx.gbl...
> Mike
> This counter measured how busy a physical array is. In general it should
be
> less 55% otherwise you probably jave IO bottleneck.
>
>
> "Mike" <anonymous@.discussions.microsoft.com> wrote in message
> news:c13201c47a1b$4db045c0$a401280a@.phx.gbl...
>
|||Brian
First of all i have said 'probably'
Secondly if you have this counter higher than 55% for continuous periods
(let me say 15 min)then your SQL Server
may be experiencing an I/O bottleneck.
"Brian Moran" <brian@.solidqualitylearning.com> wrote in message
news:%23r7TWIieEHA.2848@.TK2MSFTNGP10.phx.gbl...
> I don't tend to agree with that advice.
> On it's own
> <<
> '%Disk Time
> Is not a very reliable indicator of an IO bottleneck. You could easily
have
> a bottleneck if this counter is much lower than 55%. You might NOT have a
> bottleneck if this counter is 55% or higher.
> There are many, many more counters which you should take a look. You need
to
> take a look at queue lenght, wait times for disk transfers, time it takes
> per read and write. In addition, there are a host of other counters
> specific to your SAN that you should take a look at.
>
> Unfortunately, I'm running a bit late right now and I don't have time to
> write a long message. Tom Davidson from MS has a nice article in SQL
Server
> Magazine that dicusses some of these counters. I believe it might also be
on
> MSDN. It shouldn't be too hard to track down if you search by his name on
> each site.
> --
> Brian Moran
> "Uri Dimant" <urid@.iscar.co.il> wrote in message
> news:u1rJ0yheEHA.3520@.TK2MSFTNGP10.phx.gbl...
> be
>

% Disk Time (Perfmon)

I have a Windows 2000 Advanced Server with SQL Server 2000
Enterprise Edition on a SAN. What units are the %Disk
Time measured in from Perfmon?
Is the (% Disk Time E) = (Disk Time E) / (Total % Disk
Time) ?
Please help me with these questions.
Thanks,
MikeMike
This counter measured how busy a physical array is. In general it should be
less 55% otherwise you probably jave IO bottleneck.
"Mike" <anonymous@.discussions.microsoft.com> wrote in message
news:c13201c47a1b$4db045c0$a401280a@.phx.gbl...
> I have a Windows 2000 Advanced Server with SQL Server 2000
> Enterprise Edition on a SAN. What units are the %Disk
> Time measured in from Perfmon?
> Is the (% Disk Time E) = (Disk Time E) / (Total % Disk
> Time) ?
> Please help me with these questions.
> Thanks,
> Mike|||I don't tend to agree with that advice.
On it's own
<<
'%Disk Time
Is not a very reliable indicator of an IO bottleneck. You could easily have
a bottleneck if this counter is much lower than 55%. You might NOT have a
bottleneck if this counter is 55% or higher.
There are many, many more counters which you should take a look. You need to
take a look at queue lenght, wait times for disk transfers, time it takes
per read and write. In addition, there are a host of other counters
specific to your SAN that you should take a look at.
Unfortunately, I'm running a bit late right now and I don't have time to
write a long message. Tom Davidson from MS has a nice article in SQL Server
Magazine that dicusses some of these counters. I believe it might also be on
MSDN. It shouldn't be too hard to track down if you search by his name on
each site.
--
Brian Moran
"Uri Dimant" <urid@.iscar.co.il> wrote in message
news:u1rJ0yheEHA.3520@.TK2MSFTNGP10.phx.gbl...[vbcol=seagreen]
> Mike
> This counter measured how busy a physical array is. In general it should
be
> less 55% otherwise you probably jave IO bottleneck.
>
>
> "Mike" <anonymous@.discussions.microsoft.com> wrote in message
> news:c13201c47a1b$4db045c0$a401280a@.phx.gbl...
>|||Brian
First of all i have said 'probably'
Secondly if you have this counter higher than 55% for continuous periods
(let me say 15 min)then your SQL Server
may be experiencing an I/O bottleneck.
"Brian Moran" <brian@.solidqualitylearning.com> wrote in message
news:%23r7TWIieEHA.2848@.TK2MSFTNGP10.phx.gbl...
> I don't tend to agree with that advice.
> On it's own
> <<
> '%Disk Time
> Is not a very reliable indicator of an IO bottleneck. You could easily
have
> a bottleneck if this counter is much lower than 55%. You might NOT have a
> bottleneck if this counter is 55% or higher.
> There are many, many more counters which you should take a look. You need
to
> take a look at queue lenght, wait times for disk transfers, time it takes
> per read and write. In addition, there are a host of other counters
> specific to your SAN that you should take a look at.
>
> Unfortunately, I'm running a bit late right now and I don't have time to
> write a long message. Tom Davidson from MS has a nice article in SQL
Server
> Magazine that dicusses some of these counters. I believe it might also be
on
> MSDN. It shouldn't be too hard to track down if you search by his name on
> each site.
> --
> Brian Moran
> "Uri Dimant" <urid@.iscar.co.il> wrote in message
> news:u1rJ0yheEHA.3520@.TK2MSFTNGP10.phx.gbl...
> be
>