One SSD Vs. Two In RAID: Which Is Better?

[mapesdhs]

Please trim your quoting, it just clogs up the page.


Suferbus writes:
> I am not admitting sh*t because that is not what I meant, ...

Then I can only assume you didn't read what you'd typed before submitting
the post, because the text is clearly factually wrong. I don't care
whether you admit that or not; what I do care about is such information
misleading others who wish to know the truth in order to make better
build/buying decisions.


> ... and your not right either. ...

Yes I am.


> ... Raid0 is not needed, ...

I didn't comment on whether it's needed or not. That is a decision for
each user, whether they wish to accept the risks, and if so then what
kind of backup solution they use. I personally would not use RAID0 for a
system drive, but that's my personal choice.


> ... The performance gain a user gets just from
> upgrading to the SSD is enough, ...

For you, maybe. Some people want more, eg. much higher sequential I/O,
which works very well with SSD RAID0, or better performance with
multiplayer games that involve a lot of server traffic. And read posts
I've made elsewhere, you'll find I always recommend RAID10 instead. Don't
try and infer something I haven't said; not once have I said whether I
think it's objectively a good idea or not - that's up to each user to
make their own decision based on their intended task.


> not mean my first comment any other way than that way, and your full of
> sh*t period.

Ah, resorting to direct insults and bad language, a sure sign you know
you initially posted in error. 8)


> There you go just putting words in my mouth again. ...

Read your own post, you're the one who said something factually wrong.


> do you understand that or should i repeat it one more time.. here

I never said failure was less likely. I said the risks are such that many
nevertheless find them acceptable. I have 36 disks in RAID0 because it's
just a scratch drive for analysing Flame performance; if the array fails,
it doesn't matter a hoot, I can reload the data in minutes.


> lets try it one letter at a time.. ...

Maybe you should have done that with your first post, perhaps it would have
been more obvious you were about to say something so utterly incorrect.


> worth the performance gain if you are already using SSD drives....

Makes me laugh you think *your* opinion on the risks is somehow universal
fact. Plenty of people use SSD RAID0 perfectly fine with no problems (or
so they say). On the other hand, many others have indeed had failures and
loss of data. Peoples' experiences vary, but you're trying to claim
there's some kind of universal view which holds for everyone, which is
just silly. MTBF does not mean that X% of one's drives will definitely
fail within a given time period, probabilities don't work that way. One
person might use RAID0 for years and have no issues, while someone else
might find it a nightmare of multiple failures. Whether or not it's a
risk worth taking is not your right to determine in some kind of
general manner, so quit preaching. What I tell people who ask is that,
if they want to use RAID0, then they should consider whether it fits
their task and also what backup solution they intend using.


> make you a hard drive expert, ...

Seems like I know more about these issues than you though. Read your
original post, it's just plain garbage. Are you really still saying
your original post was correct? I think I'm heading for another ROFL...

I'll quote you again: "If you want 1 tb of storage, you cannot run 2 in raid 0."

So are you saying that comment is true? Because it's not true at all,
it's utter nonsense.


Btw, your comments about failure probability are incredibly simplistic.
For a start, compound failure rates are not linear. A multi-drive array
has a higher chance of failure over a given time frame, but the MTBF of
each drive essentially doesn't change. Combined probabilities multiply in
a complex manner, they don't just add as you seem to think they do (5%
MTBF of each drive does not mean the MTBF of the array is 10%, that's wrong).
It also matters greatly what the workload is and the type of disk - Enterprise
disks are more reliable, thus Google's results showing pretty horrible results
for consumer drives. Likewise, SSDs vary aswell.

Ian.

 

[sna]

Thanks for this review Toms! I would have liked to see a comparison with say OCZ SSDs. Any idea why the Samsung SSD's have such low read speeds? 240 mb/s is pretty slow in comparison.

The Samsung 840 Pro are faster. go for the 840 pro

 

[Suferbus]

Please trim your quoting, it just clogs up the page.


Suferbus writes:
> I am not admitting sh*t because that is not what I meant, ...

Then I can only assume you didn't read what you'd typed before submitting
the post, because the text is clearly factually wrong. I don't care
whether you admit that or not; what I do care about is such information
misleading others who wish to know the truth in order to make better
build/buying decisions.


> ... and your not right either. ...

Yes I am.


> ... Raid0 is not needed, ...

I didn't comment on whether it's needed or not. That is a decision for
each user, whether they wish to accept the risks, and if so then what
kind of backup solution they use. I personally would not use RAID0 for a
system drive, but that's my personal choice.


> ... The performance gain a user gets just from
> upgrading to the SSD is enough, ...

For you, maybe. Some people want more, eg. much higher sequential I/O,
which works very well with SSD RAID0, or better performance with
multiplayer games that involve a lot of server traffic. And read posts
I've made elsewhere, you'll find I always recommend RAID10 instead. Don't
try and infer something I haven't said; not once have I said whether I
think it's objectively a good idea or not - that's up to each user to
make their own decision based on their intended task.


> not mean my first comment any other way than that way, and your full of
> sh*t period.

Ah, resorting to direct insults and bad language, a sure sign you know
you initially posted in error. 8)


> There you go just putting words in my mouth again. ...

Read your own post, you're the one who said something factually wrong.


> do you understand that or should i repeat it one more time.. here

I never said failure was less likely. I said the risks are such that many
nevertheless find them acceptable. I have 36 disks in RAID0 because it's
just a scratch drive for analysing Flame performance; if the array fails,
it doesn't matter a hoot, I can reload the data in minutes.


> lets try it one letter at a time.. ...

Maybe you should have done that with your first post, perhaps it would have
been more obvious you were about to say something so utterly incorrect.


> worth the performance gain if you are already using SSD drives....

Makes me laugh you think *your* opinion on the risks is somehow universal
fact. Plenty of people use SSD RAID0 perfectly fine with no problems (or
so they say). On the other hand, many others have indeed had failures and
loss of data. Peoples' experiences vary, but you're trying to claim
there's some kind of universal view which holds for everyone, which is
just silly. MTBF does not mean that X% of one's drives will definitely
fail within a given time period, probabilities don't work that way. One
person might use RAID0 for years and have no issues, while someone else
might find it a nightmare of multiple failures. Whether or not it's a
risk worth taking is not your right to determine in some kind of
general manner, so quit preaching. What I tell people who ask is that,
if they want to use RAID0, then they should consider whether it fits
their task and also what backup solution they intend using.


> make you a hard drive expert, ...

Seems like I know more about these issues than you though. Read your
original post, it's just plain garbage. Are you really still saying
your original post was correct? I think I'm heading for another ROFL...

I'll quote you again: "If you want 1 tb of storage, you cannot run 2 in raid 0."

So are you saying that comment is true? Because it's not true at all,
it's utter nonsense.


Btw, your comments about failure probability are incredibly simplistic.
For a start, compound failure rates are not linear. A multi-drive array
has a higher chance of failure over a given time frame, but the MTBF of
each drive essentially doesn't change. Combined probabilities multiply in
a complex manner, they don't just add as you seem to think they do (5%
MTBF of each drive does not mean the MTBF of the array is 10%, that's wrong).
It also matters greatly what the workload is and the type of disk - Enterprise
disks are more reliable, thus Google's results showing pretty horrible results
for consumer drives. Likewise, SSDs vary aswell.

Ian.

Complete Horse Sh*t

 

[Suferbus]

Please trim your quoting, it just clogs up the page.


Suferbus writes:
> I am not admitting sh*t because that is not what I meant, ...

Then I can only assume you didn't read what you'd typed before submitting
the post, because the text is clearly factually wrong. I don't care
whether you admit that or not; what I do care about is such information
misleading others who wish to know the truth in order to make better
build/buying decisions.


> ... and your not right either. ...

Yes I am.


> ... Raid0 is not needed, ...

I didn't comment on whether it's needed or not. That is a decision for
each user, whether they wish to accept the risks, and if so then what
kind of backup solution they use. I personally would not use RAID0 for a
system drive, but that's my personal choice.


> ... The performance gain a user gets just from
> upgrading to the SSD is enough, ...

For you, maybe. Some people want more, eg. much higher sequential I/O,
which works very well with SSD RAID0, or better performance with
multiplayer games that involve a lot of server traffic. And read posts
I've made elsewhere, you'll find I always recommend RAID10 instead. Don't
try and infer something I haven't said; not once have I said whether I
think it's objectively a good idea or not - that's up to each user to
make their own decision based on their intended task.


> not mean my first comment any other way than that way, and your full of
> sh*t period.

Ah, resorting to direct insults and bad language, a sure sign you know
you initially posted in error. 8)


> There you go just putting words in my mouth again. ...

Read your own post, you're the one who said something factually wrong.


> do you understand that or should i repeat it one more time.. here

I never said failure was less likely. I said the risks are such that many
nevertheless find them acceptable. I have 36 disks in RAID0 because it's
just a scratch drive for analysing Flame performance; if the array fails,
it doesn't matter a hoot, I can reload the data in minutes.


> lets try it one letter at a time.. ...

Maybe you should have done that with your first post, perhaps it would have
been more obvious you were about to say something so utterly incorrect.


> worth the performance gain if you are already using SSD drives....

Makes me laugh you think *your* opinion on the risks is somehow universal
fact. Plenty of people use SSD RAID0 perfectly fine with no problems (or
so they say). On the other hand, many others have indeed had failures and
loss of data. Peoples' experiences vary, but you're trying to claim
there's some kind of universal view which holds for everyone, which is
just silly. MTBF does not mean that X% of one's drives will definitely
fail within a given time period, probabilities don't work that way. One
person might use RAID0 for years and have no issues, while someone else
might find it a nightmare of multiple failures. Whether or not it's a
risk worth taking is not your right to determine in some kind of
general manner, so quit preaching. What I tell people who ask is that,
if they want to use RAID0, then they should consider whether it fits
their task and also what backup solution they intend using.


> make you a hard drive expert, ...

Seems like I know more about these issues than you though. Read your
original post, it's just plain garbage. Are you really still saying
your original post was correct? I think I'm heading for another ROFL...

I'll quote you again: "If you want 1 tb of storage, you cannot run 2 in raid 0."

So are you saying that comment is true? Because it's not true at all,
it's utter nonsense.


Btw, your comments about failure probability are incredibly simplistic.
For a start, compound failure rates are not linear. A multi-drive array
has a higher chance of failure over a given time frame, but the MTBF of
each drive essentially doesn't change. Combined probabilities multiply in
a complex manner, they don't just add as you seem to think they do (5%
MTBF of each drive does not mean the MTBF of the array is 10%, that's wrong).
It also matters greatly what the workload is and the type of disk - Enterprise
disks are more reliable, thus Google's results showing pretty horrible results
for consumer drives. Likewise, SSDs vary aswell.

Ian.

The dude that was quoting me is a complete a** and just ruined this forum by misquoting me and re-posting my comments therefore clogging up the page and this forum..Someone needs to poke him in the eye with a hot poker....

 

[mapesdhs]

Complete Horse Sh*t

I quoted you exactly. Go check your original post. I can only conclude
you're just trolling, or have some kind of mental problem. And you're
the one clogging up the page by quoting someone's entire post (twice!!).

Ian.

 

[piklar]

Try making a RamDisk with a decent portion of DDR3 2400 . Using crystalmark your looking at read times of 9000 MB/s + and Write times of 11000 MB/s in Seq and 512K . Incredible speed compared to todays SSDs.

 

[Shako73]

I have every SSD in pairs RAID 0 with high end Adaptec RAID Card and I do backup everyday. Always have 1GB/s Read/Write. I backup and have peace of mind.

 

[Shako73]

I have every SSD in pairs RAID 0 with high end Adaptec RAID Card and I do backup everyday. Always have 1GB/s Read/Write. I backup and have peace of mind.

 

[Haravikk]

I'd be interested to see a comparison including three disks in a RAID-5 as well. With the right RAID controller or software a RAID-5 should be capable of far superior random read/write performance, at the cost of some write performance, but with the added bonus of redundancy.

In fact, while RAID-5 enables the same roughly 2x speed for sequential reading that RAID-0 does, it can theoretically offer up to 3x random read/write speed since data is distributed between three disks.

 

[wsmith]

Thanks for this article, which is pertinent to my work interest: pro video editing.

However, if I may suggest the following, for a possible future article on this subject.

Sequential Read is paramount, followed by Sequential Write. But we need to know how many concurrent streams of video and the aggregate throughput performance of x number of streams being played back - under various CODECs.

Example: I often use Adobe Premiere's Multicam editing function. That involves simultaneously playing back a number of streams concurrently (each seen in an individual player window) and using keyboard (or mouse) commands to select which of those sources is actually laid to the program.

In my case I often with shoot 3 cameras and use Multicam in post rather than employ a live switcher on set. Others use more cameras. Premiere now supposedly allows an indefinite number of sources (number determined by a given system's performance capability). 4, 5, even 6 camera/sources is not uncommon for some producers.

So, we wonder, how many current streams can these drives handle? If using highly popular and highly compressed AVCHD or H.264, the throughput requirement is less per stream. If using the highly professional Apple Prores or Avid DNxHD CODECs - which are far less compressed and thus result in much larger files - the throughput requirement is far greater.

In between these two extremes lie the 50, 100, and now with AVC Ultra, 200Mbps.

I know seeing results for all above scenarios would be of interest to indie producers and editors.

I assume that synthetic benchmark for Multicam performance probably doesn't exist AFAIK. All of the editing and capture utilities have Dropped Frame warnings and logging. That's all we need; have any frames been dropped during playback. And if the streams play smoothly.

(I highly doubt uncompressed video would be used by anyone for such things as Multi Camera productions. Just not needed and the storage requirements are off the price scale. At least for now.)

You do the benchmarks and I'll see to it that every pro video site I know of hears about it and Tom's! We poor indie video pros don't have a site that that does such serious analytical benchmarking of video production performance specs - aside from the usual Photoshop, Premiere, After Effects, Maya, etc. etc. But those tests don't address the above concerns.

Thanks!