zfs discuss - Dec 2008 - Hardware Raid Vs ZFS implementation on Sun X4150/X4450

Hi,

 

Has anyone implemented the Hardware RAID 1/5 on Sun X4150/X4450 class of
servers . 

Also any comparison between ZFS Vs H/W Raid ?

 

I would like to know the experience (good/bad) and the pros/cons?

 

Regards,

Vikash

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20081202/9a21bb33/attachment.html>

I''ve done some basic testing with a X4150 machine using 6 disks in a
RAID 5
and RAID Z configuration.  They perform very similarly, but RAIDZ definitely
has more system overhead.  In many cases this won''t be a big deal, but
if
you need as many CPU cycles as you can muster, hardware RAID may be your
better choice.

-Aaron

On Tue, Dec 2, 2008 at 4:22 AM, Vikash Gupta <vikashg at cadence.com>
wrote:
>  Hi,
>
>
>
> Has anyone implemented the Hardware RAID 1/5 on Sun X4150/X4450 class of
> servers .
>
> Also any comparison between ZFS Vs H/W Raid ?
>
>
>
> I would like to know the experience (good/bad) and the pros/cons?
>
>
>
> Regards,
>
> Vikash
>
>
>
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20081203/61699eec/attachment.html>

Aaron Blew <aaronblew <at> gmail.com>
writes:
> 
> I''ve done some basic testing with a X4150 machine using 6 disks in
a
> RAID 5 and RAID Z configuration.? They perform very similarly, but RAIDZ
> definitely has more system overhead.
Since hardware RAID 5 implementations usually do not checksum data (they only 
compute the parity, which is not the same thing), for an apples-to-apples 
performance comparison you should have benchmarked raidz with checksum=off. Is 
it what you did ?

-marc

Keep in mind that if you use ZFS you get a lot of additional functionality like
snapshots, compression, clones.
-- 
This message posted from opensolaris.org

Thanks, but compared to what?
To Windows, are you sure we can say "lot of additional"?
To Linux, maybe, since I am not a Linux fan.
To leading NAS appliances, these are not competitive advantages.
"opensolaris.org" posted this, I would like an official answer!

The Open-spirit should be encouraged, but the wrong marketing positioning 
messages are not!!!
Please, don''t bring shame to the open community.

Thank you!
zStorageAnalyst

----- Original Message ----- 
From: "William D. Hathaway" <william.hathaway at versatile.com>
To: <zfs-discuss at opensolaris.org>
Sent: Thursday, December 04, 2008 7:35 AM
Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on Sun 
X4150/X4450

> Keep in mind that if you use ZFS you get a lot of additional functionality 
> like snapshots, compression, clones.
> -- 
> This message posted from opensolaris.org
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss

I don''t understand your statement/questions.  This wasn''t a
response to "ZFS versus every possible storage platform in the world".
The original poster was asking about comparing  ZFS versus hardware RAID on
specific machines as mentioned in the title.  AFAIK you don''t get
compression, snapshots and clones with standard hardware RAID cards.
-- 
This message posted from opensolaris.org

Yeah?
http://www.adaptec.com/en-US/products/Controllers/Hardware/sas/value/SAS-31605/_details/Series3_FAQs.htm
Snapshot is a big deal?

Windows OS does that too.

Compression -- where is the performance data showing compression in 
OpenSolaris has little overhead?

Clones -- tell me the benefit of Clone when we have point-in-time copies 
with continuous, policy-based protection?  And snapshot images are mostly 
writable and sync-able today?

Man, I am an open storage analyst, please, tell me I am wrong!!!!!
zStorageAnalyst

----- Original Message ----- 
From: "William D. Hathaway" <william.hathaway at versatile.com>
To: <zfs-discuss at opensolaris.org>
Sent: Saturday, December 06, 2008 11:41 PM
Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on Sun 
X4150/X4450

>I don''t understand your statement/questions.  This wasn''t
a response to
>"ZFS versus every possible storage platform in the world".  The
original
>poster was asking about comparing  ZFS versus hardware RAID on specific 
>machines as mentioned in the title.  AFAIK you don''t get
compression,
>snapshots and clones with standard hardware RAID cards.
> -- 
> This message posted from opensolaris.org
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss

Is Jeff Cheeney still on this list? He had an open mind.

Jeff, if you can see this, tell me if I am wrong! Please!

Thanks!
z

----- Original Message ----- 
From: "Joseph Zhou" <jz at excelsioritsolutions.com>
To: "William D. Hathaway" <william.hathaway at versatile.com>; 
<zfs-discuss at opensolaris.org>
Cc: <chriso at ideasinternational.com>; <peterc at
ideasinternational.com>
Sent: Saturday, December 06, 2008 11:51 PM
Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on 
SunX4150/X4450

> Yeah?
>
http://www.adaptec.com/en-US/products/Controllers/Hardware/sas/value/SAS-31605/_details/Series3_FAQs.htm
> Snapshot is a big deal?
>
> Windows OS does that too.
>
> Compression -- where is the performance data showing compression in
> OpenSolaris has little overhead?
>
> Clones -- tell me the benefit of Clone when we have point-in-time copies
> with continuous, policy-based protection?  And snapshot images are mostly
> writable and sync-able today?
>
> Man, I am an open storage analyst, please, tell me I am wrong!!!!!
> zStorageAnalyst
>
> ----- Original Message ----- 
> From: "William D. Hathaway" <william.hathaway at
versatile.com>
> To: <zfs-discuss at opensolaris.org>
> Sent: Saturday, December 06, 2008 11:41 PM
> Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on Sun
> X4150/X4450
>
>
>>I don''t understand your statement/questions.  This
wasn''t a response to
>>"ZFS versus every possible storage platform in the world". 
The original
>>poster was asking about comparing  ZFS versus hardware RAID on specific
>>machines as mentioned in the title.  AFAIK you don''t get
compression,
>>snapshots and clones with standard hardware RAID cards.
>> -- 
>> This message posted from opensolaris.org
>> _______________________________________________
>> zfs-discuss mailing list
>> zfs-discuss at opensolaris.org
>> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
>
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss

Joseph Zhou wrote:
> Yeah?
>
http://www.adaptec.com/en-US/products/Controllers/Hardware/sas/value/SAS-31605/_details/Series3_FAQs.htm
> Snapshot is a big deal?
>   
Snapshot is a big deal, but you will find most "hardware" RAID 
implementations
are somewhat limited, as the above adaptec only supports 4 snapshots and 
it is an
optional feature.  You will find many array vendors will be happy to 
charge lots
of money for the snapshot feature.
> Windows OS does that too.
>   
Not the Windows OS I run on my laptop.  But the feature seems to be best 
integrated
on Max OSX.
> Compression -- where is the performance data showing compression in 
> OpenSolaris has little overhead?
>   
If you search these archives you will find instances where compression
performance is much faster than not, and you will find instances where
compression has significant overhead.  YMMV.  As with most things,
there are engineering and design trade-offs that you should consider.
> Clones -- tell me the benefit of Clone when we have point-in-time copies 
> with continuous, policy-based protection?  And snapshot images are mostly 
> writable and sync-able today?
>   
In ZFS, snapshots are read-only.  Clones are created from a snapshot
and can be writable.  We use clones extensively for OS upgrading and
patching.  For example, when you upgrade OpenSolaris, we clone the
OS file systems and upgrade the clone, so that you can move forward
or roll back to different versions.  Many people use clones for virtual
machines.
> Man, I am an open storage analyst, please, tell me I am wrong!!!!!
>   
I suggest you read the docs, particularly the ZFS Administration Guide.
http://www.opensolaris.org/os/community/zfs/docs
 -- richard
> zStorageAnalyst
>
> ----- Original Message ----- 
> From: "William D. Hathaway" <william.hathaway at
versatile.com>
> To: <zfs-discuss at opensolaris.org>
> Sent: Saturday, December 06, 2008 11:41 PM
> Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on Sun 
> X4150/X4450
>
>
>   
>> I don''t understand your statement/questions.  This
wasn''t a response to
>> "ZFS versus every possible storage platform in the world". 
The original
>> poster was asking about comparing  ZFS versus hardware RAID on specific
>> machines as mentioned in the title.  AFAIK you don''t get
compression,
>> snapshots and clones with standard hardware RAID cards.
>> -- 
>> This message posted from opensolaris.org
>> _______________________________________________
>> zfs-discuss mailing list
>> zfs-discuss at opensolaris.org
>> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss 
>>     
>
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
>

Richard, thank you so very much!
This is the kind of answer I expected from Sun Storage.
I will do more studies before I speak again.
Happy holidays!!!
z

----- Original Message ----- 
From: "Richard Elling" <Richard.Elling at Sun.COM>
To: "Joseph Zhou" <jz at excelsioritsolutions.com>
Cc: "William D. Hathaway" <william.hathaway at versatile.com>; 
<zfs-discuss at opensolaris.org>; <chriso at
ideasinternational.com>;
<peterc at ideasinternational.com>
Sent: Sunday, December 07, 2008 12:57 AM
Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on 
SunX4150/X4450

> Joseph Zhou wrote:
>> Yeah?
>>
http://www.adaptec.com/en-US/products/Controllers/Hardware/sas/value/SAS-31605/_details/Series3_FAQs.htm
>> Snapshot is a big deal?
>>
>
> Snapshot is a big deal, but you will find most "hardware" RAID 
> implementations
> are somewhat limited, as the above adaptec only supports 4 snapshots and 
> it is an
> optional feature.  You will find many array vendors will be happy to 
> charge lots
> of money for the snapshot feature.
>
>> Windows OS does that too.
>>
>
> Not the Windows OS I run on my laptop.  But the feature seems to be best 
> integrated
> on Max OSX.
>
>> Compression -- where is the performance data showing compression in 
>> OpenSolaris has little overhead?
>>
>
> If you search these archives you will find instances where compression
> performance is much faster than not, and you will find instances where
> compression has significant overhead.  YMMV.  As with most things,
> there are engineering and design trade-offs that you should consider.
>
>> Clones -- tell me the benefit of Clone when we have point-in-time
copies
>> with continuous, policy-based protection?  And snapshot images are
mostly
>> writable and sync-able today?
>>
>
> In ZFS, snapshots are read-only.  Clones are created from a snapshot
> and can be writable.  We use clones extensively for OS upgrading and
> patching.  For example, when you upgrade OpenSolaris, we clone the
> OS file systems and upgrade the clone, so that you can move forward
> or roll back to different versions.  Many people use clones for virtual
> machines.
>
>> Man, I am an open storage analyst, please, tell me I am wrong!!!!!
>>
>
> I suggest you read the docs, particularly the ZFS Administration Guide.
> http://www.opensolaris.org/os/community/zfs/docs
> -- richard
>> zStorageAnalyst
>>
>> ----- Original Message ----- 
>> From: "William D. Hathaway" <william.hathaway at
versatile.com>
>> To: <zfs-discuss at opensolaris.org>
>> Sent: Saturday, December 06, 2008 11:41 PM
>> Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on Sun 
>> X4150/X4450
>>
>>
>>
>>> I don''t understand your statement/questions.  This
wasn''t a response to
>>> "ZFS versus every possible storage platform in the
world".  The original
>>> poster was asking about comparing  ZFS versus hardware RAID on
specific
>>> machines as mentioned in the title.  AFAIK you don''t get
compression,
>>> snapshots and clones with standard hardware RAID cards.
>>> -- 
>>> This message posted from opensolaris.org
>>> _______________________________________________
>>> zfs-discuss mailing list
>>> zfs-discuss at opensolaris.org
>>> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
>>
>> _______________________________________________
>> zfs-discuss mailing list
>> zfs-discuss at opensolaris.org
>> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
>>
>

Richard Elling wrote:
> Joseph Zhou wrote:
>   
>> Yeah?
>>
http://www.adaptec.com/en-US/products/Controllers/Hardware/sas/value/SAS-31605/_details/Series3_FAQs.htm
>> Snapshot is a big deal?
>>   
>>     
>
> Snapshot is a big deal, but you will find most "hardware" RAID 
> implementations
> are somewhat limited, as the above adaptec only supports 4 snapshots and 
> it is an
> optional feature.  You will find many array vendors will be happy to 
> charge lots
> of money for the snapshot feature.
On top of that since the ZFS snapshot is at the file system level it''s 
much easier to use. You don''t have to quiesce the file system first or 
hope that when you take the snapshot you get a consistent data set.
I''ve
seen plenty of folks take hw raid snapshots without locking the file 
system first, let alone quiescing the app, and getting garbage.

Torrey, now this impressive as the old days with Sun Storage.

Ok, ZFS PiT is only a software solution.
The Windows VSS is not only a software solution, but also a 3rd party 
integration standard from MS.
What''s your comment on ZFS PiT is better than MS PiT, in light of
openness
and 3rd-party integration???

Talking about garbage!
z


----- Original Message ----- 
From: "Torrey McMahon" <tmcmahon2 at yahoo.com>
To: "Richard Elling" <Richard.Elling at Sun.COM>
Cc: "Joseph Zhou" <jz at excelsioritsolutions.com>;
"William D. Hathaway"
<william.hathaway at versatile.com>; <chriso at
ideasinternational.com>;
<zfs-discuss at opensolaris.org>; <peterc at ideasinternational.com>
Sent: Sunday, December 07, 2008 1:58 AM
Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on Sun 
X4150/X4450

> Richard Elling wrote:
>> Joseph Zhou wrote:
>>
>>> Yeah?
>>>
http://www.adaptec.com/en-US/products/Controllers/Hardware/sas/value/SAS-31605/_details/Series3_FAQs.htm
>>> Snapshot is a big deal?
>>>
>>
>> Snapshot is a big deal, but you will find most "hardware"
RAID
>> implementations
>> are somewhat limited, as the above adaptec only supports 4 snapshots
and
>> it is an
>> optional feature.  You will find many array vendors will be happy to 
>> charge lots
>> of money for the snapshot feature.
>
> On top of that since the ZFS snapshot is at the file system level
it''s
> much easier to use. You don''t have to quiesce the file system
first or
> hope that when you take the snapshot you get a consistent data set.
I''ve
> seen plenty of folks take hw raid snapshots without locking the file 
> system first, let alone quiescing the app, and getting garbage.

Ok, I am tired and going to bed.
Thanks, Real Sun Storage folks, this is the best discussion I have had in 
months.

I am satisfied.   ;-)

Goodnight, and long live the open spirit!!!

zStorageAnalyst


----- Original Message ----- 
From: "Joseph Zhou" <jz at excelsioritsolutions.com>
To: "Torrey McMahon" <tmcmahon2 at yahoo.com>; "Richard
Elling"
<Richard.Elling at Sun.COM>
Cc: "William D. Hathaway" <william.hathaway at versatile.com>; 
<chriso at ideasinternational.com>; <zfs-discuss at
opensolaris.org>;
<peterc at ideasinternational.com>
Sent: Sunday, December 07, 2008 2:03 AM
Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementationon Sun 
X4150/X4450

> Torrey, now this impressive as the old days with Sun Storage.
>
> Ok, ZFS PiT is only a software solution.
> The Windows VSS is not only a software solution, but also a 3rd party
> integration standard from MS.
> What''s your comment on ZFS PiT is better than MS PiT, in light of
openness
> and 3rd-party integration???
>
> Talking about garbage!
> z
>
>
> ----- Original Message ----- 
> From: "Torrey McMahon" <tmcmahon2 at yahoo.com>
> To: "Richard Elling" <Richard.Elling at Sun.COM>
> Cc: "Joseph Zhou" <jz at excelsioritsolutions.com>;
"William D. Hathaway"
> <william.hathaway at versatile.com>; <chriso at
ideasinternational.com>;
> <zfs-discuss at opensolaris.org>; <peterc at
ideasinternational.com>
> Sent: Sunday, December 07, 2008 1:58 AM
> Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on Sun
> X4150/X4450
>
>
>> Richard Elling wrote:
>>> Joseph Zhou wrote:
>>>
>>>> Yeah?
>>>>
http://www.adaptec.com/en-US/products/Controllers/Hardware/sas/value/SAS-31605/_details/Series3_FAQs.htm
>>>> Snapshot is a big deal?
>>>>
>>>
>>> Snapshot is a big deal, but you will find most "hardware"
RAID
>>> implementations
>>> are somewhat limited, as the above adaptec only supports 4
snapshots and
>>> it is an
>>> optional feature.  You will find many array vendors will be happy
to
>>> charge lots
>>> of money for the snapshot feature.
>>
>> On top of that since the ZFS snapshot is at the file system level
it''s
>> much easier to use. You don''t have to quiesce the file system
first or
>> hope that when you take the snapshot you get a consistent data set.
I''ve
>> seen plenty of folks take hw raid snapshots without locking the file
>> system first, let alone quiescing the app, and getting garbage.
>
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss

Compared to hw raid only snapshots ZFS is still, imho, easier to use.

If you start talking about VSS, aka shadow copy for Windows, you''re now
at the fs level. I can see that VSS offers an API for 3rd parties to use 
but, as I literally just started reading about it, I''m not an expert. 
 From a quick glance I think the ZFS feature set is comparable. Is there 
a C++ API to ZFS? Not that I know of. Do you need one? Can''t think of a
reason off the top of my head given the way the zpool/zfs commands work.

Joseph Zhou wrote:
> Torrey, now this impressive as the old days with Sun Storage.
>
> Ok, ZFS PiT is only a software solution.
> The Windows VSS is not only a software solution, but also a 3rd party 
> integration standard from MS.
> What''s your comment on ZFS PiT is better than MS PiT, in light of 
> openness and 3rd-party integration???
>
> Talking about garbage!
> z
>
>
> ----- Original Message ----- From: "Torrey McMahon" <tmcmahon2
at yahoo.com>
> To: "Richard Elling" <Richard.Elling at Sun.COM>
> Cc: "Joseph Zhou" <jz at excelsioritsolutions.com>;
"William D. Hathaway"
> <william.hathaway at versatile.com>; <chriso at
ideasinternational.com>;
> <zfs-discuss at opensolaris.org>; <peterc at
ideasinternational.com>
> Sent: Sunday, December 07, 2008 1:58 AM
> Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on Sun 
> X4150/X4450
>
>
>> Richard Elling wrote:
>>> Joseph Zhou wrote:
>>>
>>>> Yeah?
>>>>
http://www.adaptec.com/en-US/products/Controllers/Hardware/sas/value/SAS-31605/_details/Series3_FAQs.htm
>>>>
>>>> Snapshot is a big deal?
>>>>
>>>
>>> Snapshot is a big deal, but you will find most "hardware"
RAID
>>> implementations
>>> are somewhat limited, as the above adaptec only supports 4
snapshots
>>> and it is an
>>> optional feature.  You will find many array vendors will be happy
to
>>> charge lots
>>> of money for the snapshot feature.
>>
>> On top of that since the ZFS snapshot is at the file system level 
>> it''s much easier to use. You don''t have to quiesce
the file system
>> first or hope that when you take the snapshot you get a consistent 
>> data set. I''ve seen plenty of folks take hw raid snapshots
without
>> locking the file system first, let alone quiescing the app, and 
>> getting garbage. 
>
>

Ok, Torrey, I like you, so one more comment before I go to bed --

Please go study the EMC NetWorker 7.5, and why EMC can claim leadership in 
VSS support.
Then, if you still don''t understand the importance of VSS, just ask me
in an
open fashion, I will teach you.

The importance of storage in system and application optimization can be very 
significant.
You do coding, do you know what''s TGT from IBM in COBOL, to be able to
claim
"enterprise technology"?
If not, please study.
http://publib.boulder.ibm.com/infocenter/pdthelp/v1r1/index.jsp?topic=/com.ibm.entcobol.doc_4.1/PGandLR/ref/rpbug10.htm

Open Storage is a great concept, but we can only win with realy advantages, 
not fake marketing lines.
I hope everyone enjoyed the discussion. I did.

zStorageAnalyst


----- Original Message ----- 
From: "Torrey McMahon" <tmcmahon2 at yahoo.com>
To: "Joseph Zhou" <jz at excelsioritsolutions.com>
Cc: "Richard Elling" <Richard.Elling at Sun.COM>; "William
D. Hathaway"
<william.hathaway at versatile.com>; <chriso at
ideasinternational.com>;
<zfs-discuss at opensolaris.org>; <peterc at ideasinternational.com>
Sent: Sunday, December 07, 2008 2:40 AM
Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on Sun 
X4150/X4450

> Compared to hw raid only snapshots ZFS is still, imho, easier to use.
>
> If you start talking about VSS, aka shadow copy for Windows,
you''re now at
> the fs level. I can see that VSS offers an API for 3rd parties to use but, 
> as I literally just started reading about it, I''m not an expert.
From a
> quick glance I think the ZFS feature set is comparable. Is there a C++ API 
> to ZFS? Not that I know of. Do you need one? Can''t think of a
reason off
> the top of my head given the way the zpool/zfs commands work.
>
> Joseph Zhou wrote:
>> Torrey, now this impressive as the old days with Sun Storage.
>>
>> Ok, ZFS PiT is only a software solution.
>> The Windows VSS is not only a software solution, but also a 3rd party 
>> integration standard from MS.
>> What''s your comment on ZFS PiT is better than MS PiT, in light
of
>> openness and 3rd-party integration???
>>
>> Talking about garbage!
>> z
>>
>>
>> ----- Original Message ----- From: "Torrey McMahon"
<tmcmahon2 at yahoo.com>
>> To: "Richard Elling" <Richard.Elling at Sun.COM>
>> Cc: "Joseph Zhou" <jz at excelsioritsolutions.com>;
"William D. Hathaway"
>> <william.hathaway at versatile.com>; <chriso at
ideasinternational.com>;
>> <zfs-discuss at opensolaris.org>; <peterc at
ideasinternational.com>
>> Sent: Sunday, December 07, 2008 1:58 AM
>> Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on Sun 
>> X4150/X4450
>>
>>
>>> Richard Elling wrote:
>>>> Joseph Zhou wrote:
>>>>
>>>>> Yeah?
>>>>>
http://www.adaptec.com/en-US/products/Controllers/Hardware/sas/value/SAS-31605/_details/Series3_FAQs.htm
>>>>> Snapshot is a big deal?
>>>>>
>>>>
>>>> Snapshot is a big deal, but you will find most
"hardware" RAID
>>>> implementations
>>>> are somewhat limited, as the above adaptec only supports 4
snapshots
>>>> and it is an
>>>> optional feature.  You will find many array vendors will be
happy to
>>>> charge lots
>>>> of money for the snapshot feature.
>>>
>>> On top of that since the ZFS snapshot is at the file system level
it''s
>>> much easier to use. You don''t have to quiesce the file
system first or
>>> hope that when you take the snapshot you get a consistent data set.
I''ve
>>> seen plenty of folks take hw raid snapshots without locking the
file
>>> system first, let alone quiescing the app, and getting garbage.
>>
>>
>

On Wed, Dec 3, 2008 at 6:37 PM, Aaron Blew <aaronblew at gmail.com> wrote:
> I''ve done some basic testing with a X4150 machine using 6 disks in
a RAID 5
> and RAID Z configuration.  They perform very similarly, but RAIDZ
definitely
> has more system overhead.  In many cases this won''t be a big deal,
but if
> you need as many CPU cycles as you can muster, hardware RAID may be your
> better choice.
>

Some people keep stressing the point that HW raid does not include snapshots
or what ever other features, or does so at cost, or ... or ... or .  It
seems to me like we assume that the above poster intended or implied the use
of another file system on the HW raid system.

The poster above did not specify a file system, so I may as well assume the
comparisons is between using ZFS with JBOD vs ZFS on HW-raid.

Then the features available to the administrator are essentially the same.
Not the question becomes: What are the pros and cons for each?

I have not tested this, but I would assume that the HW raid (forget about
cheap motherboard chipset integrated "fake-raid") will save some CPU
time
because the raid controller has got a dedicated processor to do the stripe
parity calculations.  In addition the ZFS routines may have an easier time
ITO selecting which disk to store the data on (only one disk to choose
from).

On the other hand, ZFS promises better fault detection, but presently this
is temptered by several open bugs against ZFS during situations where
degraded pools are present, eg pools freezing, etc.  HW raid seem to have
this sort of situation under control.

Some HW raids may offer re-layout without losing data.  ZFS does not (yet)
offer this.

ZFS claims better write performance in scenarios where less than a full
stripe width is updated, and raid5 suffers from the "write-hole"
problem.
Nicely defined here: http://blog.dentarg.net/2007/1/10/raid5-write-hole

ZFS updates are "atomic" - you never need to fsck the file system.

ZFS will work regardless of whether or not you have a HW raid disk
subsystem.

So... what other benefits has ZFS got (as defined in my second paragraph)

For what it is worth, have a look at my ZFS feature wishlist / AKA what it
would take to make ZFS _THE_ last word in storage management:
http://initialprogramload.blogspot.com/2008/07/zfs-missing-features.html

  _J

-- 
Any sufficiently advanced technology is indistinguishable from magic.
   Arthur C. Clarke

My blog: http://initialprogramload.blogspot.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20081207/b2dee15d/attachment.html>

On 07 December, 2008 - Johan Hartzenberg sent me these 6,3K bytes:
> For what it is worth, have a look at my ZFS feature wishlist / AKA what it
> would take to make ZFS _THE_ last word in storage management:
> http://initialprogramload.blogspot.com/2008/07/zfs-missing-features.html
#2 can kinda be solved with L2ARC.. Not entirely, but somewhat..

#3 is coming, but there is no hard ETA (according to Sun when I poked
them).

/Tomas
-- 
Tomas ?gren, stric at acc.umu.se, http://www.acc.umu.se/~stric/
|- Student at Computing Science, University of Ume?
`- Sysadmin at {cs,acc}.umu.se

I''m pretty sure I understand the importance of a snapshot API. (You
take
the snap, then you do the backup or whatever....) My point is that, at 
least on my quick read, you can do most of the same things with the ZFS 
command line utilities. The relevant question would then be how stable 
that is for the type of work we''re talking about.

Joseph Zhou wrote:
> Ok, Torrey, I like you, so one more comment before I go to bed --
>
> Please go study the EMC NetWorker 7.5, and why EMC can claim 
> leadership in VSS support.
> Then, if you still don''t understand the importance of VSS, just
ask me
> in an open fashion, I will teach you.
>
> The importance of storage in system and application optimization can 
> be very significant.
> You do coding, do you know what''s TGT from IBM in COBOL, to be
able to
> claim "enterprise technology"?
> If not, please study.
>
http://publib.boulder.ibm.com/infocenter/pdthelp/v1r1/index.jsp?topic=/com.ibm.entcobol.doc_4.1/PGandLR/ref/rpbug10.htm
>
>
> Open Storage is a great concept, but we can only win with realy 
> advantages, not fake marketing lines.
> I hope everyone enjoyed the discussion. I did.
>
> zStorageAnalyst
>
>
> ----- Original Message ----- From: "Torrey McMahon" <tmcmahon2
at yahoo.com>
> To: "Joseph Zhou" <jz at excelsioritsolutions.com>
> Cc: "Richard Elling" <Richard.Elling at Sun.COM>;
"William D. Hathaway"
> <william.hathaway at versatile.com>; <chriso at
ideasinternational.com>;
> <zfs-discuss at opensolaris.org>; <peterc at
ideasinternational.com>
> Sent: Sunday, December 07, 2008 2:40 AM
> Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on Sun 
> X4150/X4450
>
>
>> Compared to hw raid only snapshots ZFS is still, imho, easier to use.
>>
>> If you start talking about VSS, aka shadow copy for Windows,
you''re
>> now at the fs level. I can see that VSS offers an API for 3rd parties 
>> to use but, as I literally just started reading about it, I''m
not an
>> expert. From a quick glance I think the ZFS feature set is 
>> comparable. Is there a C++ API to ZFS? Not that I know of. Do you 
>> need one? Can''t think of a reason off the top of my head given
the
>> way the zpool/zfs commands work.
>>
>> Joseph Zhou wrote:
>>> Torrey, now this impressive as the old days with Sun Storage.
>>>
>>> Ok, ZFS PiT is only a software solution.
>>> The Windows VSS is not only a software solution, but also a 3rd 
>>> party integration standard from MS.
>>> What''s your comment on ZFS PiT is better than MS PiT, in
light of
>>> openness and 3rd-party integration???
>>>
>>> Talking about garbage!
>>> z
>>>
>>>
>>> ----- Original Message ----- From: "Torrey McMahon" 
>>> <tmcmahon2 at yahoo.com>
>>> To: "Richard Elling" <Richard.Elling at Sun.COM>
>>> Cc: "Joseph Zhou" <jz at excelsioritsolutions.com>;
"William D.
>>> Hathaway" <william.hathaway at versatile.com>; 
>>> <chriso at ideasinternational.com>; <zfs-discuss at
opensolaris.org>;
>>> <peterc at ideasinternational.com>
>>> Sent: Sunday, December 07, 2008 1:58 AM
>>> Subject: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on 
>>> Sun X4150/X4450
>>>
>>>
>>>> Richard Elling wrote:
>>>>> Joseph Zhou wrote:
>>>>>
>>>>>> Yeah?
>>>>>>
http://www.adaptec.com/en-US/products/Controllers/Hardware/sas/value/SAS-31605/_details/Series3_FAQs.htm
>>>>>>
>>>>>> Snapshot is a big deal?
>>>>>>
>>>>>
>>>>> Snapshot is a big deal, but you will find most
"hardware" RAID
>>>>> implementations
>>>>> are somewhat limited, as the above adaptec only supports 4 
>>>>> snapshots and it is an
>>>>> optional feature.  You will find many array vendors will be
happy
>>>>> to charge lots
>>>>> of money for the snapshot feature.
>>>>
>>>> On top of that since the ZFS snapshot is at the file system
level
>>>> it''s much easier to use. You don''t have to
quiesce the file system
>>>> first or hope that when you take the snapshot you get a
consistent
>>>> data set. I''ve seen plenty of folks take hw raid
snapshots without
>>>> locking the file system first, let alone quiescing the app, and
>>>> getting garbage.
>>>
>>>
>>
>
>

On Mon 08/12/08 08:14 , Torrey McMahon tmcmahon2 at yahoo.com
sent:
> I''m pretty sure I understand the importance of a snapshot API.
(You take
> the snap, then you do the backup or whatever....) My point is that, at 
> least on my quick read, you can do most of the same things with the ZFS
> command line utilities. The relevant question would then be how stable 
> that is for the type of work we''re talking about.
> 
Or through the APIs provided by libzfs.

-- 
Ian.

Ian Collins wrote:
>
>  On Mon 08/12/08 08:14 , Torrey McMahon tmcmahon2 at yahoo.com sent:
>   
>> I''m pretty sure I understand the importance of a snapshot API.
(You take
>> the snap, then you do the backup or whatever....) My point is that, at 
>> least on my quick read, you can do most of the same things with the ZFS
>> command line utilities. The relevant question would then be how stable 
>> that is for the type of work we''re talking about.
>>
>>     
> Or through the APIs provided by libzfs.
I''m not sure if those are published/supported as opposed to just being 
readable in the source. I think the ADM project is the droid we''re 
looking for.

Automatic Data Migration <http://opensolaris.org/os/project/adm/>
    ADM is designed to use the Data Storage Management API (aka XDSM) as
    defined in the CAE Specification XDSM as documented by the Open
    Group. XDSM provides an Open Standard API to Data Migration
    Applications (DMAPI) to manage file backup and recovery, automatic
    file migration, and file replication. ADM will take advantage of
    these APIs as a privileged application and extension to ZFS.

On Mon 08/12/08 09:14 , Torrey McMahon tmcmahon2 at yahoo.com
sent:
> Ian Collins wrote:
> > Or through the APIs provided by libzfs.
> I''m not sure if those are published/supported as opposed to just
being
> readable in the source. I think the ADM project is the droid we''re
> looking for.
> 
Fair point, I''ve been working with my own (C++) wrapper which abstracts
the differences.

-- 
Ian

Yes, yes, Torrey, that''s why I like you!

You are getting there -- the argument of snopshot is not key in its absolute 
elegance, but what it does in the overall solution. When you are talking 
about PiT with ADM, it made more sense, didn''t it?

Please keep in mind that OpenSolaris and ZFS don''t need to be the
greatest
technology today, and we need to respect the older generation
engineers''
thoughts -- it''s an evolution of transfering enterprise capabilities to
industry-standard solutions -- not a revolution that Sun Storage just 
re-invented everything.

And think strategically, is VSS just an API?   Even it is, by some logic, 
but what this API doos, in MS long term marketing strategy and its intent to 
claim "enterprise". -- and how OpenSolaris and ZFS can claim
"more
enterprise", one day???

I have lots other work to do, cannot chat no more.
But this is the first year since 2002 that I did not visit Sun Storage, and 
chat with Real Sun Storage folks over drinks. Miss you guys!
As every year, here is my contribution to open storage -- my frank comments.

Happy holidays!
zStorageAnalyst

----- Original Message ----- 
From: "Ian Collins" <ian at ianshome.com>
To: <ian at ianshome.com>; "Torrey McMahon" <tmcmahon2 at
yahoo.com>
Cc: "Joseph Zhou" <jz at excelsioritsolutions.com>;
"William D. Hathaway"
<william.hathaway at versatile.com>; <chriso at
ideasinternational.com>;
<zfs-discuss at opensolaris.org>; <peterc at
ideasinternational.com>; "Richard
Elling" <Richard.Elling at Sun.COM>
Sent: Sunday, December 07, 2008 3:26 PM
Subject: Re: Re: [zfs-discuss] Hardware Raid Vs ZFS implementation on Sun 
X4150/X4450


On Mon 08/12/08 09:14 , Torrey McMahon tmcmahon2 at yahoo.com
sent:
> Ian Collins wrote:
> > Or through the APIs provided by libzfs.
> I''m not sure if those are published/supported as opposed to just
being
> readable in the source. I think the ADM project is the droid we''re
> looking for.
>
Fair point, I''ve been working with my own (C++) wrapper which abstracts
the
differences.

-- 
Ian

On Sun, 7 Dec 2008, Joseph Zhou wrote:
>
> Please keep in mind that OpenSolaris and ZFS don''t need to be the
greatest
> technology today, and we need to respect the older generation
engineers''
> thoughts -- it''s an evolution of transfering enterprise
capabilities to
> industry-standard solutions -- not a revolution that Sun Storage just
> re-invented everything.
I am not sure what you are trying to say.  Sometimes revolution is 
necessary in order for there to be substantial improvement.  ZFS is a 
revolution rather than an evolution.
> And think strategically, is VSS just an API?   Even it is, by some logic,
> but what this API doos, in MS long term marketing strategy and its intent
to
> claim "enterprise". -- and how OpenSolaris and ZFS can claim
"more
> enterprise", one day???
VSS is an NTFS filesystem feature which seems to only have become 
usable as of Windows Server 2003.  It includes arbitrary limitations 
which don''t exist in ZFS.  Clearly you are "sold" on this 
closed-source technology.

To my way of thinking individual components are not in themselves 
"enterprise".  The notion of "enterprise" is that there is a
system of
well integrated components which provide the performance, reliability, 
and maintainability required for mission critical installations. 
Since Microsoft is not a vertically integrated system vendor it can 
only qualify its products as being "enterprise" in conjuction with a 
real system vendor in order to offer an integrated solution. 
Otherwise it is just a collection of parts which may or may not even 
function together.

Bob
=====================================Bob Friesenhahn
bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer,    http://www.GraphicsMagick.org/

>>>>> "jh" == Johan Hartzenberg <jhartzen at
gmail.com> writes:
    jh> raid5 suffers from the "write-hole" problem.

this is only when you use it without a battery.

I can tell you a little about Windows VSS snapshots compared to ZFS ones, since
one of the main reasons I''m so interested in ZFS is because windows
snapshots are so useless.

For windows VSS:
* You have OS overhead for taking the snapshot, as opposed to it being
instantaneous for ZFS.  Microsoft actually recommend the snapshots are stored on
a separate disk.
* You have to reserve space in advance for them, so if you guess wrong
you''re out of luck.
* Microsoft''s snapshots can have one schedule.  They support hourly,
daily or weekly snapshots, but you can only pick one period.
* You are limited to 64 snapshots.

So if you want hourly snapshots of your data, you''re not even going to
have 3 days worth of backups.  If you can live with daily backups you can manage
2 months worth.

When you compare that to Tim''s excellent auto backup service it makes
VSS look like a joke.  While ZFS doesn''t actually limit how many
snapshots you keep, with just 90 you can run:

8x 15 minute snapshots
48x hourly snapshots
14x daily snapshots
8x weekly snapshots
12x monthly snapshots

So you have snapshots being taken *far* more regularly than VSS can manage, and
they go back a full year with considerable overlap between the different
periods.
-- 
This message posted from opensolaris.org