|
|
|
date: Thu, 9 Feb 2006 19:10:26 -0800,
group: microsoft.public.exchange.design
back
Re: SMTP Queue Disk Capacity
Another consideration would be look at your timeout value for SMTP and work
backwards from there. For example, if that's 72 hours and you want to be
able to withstand a worst case delivery outage (this totally depends on your
architecture) + growth, you'd want to have enough space to hold 72 hours
worth of SMTP messages + growth + system overhead + buffer space (in case
you're wrong about how much space you need).
Disk is generally cheap and comes in large sizes by default. You'll want to,
for performance reasons, separate that disk I/O as it is so unless this is a
SAN, you'll pretty much be limited to a 72GB (smallest I've seen lately)
drive size. That's larger than your Logs so I'll have to guess that you're
calculation for logs is based on SAN based storage that you can carve up
into smaller space.
Since Bharat has been kind and pointed out that size matters, you'll have to
dig a little more to work out the amount, but I see no reason to make it
smaller than the log drives. During normal operation you won't need it and
it'll look like it is being wasted. When delivery stops and mail backs up,
and then dequeues, you'll wish you had used enough space to keep the server
operational until you could give it the attention it needed.
I'm sure I could wander on about this for hours, but I think you get the
idea. :)
Al
"Bharat Suneja" wrote in message
news:e7ti8EmLGHA.2336@TK2MSFTNGP12.phx.gbl...
> Depends on how large those messages are (avg message size.. 30k? 50?) and
> the max time you'd have those queues backed up in case of a problem - 8
> hours? a day?
>
> Also factor in inbound traffic.
> --
> Bharat Suneja
> MCSE, MCT
> www.zenprise.com
> blog: www.suneja.com/blog
> -----------------------------------------
>
>
> "Doug C." wrote in message
> news:0CBD3BD6-CB3B-49AD-A74A-19182C6119FA@microsoft.com...
>> Quick quetion on disk sizing. Is there a rule of thumb for the disk
>> capacity
>> on the SMTP Queue disk?
>>
>> The environment I am sizing has:
>> 4 Storage Groups 175Gb
>> Logs 67Gb
>> SMTP-?
>>
>> 4000 users pushing 130 messages per day @ 1.0 IOPS
>>
>
>
date: Fri, 10 Feb 2006 17:35:37 -0500
author: Al Mulnick
Re: SMTP Queue Disk Capacity
Great post - that (considering SMTP timeouts) makes sense!
If you're on a SAN, you can grow or shrink the volumes as long as you "have
more storage in the pool... " , but it helps to be able to size it right (or
as close as possible) and always get more than you need.
--
Bharat Suneja
MCSE, MCT
www.zenprise.com
blog: www.suneja.com/blog
-----------------------------------------
"Al Mulnick" wrote in message
news:%23oLQRJpLGHA.1676@TK2MSFTNGP09.phx.gbl...
> Another consideration would be look at your timeout value for SMTP and
> work backwards from there. For example, if that's 72 hours and you want
> to be able to withstand a worst case delivery outage (this totally depends
> on your architecture) + growth, you'd want to have enough space to hold 72
> hours worth of SMTP messages + growth + system overhead + buffer space (in
> case you're wrong about how much space you need).
>
> Disk is generally cheap and comes in large sizes by default. You'll want
> to, for performance reasons, separate that disk I/O as it is so unless
> this is a SAN, you'll pretty much be limited to a 72GB (smallest I've seen
> lately) drive size. That's larger than your Logs so I'll have to guess
> that you're calculation for logs is based on SAN based storage that you
> can carve up into smaller space.
>
> Since Bharat has been kind and pointed out that size matters, you'll have
> to dig a little more to work out the amount, but I see no reason to make
> it smaller than the log drives. During normal operation you won't need it
> and it'll look like it is being wasted. When delivery stops and mail
> backs up, and then dequeues, you'll wish you had used enough space to keep
> the server operational until you could give it the attention it needed.
>
> I'm sure I could wander on about this for hours, but I think you get the
> idea. :)
>
> Al
>
>
>
> "Bharat Suneja" wrote in message
> news:e7ti8EmLGHA.2336@TK2MSFTNGP12.phx.gbl...
>> Depends on how large those messages are (avg message size.. 30k? 50?) and
>> the max time you'd have those queues backed up in case of a problem - 8
>> hours? a day?
>>
>> Also factor in inbound traffic.
>> --
>> Bharat Suneja
>> MCSE, MCT
>> www.zenprise.com
>> blog: www.suneja.com/blog
>> -----------------------------------------
>>
>>
>> "Doug C." wrote in message
>> news:0CBD3BD6-CB3B-49AD-A74A-19182C6119FA@microsoft.com...
>>> Quick quetion on disk sizing. Is there a rule of thumb for the disk
>>> capacity
>>> on the SMTP Queue disk?
>>>
>>> The environment I am sizing has:
>>> 4 Storage Groups 175Gb
>>> Logs 67Gb
>>> SMTP-?
>>>
>>> 4000 users pushing 130 messages per day @ 1.0 IOPS
>>>
>>
>>
>
>
date: Fri, 10 Feb 2006 16:10:11 -0800
author: Bharat Suneja
Re: SMTP Queue Disk Capacity
Oh, I agree about the SAN. That concept can work well for slow growing
areas such as the stores, but SMTP Queues have a tendency to be
underutilized for 99% of the time. It's that 1% that you really need it
that counts and until dynamic allocation of SAN resources is available, I
think it best to go with high-water marks.
In practice, I've also seen where it's far more difficult to acquire an
expensive resource (SAN space) than is thought during design and initial
deployment phase. I agree with you Bharat, that you should get more than you
need out of the gate to save the pain and embarrassment of not having enough
space when you critically need it.
"Bharat Suneja" wrote in message
news:OOtVI%23pLGHA.3100@tk2msftngp13.phx.gbl...
> Great post - that (considering SMTP timeouts) makes sense!
>
> If you're on a SAN, you can grow or shrink the volumes as long as you
> "have more storage in the pool... " , but it helps to be able to size it
> right (or as close as possible) and always get more than you need.
> --
> Bharat Suneja
> MCSE, MCT
> www.zenprise.com
> blog: www.suneja.com/blog
> -----------------------------------------
>
>
> "Al Mulnick" wrote in message
> news:%23oLQRJpLGHA.1676@TK2MSFTNGP09.phx.gbl...
>> Another consideration would be look at your timeout value for SMTP and
>> work backwards from there. For example, if that's 72 hours and you want
>> to be able to withstand a worst case delivery outage (this totally
>> depends on your architecture) + growth, you'd want to have enough space
>> to hold 72 hours worth of SMTP messages + growth + system overhead +
>> buffer space (in case you're wrong about how much space you need).
>>
>> Disk is generally cheap and comes in large sizes by default. You'll want
>> to, for performance reasons, separate that disk I/O as it is so unless
>> this is a SAN, you'll pretty much be limited to a 72GB (smallest I've
>> seen lately) drive size. That's larger than your Logs so I'll have to
>> guess that you're calculation for logs is based on SAN based storage that
>> you can carve up into smaller space.
>>
>> Since Bharat has been kind and pointed out that size matters, you'll have
>> to dig a little more to work out the amount, but I see no reason to make
>> it smaller than the log drives. During normal operation you won't need it
>> and it'll look like it is being wasted. When delivery stops and mail
>> backs up, and then dequeues, you'll wish you had used enough space to
>> keep the server operational until you could give it the attention it
>> needed.
>>
>> I'm sure I could wander on about this for hours, but I think you get the
>> idea. :)
>>
>> Al
>>
>>
>>
>> "Bharat Suneja" wrote in message
>> news:e7ti8EmLGHA.2336@TK2MSFTNGP12.phx.gbl...
>>> Depends on how large those messages are (avg message size.. 30k? 50?)
>>> and the max time you'd have those queues backed up in case of a
>>> problem - 8 hours? a day?
>>>
>>> Also factor in inbound traffic.
>>> --
>>> Bharat Suneja
>>> MCSE, MCT
>>> www.zenprise.com
>>> blog: www.suneja.com/blog
>>> -----------------------------------------
>>>
>>>
>>> "Doug C." wrote in message
>>> news:0CBD3BD6-CB3B-49AD-A74A-19182C6119FA@microsoft.com...
>>>> Quick quetion on disk sizing. Is there a rule of thumb for the disk
>>>> capacity
>>>> on the SMTP Queue disk?
>>>>
>>>> The environment I am sizing has:
>>>> 4 Storage Groups 175Gb
>>>> Logs 67Gb
>>>> SMTP-?
>>>>
>>>> 4000 users pushing 130 messages per day @ 1.0 IOPS
>>>>
>>>
>>>
>>
>>
>
>
date: Fri, 10 Feb 2006 21:53:16 -0500
author: Al Mulnick
Re: SMTP Queue Disk Capacity
Thanks for the response guys. Lots of good info. In the past, I have used a
rule of thumb of using the same size LUN as my log devices as long as there
are at least 4 spindles in the grouping. However, the client got an outside
recommendation of 275GB for an SMTP LUN!!!! That is REALLY large and I have
not found any reason to use something with that much space unless they had a
requirment to support queueing in the event the SMTP service went down for 3
weeks.
Thaks again for the assistance
--
Douglas Collimore
Solution Consultant
New Jersey
"Al Mulnick" wrote:
> Oh, I agree about the SAN. That concept can work well for slow growing
> areas such as the stores, but SMTP Queues have a tendency to be
> underutilized for 99% of the time. It's that 1% that you really need it
> that counts and until dynamic allocation of SAN resources is available, I
> think it best to go with high-water marks.
>
> In practice, I've also seen where it's far more difficult to acquire an
> expensive resource (SAN space) than is thought during design and initial
> deployment phase. I agree with you Bharat, that you should get more than you
> need out of the gate to save the pain and embarrassment of not having enough
> space when you critically need it.
>
> "Bharat Suneja" wrote in message
> news:OOtVI%23pLGHA.3100@tk2msftngp13.phx.gbl...
> > Great post - that (considering SMTP timeouts) makes sense!
> >
> > If you're on a SAN, you can grow or shrink the volumes as long as you
> > "have more storage in the pool... " , but it helps to be able to size it
> > right (or as close as possible) and always get more than you need.
> > --
> > Bharat Suneja
> > MCSE, MCT
> > www.zenprise.com
> > blog: www.suneja.com/blog
> > -----------------------------------------
> >
> >
> > "Al Mulnick" wrote in message
> > news:%23oLQRJpLGHA.1676@TK2MSFTNGP09.phx.gbl...
> >> Another consideration would be look at your timeout value for SMTP and
> >> work backwards from there. For example, if that's 72 hours and you want
> >> to be able to withstand a worst case delivery outage (this totally
> >> depends on your architecture) + growth, you'd want to have enough space
> >> to hold 72 hours worth of SMTP messages + growth + system overhead +
> >> buffer space (in case you're wrong about how much space you need).
> >>
> >> Disk is generally cheap and comes in large sizes by default. You'll want
> >> to, for performance reasons, separate that disk I/O as it is so unless
> >> this is a SAN, you'll pretty much be limited to a 72GB (smallest I've
> >> seen lately) drive size. That's larger than your Logs so I'll have to
> >> guess that you're calculation for logs is based on SAN based storage that
> >> you can carve up into smaller space.
> >>
> >> Since Bharat has been kind and pointed out that size matters, you'll have
> >> to dig a little more to work out the amount, but I see no reason to make
> >> it smaller than the log drives. During normal operation you won't need it
> >> and it'll look like it is being wasted. When delivery stops and mail
> >> backs up, and then dequeues, you'll wish you had used enough space to
> >> keep the server operational until you could give it the attention it
> >> needed.
> >>
> >> I'm sure I could wander on about this for hours, but I think you get the
> >> idea. :)
> >>
> >> Al
> >>
> >>
> >>
> >> "Bharat Suneja" wrote in message
> >> news:e7ti8EmLGHA.2336@TK2MSFTNGP12.phx.gbl...
> >>> Depends on how large those messages are (avg message size.. 30k? 50?)
> >>> and the max time you'd have those queues backed up in case of a
> >>> problem - 8 hours? a day?
> >>>
> >>> Also factor in inbound traffic.
> >>> --
> >>> Bharat Suneja
> >>> MCSE, MCT
> >>> www.zenprise.com
> >>> blog: www.suneja.com/blog
> >>> -----------------------------------------
> >>>
> >>>
> >>> "Doug C." wrote in message
> >>> news:0CBD3BD6-CB3B-49AD-A74A-19182C6119FA@microsoft.com...
> >>>> Quick quetion on disk sizing. Is there a rule of thumb for the disk
> >>>> capacity
> >>>> on the SMTP Queue disk?
> >>>>
> >>>> The environment I am sizing has:
> >>>> 4 Storage Groups 175Gb
> >>>> Logs 67Gb
> >>>> SMTP-?
> >>>>
> >>>> 4000 users pushing 130 messages per day @ 1.0 IOPS
> >>>>
> >>>
> >>>
> >>
> >>
> >
> >
>
>
>
date: Sun, 12 Feb 2006 13:01:26 -0800
author: Douglas Collimore
|
|
