[aggregated reply to multiple posts from the same sender]

On Thu, May 28, 2009 at 6:34 PM, Brian Desmond <br...@briandesmond.com> wrote:
> In prior releases you had a single table for messages, a single table for 
> attachments, and so forth. This leads to A LOT of random
> I/O which is very expensive, especially on slower drives. In Exchange 2010, 
> you have one table _per mailbox_ for folders,
> one per mailbox for message headers, and one per mailbox for bodies and so 
> forth. What you get from this is
> huge gains in large sequential I/O because you have contiguous data.

  I presume the database takes pains to store these several tables for
each mailbox together on disk?  (Otherwise, there would be just as
much random I/O jumping around between table fragments.  Right?)

  It sounds like Exchange 2010 has been optimized for a large number
of mailboxes at the cost of the performance and space savings SIS can
give you.  Which is great if you have a large number of mailboxes.
We've got around 70.  :-(

  The compression of content that MBS mentions sounds interesting, but
what about already highly compressed files, like digital images and
movies?  There's a lot of that floating around in our IS.

> The page size has also been increased to 32KB (4-fold increase) which gets 
> you a substantial reduction in I/O operations.

  That would apply regardless of SIS or not, no?

> If you're running in a scenario where you have high availability (the new DAG 
> [Distributed Availability Group] design),
> the checkpoint depth has been increased substantially which gets a huge 
> reduction in write operations.

  That would apply regardless of SIS or not, no?

> The net result here is that you can buy fewer cheaper disks, put 
> substantially more users on them
> (think potential factor of 3-4), and deliver larger mailbox quotas. That's a 
> win.

  What if we only have one Exchange server?  I'm not looking to
increase the number of users I can put on that one server.  Does this
scale in the other direction, i.e., can I put the same number of users
on a smaller server, or smaller VM slice?  Or will I be told Exchange
expects gobs of RAM and CPU, because after all, hardware is cheap?

On Thu, May 28, 2009 at 6:51 PM, Brian Desmond <br...@briandesmond.com> wrote:
>> If the SATA drives you're buying are high-end, server-only products, I would
>> expect them to cost as much as SCSI/SAS.
>
> They don't cost the same - they're more than the NewEgg variety but they're 
> not the same cost as a typical SCSI spindle.

  I can't find much in the way of real data, but I found one anecdote
which would would seem to corroborate that.  (ST31000340NS vs
ST31000640SS; $80 difference at CDW.)

  But now I have to wonder, why does the SAS drive cost more?  What am
I giving up for that $80?  Is it just that manufacturers can charge
more for SAS because SAS is perceived as better?  Or will the SATA
disk quit a few years before the SAS disk?

  I know, in the distant past (circa 1995), it used to be that SCSI
disks would generally our-perform and out-last IDE disks, because
anyone using SCSI had higher standards.

  I know, on our piddling little 70-user Exchange server, I still
sprung for a couple of 15K SAS drives for Exchange, because I know I/O
can never be fast enough.  Sounds like I'll have to spring for more
drives now.  Or buy cheaper drives and hope they perform as well.

> --> You can have up to 16 copies of a given database (across 16 servers, 2TB 
> DB max size)
>        --> This means you can spend the money on an extra server/replica 
> potentially as opposed to the cheap disks

  That only helps if you're a large multi-server shop.  For us, the
cost of another Windows Server license + another Exchange Server
license + another server + another server support contract = very
significant money.  We're not about to do that just because Microsoft
tells us the emperor's new clothes are so stylish.

On Thu, May 28, 2009 at 6:36 PM, Brian Desmond <br...@briandesmond.com> wrote:
> Sounds like you’re using Exchange as a file server

  This is how users want to use the product.  Saying "the product
isn't designed for that" means the design is not meeting the user's
desires.  If this was an inescapable technological limitation, it
would be one thing, but since this is a design change away from a
satisfactory design, that's obviously not the case.  When you tell
people doing this "don't use Exchange that way", you're really telling
them "Exchange isn't the right product for you".  I know Exchange has
a bazillion other features, but the thing most people want is for it
to send and store messages.  And "messages" means "whatever content
the user wants", not just text/plain.

-- Ben

~ Finally, powerful endpoint security that ISN'T a resource hog! ~
~ <http://www.sunbeltsoftware.com/Business/VIPRE-Enterprise/>  ~

Reply via email to