[Novalug] Seagate Momentus XT hybrid Drive
James Ewing Cottrell 3rd
JECottrell3 at Comcast.NET
Tue Jun 1 13:36:32 EDT 2010
In a way, I am with you. Two mirrors *is* two (even) parity drives.
But mirroring is Much Simpler than RAID.
On 6/1/2010 11:37 AM, John Franklin wrote:
> I'm with Miguel on this one.
>
> Given a set of disks all bought and deployed at the same time in the
> same RAID array, the odds of a second drive failing during the stress
> of a rebuild is higher than I like.
What are the odds? Has anyone you ever encountered been bitten by these
odds? Are they great enough to cripple the performance when the system
is actually working?
But why stop there? If you believe in two failing, aren't three almost
as likely to fail? Four, anyone?
Are you willing to trade Certain Degradation for Implausibility-Squared
Failure Protection? Not Me.
Why not swap some of your new drive with old burned-in one out in the field?
> More so, the cost of a second parity drive in each RAID volume at each
> location is a fraction of the cost to deploy a new RAID volume and
> reload it from backups plus the cost of loss of business continuity.
Please note that we are talking about exactly 4 disks here. By the time
you have reserved two for parity, you might as well go for mirroring.
>
> If you really REALLY care about your data, you'll have another array
> in another time zone already running and kept in sync.
I'm with you there.
>
> jf
JIM
>
> On May 31, 2010, at 2:00 PM, James Ewing Cottrell 3rd wrote:
>
>> I don't know why people think they have to avoid Religious
>> Battles....we have been doing it for ages.
>>
>> In fact, I dislike the term Religious applied to either Battles *or*
>> Arguments. That tends to imply that they cannot be decided on
>> Technical Merit alone.
>>
>> While I have heard of "two disks going bad simultaneously", I have
>> never actually SEEN it, nor, I bet, has any of you either. And if any
>> of you have, it's just the exception that proves the rule. In most
>> cases, these are infant failures related to two disks that were
>> produced on the same assembly line...from a "bad batch".
>>
>> Besides, you only die if the second one goes bad before you can
>> rebuild your mirror/stripes/raid. Besides...a RAID rebuild is more
>> expensive than a mirror resilvering.
>>
>> Like I said, I optimize for the Working Case than for the Failure Case.
>>
>> If you case THAT much about your data, you probably need 3 way
>> mirrors. And backing up to Live Disk helps a lot with Restoration as
>> opposed to backing up to tape.
>>
>> JIM
>>
>> P.S. I said "NetApp notwithstanding" because I am a Great Fan as
>> well. But they are not a Panacea. I used one in 2005 to hold the
>> directories of 100,000+ users mail spools, and it wasn't exactly
>> Blindingly Fast. BTW, we did hash the names...mine would have been
>> called /mail/je/co/jecottrell3 at comcast.net
>> <mailto:/mail/je/co/jecottrell3 at comcast.net>
>>
>> On 5/31/2010 1:25 PM, Miguel Gonzalez wrote:
>>> And two disks going bad is very common nowadays. First because the error ratio has increased with bigger drives and due to this, it can happen that another disk fails when a rebuild is in place...a long time ago rebuilding time could be a question of hours, nowadays it can be even days (try to rebuild a 1 Tb HD in a hot spare).
>>>
>>> It's quite pricey but Netapp is amazing. I haven't seen something so reliable in my life. I've seen several Virtual Machines running on netapp devices, bringing down one path and taking over the other path in a few seconds and resulting in a 1-2 seconds network glitch that was close to seamless (only one virtual machine from 100 was down).
>>>
>>> At work we have Dell multipath HD cabins, Sun SCSI HDs and other hardware and lately we are moving everything to Netapp. And you can sleep tight at the end of the day. And support is amazing, they call you even before you have realized a HD has failed and ship you a HD in a few hours...That's what I call REAL 4 hours critical support...
>>>
>>> Sorry for getting into the religious battle ;)
>>>
>>> Miguel
>>>
>>> --- El lun, 31/5/10, James Ewing Cottrell 3rd<JECottrell3 at Comcast.NET> escribió:
>>>
>>>
>>>> De: James Ewing Cottrell 3rd<JECottrell3 at Comcast.NET>
>>>> Asunto: Re: [Novalug] Seagate Momentus XT hybrid Drive
>>>> Para:novalug at calypso.tux.org
>>>> Fecha: lunes, 31 de mayo, 2010 12:43
>>>> NetApp notwithstanding, I would be
>>>> prepared to fight that Religious
>>>> Battle under Any and All Circumstances, especially now that
>>>> disk is
>>>> Laughably Cheap. But Not Today.
>>>>
>>>> Here I was speaking of ONE special case...Exactly Four
>>>> Disks...RAID6 vs
>>>> RAID10.
>>>>
>>>> RAID6 is superior in One Small Edge Case...it will survive
>>>> ANY two disks
>>>> going bad. RAID10 will suvive FOUR out of SIX cases of two
>>>> going bad.
>>>>
>>>> I like to optimize for the Working Case, not the rare Rare
>>>> RARE Failure
>>>> case. Now maybe on the Mars Rover...
>>>>
>>>> JIM
>>>>
>>>> P.S. Of course, Dave advice to Optimize the System as a
>>>> Whole is well
>>>> taken, altho I/O is usually one of the biggest concerns.
>>>>
>>>> On 5/28/2010 4:55 PM, American Dave Kline wrote:
>>>>
>>>>> On Fri, May 28, 2010 at 11:41:24AM -0400, James Ewing
>>>>>
>>>> Cottrell 3rd wrote:
>>>>
>>>>>
>>>>>
>>>>>> Kudos for the RAID10
>>>>>>
>>>> format. I recently worked with a groups that used
>>>>
>>>>>> RAID5 with a spare.
>>>>>>
>>>> Geeze.
>>>>
>>>>>>
>>>>>>
>>>>> RAID 5 gets dogged a lot, though in practice it works
>>>>>
>>>> quite well. It's
>>>>
>>>>> a good compromise between array size, speed, and
>>>>>
>>>> reliability. A hot
>>>>
>>>>> spare as well can buy you a bit more
>>>>>
>>>> reliability. As the devices per
>>>>
>>>>> array grows, RAID 5 makes less sense since the chance
>>>>>
>>>> of having
>>>>
>>>>> concurrent failures grows.
>>>>>
>>>>> I often see RAID levels get brought into technical
>>>>>
>>>> religious battles,
>>>>
>>>>> when the time is almost always better spent optimizing
>>>>>
>>>> the system as a whole.
>>>>
>>>>> -A. Dave
>>>>> _______________________________________________
>>>>> Novalug mailing list
>>>>> Novalug at calypso.tux.org
>>>>> http://calypso.tux.org/mailman/listinfo/novalug
>>>>>
>>>>>
>>>>>
>>>> _______________________________________________
>>>> Novalug mailing list
>>>> Novalug at calypso.tux.org
>>>> http://calypso.tux.org/mailman/listinfo/novalug
>>>>
>>>>
>>>
>>> _______________________________________________
>>> Novalug mailing list
>>> Novalug at calypso.tux.org
>>> http://calypso.tux.org/mailman/listinfo/novalug
>>>
>>>
>>>
>>> No virus found in this incoming message.
>>> Checked by AVG -www.avg.com
>>> Version: 9.0.819 / Virus Database: 271.1.1/2908 - Release Date: 05/31/10 02:25:00
>>>
>>>
>>
>> _______________________________________________
>> Novalug mailing list
>> Novalug at calypso.tux.org <mailto:Novalug at calypso.tux.org>
>> http://calypso.tux.org/mailman/listinfo/novalug
>
> --
> John Franklin
> franklin at elfie.org <mailto:franklin at elfie.org>
> ICBM: 39º 01' 58.4"N 77º 24' 49.6"W Z+84m
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://calypso.tux.org/pipermail/novalug/attachments/20100601/84491723/attachment.html
More information about the Novalug
mailing list