FIO and Storage Data Integrity testing

FIO and Storage Data Integrity testing

[Grant Grundler]

Hi Jens!

My summer intern (Juan Casse) is working on a data integrity/retention
test and has the first prototype running. (Juan might follow up with a
gerrit code review for the first cut)

The goals of this data integrity test:
o GPL implementation I can give to storage vendors as part of HW Qual test
o verify data written to storage is available and correct
o log writes (save a map) so we can repeatedly verify writes at a
later date (weeks or months later)
o provide some "bread crumbs" for debugging when data is NOT correct.
   (Not available typically will result in reported errors)
o work on any block storage device (ie no knowledge of specific device
geometry or flash vs magnetic vs optical or removable vs built-in -
some 'workloads' might be geared for specific types of storage)

o specify workload the same way fio does (multi-threaded, async,
random vs seq, read vs write mix, etc)
o collect same performance statistics that fio does (latency
histograms in particular)
o be done in < 6 weeks by a full time intern. :)

It seems like he should be doing something with fio.  I saw this query
in the Fio mailing list archive but didn't see a response:
   http://www.spinics.net/lists/fio/msg01933.html

Since these are destructive tests, I expect the primary target
"audience" is anyone working on Storage HW or wants to confirm Storage
HW is operating correctly before deploying $$$ worth of HW.

Questions:
1) You know anyone else developing data integrity/retention testing with fio?

2) Other good open source data integrity test I should know about?
Both Juan and I looked and didn't find anything better than
"badblocks". As a first exercise, Juan has written an autotest for
badblocks on ChromeOS:
   https://gerrit.chromium.org/gerrit/#/c/61786/

3) You have a preference on how this might be implemented if (a) we
used code from OR (b) integrated this functionality into fio?

(2) is probably a longer answer - refactor vs integrate is a good
first answer for now. :)

thanks,
grant

Re: FIO and Storage Data Integrity testing

[Jens Axboe]

On 07/31/2013 02:32 PM, Grant Grundler wrote:
> Hi Jens!
> 
> My summer intern (Juan Casse) is working on a data integrity/retention
> test and has the first prototype running. (Juan might follow up with a
> gerrit code review for the first cut)
> 
> The goals of this data integrity test:
> o GPL implementation I can give to storage vendors as part of HW Qual test
> o verify data written to storage is available and correct

Both of these fio already satisfies.

> o log writes (save a map) so we can repeatedly verify writes at a
> later date (weeks or months later)

One approach that I have started favoring is instead providing coverage
through the use of an lfsr. This negates the need for dedicated tracking
map or log. Fio supports this through random_generator=lfsr. The one
thing that does not YET work is lfsr and multiple block sizes. Should be
doable through layered use of multiple lfsrs. Something for your intern
to tackle?

> o provide some "bread crumbs" for debugging when data is NOT correct.
>    (Not available typically will result in reported errors)

So lets say that one of the fio verify modes was augmented to include a
time stamp (say the meta mode, or could even be added to all the verify
modes), that could be part of the bread crumbs and aid in judging
retention of the data.

> o work on any block storage device (ie no knowledge of specific device
> geometry or flash vs magnetic vs optical or removable vs built-in -
> some 'workloads' might be geared for specific types of storage)

Certainly

> o specify workload the same way fio does (multi-threaded, async,
> random vs seq, read vs write mix, etc)
> o collect same performance statistics that fio does (latency
> histograms in particular)
> o be done in < 6 weeks by a full time intern. :)

Depends on the quality of the intern :-)

> It seems like he should be doing something with fio.  I saw this query
> in the Fio mailing list archive but didn't see a response:
>    http://www.spinics.net/lists/fio/msg01933.html
> 
> Since these are destructive tests, I expect the primary target
> "audience" is anyone working on Storage HW or wants to confirm Storage
> HW is operating correctly before deploying $$$ worth of HW.
> 
> Questions:
> 1) You know anyone else developing data integrity/retention testing with fio?

I know of lots of people/companies using it for data integrity testing
(I even work for one of them), but not data retention. So I think that
would be a very interesting feature to add.

> 2) Other good open source data integrity test I should know about?
> Both Juan and I looked and didn't find anything better than
> "badblocks". As a first exercise, Juan has written an autotest for
> badblocks on ChromeOS:
>    https://gerrit.chromium.org/gerrit/#/c/61786/

No idea, sorry.

> 3) You have a preference on how this might be implemented if (a) we
> used code from OR (b) integrated this functionality into fio?

I think the the data retention aspect should be integrated into the
verify modes. The fio verification modes checksum both the stored
header, as well as the actual contents. There might be additional
tracking required on the side for retention, to be able to pass some
interesting info on where we seem to fall off a cliff.

I'll be happy to work with you guys on this, both on the initial design
phase and the final integration into fio.

-- 
Jens Axboe

Re: FIO and Storage Data Integrity testing

[Grant Grundler]

On Wed, Jul 31, 2013 at 2:23 PM, Jens Axboe <axboe@kernel.dk> wrote:
...
>> o log writes (save a map) so we can repeatedly verify writes at a
>> later date (weeks or months later)
>
> One approach that I have started favoring is instead providing coverage
> through the use of an lfsr.

"Log File Structured <something>?" Expand that and I can better assess
if I think this is a good idea or not.

I'm not too picky about the implementation that meets the requirement.
I understand maps or sparse maps can get very awkward to handle for
devices with more than a few billion blocks. ("Can plz haz 4k blox?"
:)

> This negates the need for dedicated tracking
> map or log. Fio supports this through random_generator=lfsr. The one
> thing that does not YET work is lfsr and multiple block sizes. Should be
> doable through layered use of multiple lfsrs. Something for your intern
> to tackle?

Yes. That would be fair game.

>> o provide some "bread crumbs" for debugging when data is NOT correct.
>>    (Not available typically will result in reported errors)
>
> So lets say that one of the fio verify modes was augmented to include a
> time stamp (say the meta mode, or could even be added to all the verify
> modes), that could be part of the bread crumbs and aid in judging
> retention of the data.

I want four pieces of data for bread crumbs:
o timestamp
o LBA written (e.g. if it's a partition, that means the offset into
the partition)
o magic number for that test run (think of it as a GUID - verifies the
block was written by fio)
o generation number in the case that we rewrite an LBA - so we can
detect stale data

I haven't checked if fio provides all four of those.

BTW, to eventually support adding "trim/discard command" testing into
the mix, we would need to know when a block is explicitly unmapped and
should be all zeros if we attempt to read it.

...
>> o be done in < 6 weeks by a full time intern. :)
>
> Depends on the quality of the intern :-)

Juan is capable enough. He's also extremely persistent. So I think
yes, he can get this done.

...
>> Questions:
>> 1) You know anyone else developing data integrity/retention testing with fio?
>
> I know of lots of people/companies using it for data integrity testing
> (I even work for one of them), but not data retention. So I think that
> would be a very interesting feature to add.

Good. thanks!

My review a few months ago of fio docs didn't give me the impression
the data integrity checking was providing enough bread crumbs for good
debugging or able to detect stale data. But my memory isn't very good
and I could be wrong or just out of date.

>> 3) You have a preference on how this might be implemented if (a) we
>> used code from OR (b) integrated this functionality into fio?
>
> I think the the data retention aspect should be integrated into the
> verify modes. The fio verification modes checksum both the stored
> header, as well as the actual contents. There might be additional
> tracking required on the side for retention, to be able to pass some
> interesting info on where we seem to fall off a cliff.

Ok. Let me clarify the requirement for data retention: I wanted
"verify" to be an option to the "read" workload mix. So not
necessarily all data that gets written will get verified "during" the
write workload. The reason is performance statistics need to be as
consistent as possible without "verify" in a mixed read/write
workload. To verify everything that was written or trimmed, we can
invoke fio again (think autotest invoking fio twice per test run) to
check for retention. And then invoke fio many more times while the
device is getting baked in a thermal chamber.

> I'll be happy to work with you guys on this, both on the initial design
> phase and the final integration into fio.

Awesome - thank you!

Design phase?! :) This is design phase. :)

cheers!
grant

Re: FIO and Storage Data Integrity testing

[Jens Axboe]

On Wed, Jul 31 2013, Grant Grundler wrote:
> On Wed, Jul 31, 2013 at 2:23 PM, Jens Axboe <axboe@kernel.dk> wrote:
> ...
> >> o log writes (save a map) so we can repeatedly verify writes at a
> >> later date (weeks or months later)
> >
> > One approach that I have started favoring is instead providing coverage
> > through the use of an lfsr.
> 
> "Log File Structured <something>?" Expand that and I can better assess
> if I think this is a good idea or not.

Linear feedback shift registers. Basically a way to generate a "random"
sequence of numbers that are guarenteed not to repeat until the cycle is
repeated. Then you never have to do on-the-side tracking to avoid
overlaps or overwrites.

For verify, you simple re-set the seed to the value when the sequence
started to hit all of the same blocks again. Or variations around that
theme, if you only want some of them.

> I'm not too picky about the implementation that meets the requirement.
> I understand maps or sparse maps can get very awkward to handle for
> devices with more than a few billion blocks. ("Can plz haz 4k blox?"
> :)

Indeed, fio has the "axmap" to track it otherwise, which gets very close
to 1 bit per block without having the pathological behavior when the map
gets near full.

> 
> > This negates the need for dedicated tracking
> > map or log. Fio supports this through random_generator=lfsr. The one
> > thing that does not YET work is lfsr and multiple block sizes. Should be
> > doable through layered use of multiple lfsrs. Something for your intern
> > to tackle?
> 
> Yes. That would be fair game.

Excellent! 

> >> o provide some "bread crumbs" for debugging when data is NOT correct.
> >>    (Not available typically will result in reported errors)
> >
> > So lets say that one of the fio verify modes was augmented to include a
> > time stamp (say the meta mode, or could even be added to all the verify
> > modes), that could be part of the bread crumbs and aid in judging
> > retention of the data.
> 
> I want four pieces of data for bread crumbs:
> o timestamp

Don't have that, trivial to add.

> o LBA written (e.g. if it's a partition, that means the offset into
> the partition)

Got it.

> o magic number for that test run (think of it as a GUID - verifies the
> block was written by fio)

Got it, but it's a fio generic magic. We could add a specific magic as
well, would be trivial.

> o generation number in the case that we rewrite an LBA - so we can
> detect stale data

Don't have that, trivial to add.

> I haven't checked if fio provides all four of those.
> 
> BTW, to eventually support adding "trim/discard command" testing into
> the mix, we would need to know when a block is explicitly unmapped and
> should be all zeros if we attempt to read it.

This possibly again could be done without on-the-side tracking, if we
used a separate lfsr to generate the read/write/trim part. This would
keep the memory foot print down. Fio does support trim already.

> >> o be done in < 6 weeks by a full time intern. :)
> >
> > Depends on the quality of the intern :-)
> 
> Juan is capable enough. He's also extremely persistent. So I think
> yes, he can get this done.

Excellent!

> >> Questions:
> >> 1) You know anyone else developing data integrity/retention testing with fio?
> >
> > I know of lots of people/companies using it for data integrity testing
> > (I even work for one of them), but not data retention. So I think that
> > would be a very interesting feature to add.
> 
> Good. thanks!
> 
> My review a few months ago of fio docs didn't give me the impression
> the data integrity checking was providing enough bread crumbs for good
> debugging or able to detect stale data. But my memory isn't very good
> and I could be wrong or just out of date.

Depends on your use case. Fio checksums the verify header separate. If
that is good, we can check the actual data. If that is not good, we can
recreate the original content and compare with what is on disk. That
gives you a pretty good idea of what was destroyed and how. But it does
not have the required bits for real retention testing, like timestamp
and/or sequence. That could be added to the verify_header structure, or
it could be a specific part of eg the meta verify. The latter has the
offset written already, for instance.

> >> 3) You have a preference on how this might be implemented if (a) we
> >> used code from OR (b) integrated this functionality into fio?
> >
> > I think the the data retention aspect should be integrated into the
> > verify modes. The fio verification modes checksum both the stored
> > header, as well as the actual contents. There might be additional
> > tracking required on the side for retention, to be able to pass some
> > interesting info on where we seem to fall off a cliff.
> 
> Ok. Let me clarify the requirement for data retention: I wanted
> "verify" to be an option to the "read" workload mix. So not
> necessarily all data that gets written will get verified "during" the
> write workload. The reason is performance statistics need to be as
> consistent as possible without "verify" in a mixed read/write
> workload.

Fio already supports that. Simply do the write workload with
do_verify=0, then do a similar read workload with do_verify=1 and the
same verify checksum etc settings.

> To verify everything that was written or trimmed, we can invoke fio
> again (think autotest invoking fio twice per test run) to check for
> retention. And then invoke fio many more times while the device is
> getting baked in a thermal chamber.

Trim verification can be done if the device supports persistent and
guaranteed zero return on a completed trim. trim_verify_zero. If that
isn't set, trimmed regions are simply ignored for a verify.

> > I'll be happy to work with you guys on this, both on the initial
> > design phase and the final integration into fio.
> 
> Awesome - thank you!
> 
> Design phase?! :) This is design phase. :)

Agree :-)

-- 
Jens Axboe

Re: FIO and Storage Data Integrity testing

[Jens Axboe]

On 07/31/2013 08:25 PM, Jens Axboe wrote:
>> I want four pieces of data for bread crumbs:
>> o timestamp
> 
> Don't have that, trivial to add.

Just went and looked at what we had for the meta verify. There is
actually already a timestamp there (sec + usec), and there's the notion
of a generation number as well (it includes what number write this was).

The structure looks like this:

struct vhdr_meta
{
        uint64_t offset;
        unsigned char thread;
        unsigned short numberio;
        unsigned long time_sec;
        unsigned long time_usec;
};

Though the 'generation' number might have to be either used differently
that it is now, or made into a different type. The meta verify is also
the only one that isn't already 32 vs 64-bit agnostic.

-- 
Jens Axboe

Re: FIO and Storage Data Integrity testing

[Juan Casse]

[-- Attachment #1: Type: text/plain, Size: 2999 bytes --]

Jens, Grant,

Bellow is a summary of the thread as I understood it.
(To keep it simple, I'm using the TODO keyword for both action items and
questions.)

*Requirements for data integrity and retention test*

   1. gpl implementaion
      - fio satisfies this
   2. Verify data written to storage is available and correct
      - fio has axmap to log writes, but
      - prefer linear feedback shift registers (lfsr) to reproduce data for
      verification; fio currently supports this with random_generator=lfsr
         - TODO: Jens says lfsr does not currently work with multiple block
         sizes? I'm missing something here; I thought a single run would have a
         single block size. Please explain.
      - Breadcrumbs for debugging data that is not correct
         - lba - fio has uint64 offset. OK.
         - generation # - fio has unsigned short numberio
            - TODO: Jens says must be used differently? How does it
            currently work? Is it also incremented and written to
storage upon a read?
         - timestamp - fio has unsigned long time_sec and time_usec. OK.
         - magic # - fio has a generic magic number
            - TODO: what does generic mean? is it always the same number
            dor every run of fio?
         - TODO: make this data structure (vhdr_meta) agnostic to 32- vs.
         64-bit
      3. Data retention
      - fio currently can do
         - write workload with do_verify=0
         - read workload with do_verify=1
      - fio currently checksums
         - header information
         - actual data
      4. Test trim/discard
      - TODO: use sepparate lfsr to generate read/wrte/trim part

Grant, if fio currently has lfsr and the breadcrumbs and can do writes with
do_verify=0 and reads with do_verify=1, it seems to me that fio already
satisfies our requirements for data integrity testing, contingent on Jens'
answers to my questions above. Also, wouldn't these features be sufficient
for data retention testing as well? If you write do_verify=0 and then 3
months later read do_verify=1, wouldn't that be enough?

Juan


On Thu, Aug 1, 2013 at 2:02 PM, Jens Axboe <axboe@kernel.dk> wrote:

> On 07/31/2013 08:25 PM, Jens Axboe wrote:
> >> I want four pieces of data for bread crumbs:
> >> o timestamp
> >
> > Don't have that, trivial to add.
>
> Just went and looked at what we had for the meta verify. There is
> actually already a timestamp there (sec + usec), and there's the notion
> of a generation number as well (it includes what number write this was).
>
> The structure looks like this:
>
> struct vhdr_meta
> {
>         uint64_t offset;
>         unsigned char thread;
>         unsigned short numberio;
>         unsigned long time_sec;
>         unsigned long time_usec;
> };
>
> Though the 'generation' number might have to be either used differently
> that it is now, or made into a different type. The meta verify is also
> the only one that isn't already 32 vs 64-bit agnostic.
>
> --
> Jens Axboe
>
>

[-- Attachment #2: Type: text/html, Size: 3663 bytes --]

Re: FIO and Storage Data Integrity testing

[Juan Casse]

[-- Attachment #1: Type: text/plain, Size: 2780 bytes --]

Jens, Grant,

Bellow is a summary of the thread as I understood it.
(To keep it simple, I'm using the TODO keyword for both action items and
questions.)

Requirements for data integrity and retention test

1) gpl implementaion:

fio satisfies this

2) Verify data written to storage is available and correct:

fio has axmap to log writes, but
prefer linear feedback shift registers (lfsr) to reproduce data for
verification; fio currently supports this with random_generator=lfsr

TODO: Jens says lfsr does not currently work with multiple block sizes? I'm
missing something here; I thought a single run would have a single block
size. Please explain.

3) Breadcrumbs for debugging data that is not correct:

a) lba - fio has uint64 offset. OK.
b) generation # - fio has unsigned short numberio

TODO: Jens says must be used differently? How does it currently work? Is it
also incremented and written to storage upon a read?

c) timestamp - fio has unsigned long time_sec and time_usec. OK.
d) magic # - fio has a generic magic number

TODO: what does generic mean? is it always the same number dor every run of
fio?

TODO: make this data structure (vhdr_meta) agnostic to 32- vs. 64-bit

4) Data retention:

fio currently can do:

a) write workload with do_verify=0
b) read workload with do_verify=1

fio currently checksums:

a) header information
b) actual data

5) Test trim/discard:

TODO: use sepparate lfsr to generate read/wrte/trim part

Grant, if fio currently has lfsr and the breadcrumbs and can do writes with
do_verify=0 and reads with do_verify=1, it seems to me that fio already
satisfies our requirements for data integrity testing, contingent on Jens'
answers to my questions above. Also, wouldn't these features be sufficient
for data retention testing as well? If you write do_verify=0 and then 3
months later read do_verify=1, wouldn't that be enough?

Juan



On Thu, Aug 1, 2013 at 2:02 PM, Jens Axboe <axboe@kernel.dk> wrote:

> On 07/31/2013 08:25 PM, Jens Axboe wrote:
> >> I want four pieces of data for bread crumbs:
> >> o timestamp
> >
> > Don't have that, trivial to add.
>
> Just went and looked at what we had for the meta verify. There is
> actually already a timestamp there (sec + usec), and there's the notion
> of a generation number as well (it includes what number write this was).
>
> The structure looks like this:
>
> struct vhdr_meta
> {
>         uint64_t offset;
>         unsigned char thread;
>         unsigned short numberio;
>         unsigned long time_sec;
>         unsigned long time_usec;
> };
>
> Though the 'generation' number might have to be either used differently
> that it is now, or made into a different type. The meta verify is also
> the only one that isn't already 32 vs 64-bit agnostic.
>
> --
> Jens Axboe
>
>

[-- Attachment #2: Type: text/html, Size: 3520 bytes --]

Re: FIO and Storage Data Integrity testing

[Juan Casse]

[-- Attachment #1: Type: text/plain, Size: 2783 bytes --]

Jens, Grant,

Bellow is a summary of the thread as I understood it.
(To keep it simple, I'm using the TODO keyword for both action items and
questions.)

Requirements for data integrity and retention test

1) gpl implementaion:

fio satisfies this

2) Verify data written to storage is available and correct:


fio has axmap to log writes, but
prefer linear feedback shift registers (lfsr) to reproduce data for
verification; fio currently supports this with random_generator=lfsr

TODO: Jens says lfsr does not currently work with multiple block sizes? I'm
missing something here; I thought a single run would have a single block
size. Please explain.

3) Breadcrumbs for debugging data that is not correct:

a) lba - fio has uint64 offset. OK.
b) generation # - fio has unsigned short numberio


TODO: Jens says must be used differently? How does it currently work? Is it
also incremented and written to storage upon a read?

c) timestamp - fio has unsigned long time_sec and time_usec. OK.
d) magic # - fio has a generic magic number


TODO: what does generic mean? is it always the same number dor every run of
fio?

TODO: make this data structure (vhdr_meta) agnostic to 32- vs. 64-bit

4) Data retention:

fio currently can do:

a) write workload with do_verify=0
b) read workload with do_verify=1

fio currently checksums:

a) header information
b) actual data

5) Test trim/discard:

TODO: use sepparate lfsr to generate read/wrte/trim part

Grant, if fio currently has lfsr and the breadcrumbs and can do writes with
do_verify=0 and reads with do_verify=1, it seems to me that fio already
satisfies our requirements for data integrity testing, contingent on Jens'
answers to my questions above. Also, wouldn't these features be sufficient
for data retention testing as well? If you write do_verify=0 and then 3
months later read do_verify=1, wouldn't that be enough?

Juan



On Thu, Aug 1, 2013 at 2:02 PM, Jens Axboe <axboe@kernel.dk> wrote:

> On 07/31/2013 08:25 PM, Jens Axboe wrote:
> >> I want four pieces of data for bread crumbs:
> >> o timestamp
> >
> > Don't have that, trivial to add.
>
> Just went and looked at what we had for the meta verify. There is
> actually already a timestamp there (sec + usec), and there's the notion
> of a generation number as well (it includes what number write this was).
>
> The structure looks like this:
>
> struct vhdr_meta
> {
>         uint64_t offset;
>         unsigned char thread;
>         unsigned short numberio;
>         unsigned long time_sec;
>         unsigned long time_usec;
> };
>
> Though the 'generation' number might have to be either used differently
> that it is now, or made into a different type. The meta verify is also
> the only one that isn't already 32 vs 64-bit agnostic.
>
> --
> Jens Axboe
>
>

[-- Attachment #2: Type: text/html, Size: 3976 bytes --]

Re: FIO and Storage Data Integrity testing

[Jens Axboe]

On 08/02/2013 03:10 PM, Juan Casse wrote:
> Jens, Grant,
> 
> Bellow is a summary of the thread as I understood it.
> (To keep it simple, I'm using the TODO keyword for both action items and
> questions.)
> 
> Requirements for data integrity and retention test
> 
> 1) gpl implementaion:
> 
> fio satisfies this
> 
> 2) Verify data written to storage is available and correct:
> 
> 
> fio has axmap to log writes, but
> prefer linear feedback shift registers (lfsr) to reproduce data for
> verification; fio currently supports this with random_generator=lfsr
> 
> TODO: Jens says lfsr does not currently work with multiple block sizes?
> I'm missing something here; I thought a single run would have a single
> block size. Please explain.

A single run CAN use a single block size, but it could also be a
multiple of block sizes. It all depends on how you configure the job. If
all you care about is single block size runs, then yes, then fio already
works with lfsr and that.

> 3) Breadcrumbs for debugging data that is not correct:
> 
> a) lba - fio has uint64 offset. OK.
> b) generation # - fio has unsigned short numberio
> 
> 
> TODO: Jens says must be used differently? How does it currently work? Is
> it also incremented and written to storage upon a read?

Depends on how you want to use the generation number. Right now it's
just a truncated variant of the number of writes issued.

> c) timestamp - fio has unsigned long time_sec and time_usec. OK.
> d) magic # - fio has a generic magic number
> 
> 
> TODO: what does generic mean? is it always the same number dor every run
> of fio?

Yes, it's a fixed constant that fio uses (oxacca). Could easily be made
configurable.

> TODO: make this data structure (vhdr_meta) agnostic to 32- vs. 64-bit

And bonus points for making it little/big endian safe too, we should do
that if we change it anyway. Should be little endian disk format. Fio
has the appropriate cpu_to_leXX etc helpers included, the client/server
protocol is 32/64 little/big endian agnostic.

> 4) Data retention:
> 
> fio currently can do:
> 
> a) write workload with do_verify=0
> b) read workload with do_verify=1
> 
> fio currently checksums:
> 
> a) header information
> b) actual data
> 
> 5) Test trim/discard:
> 
> TODO: use sepparate lfsr to generate read/wrte/trim part
> 
> Grant, if fio currently has lfsr and the breadcrumbs and can do writes
> with do_verify=0 and reads with do_verify=1, it seems to me that fio
> already satisfies our requirements for data integrity testing,
> contingent on Jens' answers to my questions above. Also, wouldn't these
> features be sufficient for data retention testing as well? If you write
> do_verify=0 and then 3 months later read do_verify=1, wouldn't that be
> enough?
> 
> Juan

-- 
Jens Axboe

Re: FIO and Storage Data Integrity testing

[Juan Casse]

[-- Attachment #1: Type: text/plain, Size: 8203 bytes --]

Jens, Grant,

Below is a summary of the thread as I understood it.
(To keep it simple, I'm using the TODO keyword for both action items and
questions.)

Requirements for data integrity and retention test

1) gpl implementaion:

fio satisfies this

2) Verify data written to storage is available and correct:

fio has axmap to log writes, but
prefer linear feedback shift registers (lfsr) to reproduce data for
verification; fio currently supports this with random_generator=lfsr

TODO: Jens says lfsr does not currently work with multiple block sizes? I'm
missing something here; I thought a single run would have a single block
size. Please explain.

3) Breadcrumbs for debugging data that is not correct:

a) lba - fio has uint64 offset. OK.
b) generation # - fio has unsigned short numberio

TODO: Jens says must be used differently? How does it currently work? Is it
also incremented and written to storage upon a read?

c) timestamp - fio has unsigned long time_sec and time_usec. OK.
d) magic # - fio has a generic magic number

TODO: what does generic mean? is it always the same number dor every run of
fio?

TODO: make this data structure (vhdr_meta) agnostic to 32- vs. 64-bit

4) Data retention:

fio currently can do:

a) write workload with do_verify=0
b) read workload with do_verify=1

fio currently checksums:

a) header information
b) actual data

5) Test trim/discard:

TODO: use separate lfsr to generate read/write/trim part

Grant, if fio currently has lfsr and the breadcrumbs and can do writes with
do_verify=0 and reads with do_verify=1, it seems to me that fio already
satisfies our requirements for data integrity testing, contingent on Jens'
answers to my questions above. Also, wouldn't these features be sufficient
for data retention testing as well? If you write do_verify=0 and then 3
months later read do_verify=1, wouldn't that be enough?

Juan



On Wed, Jul 31, 2013 at 7:25 PM, Jens Axboe <axboe@kernel.dk> wrote:

> On Wed, Jul 31 2013, Grant Grundler wrote:
> > On Wed, Jul 31, 2013 at 2:23 PM, Jens Axboe <axboe@kernel.dk> wrote:
> > ...
> > >> o log writes (save a map) so we can repeatedly verify writes at a
> > >> later date (weeks or months later)
> > >
> > > One approach that I have started favoring is instead providing coverage
> > > through the use of an lfsr.
> >
> > "Log File Structured <something>?" Expand that and I can better assess
> > if I think this is a good idea or not.
>
> Linear feedback shift registers. Basically a way to generate a "random"
> sequence of numbers that are guarenteed not to repeat until the cycle is
> repeated. Then you never have to do on-the-side tracking to avoid
> overlaps or overwrites.
>
> For verify, you simple re-set the seed to the value when the sequence
> started to hit all of the same blocks again. Or variations around that
> theme, if you only want some of them.
>
> > I'm not too picky about the implementation that meets the requirement.
> > I understand maps or sparse maps can get very awkward to handle for
> > devices with more than a few billion blocks. ("Can plz haz 4k blox?"
> > :)
>
> Indeed, fio has the "axmap" to track it otherwise, which gets very close
> to 1 bit per block without having the pathological behavior when the map
> gets near full.
>
> >
> > > This negates the need for dedicated tracking
> > > map or log. Fio supports this through random_generator=lfsr. The one
> > > thing that does not YET work is lfsr and multiple block sizes. Should
> be
> > > doable through layered use of multiple lfsrs. Something for your intern
> > > to tackle?
> >
> > Yes. That would be fair game.
>
> Excellent!
>
> > >> o provide some "bread crumbs" for debugging when data is NOT correct.
> > >>    (Not available typically will result in reported errors)
> > >
> > > So lets say that one of the fio verify modes was augmented to include a
> > > time stamp (say the meta mode, or could even be added to all the verify
> > > modes), that could be part of the bread crumbs and aid in judging
> > > retention of the data.
> >
> > I want four pieces of data for bread crumbs:
> > o timestamp
>
> Don't have that, trivial to add.
>
> > o LBA written (e.g. if it's a partition, that means the offset into
> > the partition)
>
> Got it.
>
> > o magic number for that test run (think of it as a GUID - verifies the
> > block was written by fio)
>
> Got it, but it's a fio generic magic. We could add a specific magic as
> well, would be trivial.
>
> > o generation number in the case that we rewrite an LBA - so we can
> > detect stale data
>
> Don't have that, trivial to add.
>
> > I haven't checked if fio provides all four of those.
> >
> > BTW, to eventually support adding "trim/discard command" testing into
> > the mix, we would need to know when a block is explicitly unmapped and
> > should be all zeros if we attempt to read it.
>
> This possibly again could be done without on-the-side tracking, if we
> used a separate lfsr to generate the read/write/trim part. This would
> keep the memory foot print down. Fio does support trim already.
>
> > >> o be done in < 6 weeks by a full time intern. :)
> > >
> > > Depends on the quality of the intern :-)
> >
> > Juan is capable enough. He's also extremely persistent. So I think
> > yes, he can get this done.
>
> Excellent!
>
> > >> Questions:
> > >> 1) You know anyone else developing data integrity/retention testing
> with fio?
> > >
> > > I know of lots of people/companies using it for data integrity testing
> > > (I even work for one of them), but not data retention. So I think that
> > > would be a very interesting feature to add.
> >
> > Good. thanks!
> >
> > My review a few months ago of fio docs didn't give me the impression
> > the data integrity checking was providing enough bread crumbs for good
> > debugging or able to detect stale data. But my memory isn't very good
> > and I could be wrong or just out of date.
>
> Depends on your use case. Fio checksums the verify header separate. If
> that is good, we can check the actual data. If that is not good, we can
> recreate the original content and compare with what is on disk. That
> gives you a pretty good idea of what was destroyed and how. But it does
> not have the required bits for real retention testing, like timestamp
> and/or sequence. That could be added to the verify_header structure, or
> it could be a specific part of eg the meta verify. The latter has the
> offset written already, for instance.
>
> > >> 3) You have a preference on how this might be implemented if (a) we
> > >> used code from OR (b) integrated this functionality into fio?
> > >
> > > I think the the data retention aspect should be integrated into the
> > > verify modes. The fio verification modes checksum both the stored
> > > header, as well as the actual contents. There might be additional
> > > tracking required on the side for retention, to be able to pass some
> > > interesting info on where we seem to fall off a cliff.
> >
> > Ok. Let me clarify the requirement for data retention: I wanted
> > "verify" to be an option to the "read" workload mix. So not
> > necessarily all data that gets written will get verified "during" the
> > write workload. The reason is performance statistics need to be as
> > consistent as possible without "verify" in a mixed read/write
> > workload.
>
> Fio already supports that. Simply do the write workload with
> do_verify=0, then do a similar read workload with do_verify=1 and the
> same verify checksum etc settings.
>
> > To verify everything that was written or trimmed, we can invoke fio
> > again (think autotest invoking fio twice per test run) to check for
> > retention. And then invoke fio many more times while the device is
> > getting baked in a thermal chamber.
>
> Trim verification can be done if the device supports persistent and
> guaranteed zero return on a completed trim. trim_verify_zero. If that
> isn't set, trimmed regions are simply ignored for a verify.
>
> > > I'll be happy to work with you guys on this, both on the initial
> > > design phase and the final integration into fio.
> >
> > Awesome - thank you!
> >
> > Design phase?! :) This is design phase. :)
>
> Agree :-)
>
> --
> Jens Axboe
>
>

[-- Attachment #2: Type: text/html, Size: 10565 bytes --]

Re: FIO and Storage Data Integrity testing

[Juan Casse]

Jens, Grant,

Below is a summary of the thread as I understood it.
(To keep it simple, I'm using the TODO keyword for both action items
and questions.)

Requirements for data integrity and retention test

1) gpl implementaion:

fio satisfies this

2) Verify data written to storage is available and correct:

fio has axmap to log writes, but
prefer linear feedback shift registers (lfsr) to reproduce data for
verification; fio currently supports this with random_generator=lfsr

TODO: Jens says lfsr does not currently work with multiple block
sizes? I'm missing something here; I thought a single run would have a
single block size. Please explain.

3) Breadcrumbs for debugging data that is not correct:

a) lba - fio has uint64 offset. OK.
b) generation # - fio has unsigned short numberio

TODO: Jens says must be used differently? How does it currently work?
Is it also incremented and written to storage upon a read?

c) timestamp - fio has unsigned long time_sec and time_usec. OK.
d) magic # - fio has a generic magic number

TODO: what does generic mean? is it always the same number dor every run of fio?

TODO: make this data structure (vhdr_meta) agnostic to 32- vs. 64-bit

4) Data retention:

fio currently can do:

a) write workload with do_verify=0
b) read workload with do_verify=1

fio currently checksums:

a) header information
b) actual data

5) Test trim/discard:

TODO: use separate lfsr to generate read/write/trim part

Grant, if fio currently has lfsr and the breadcrumbs and can do writes
with do_verify=0 and reads with do_verify=1, it seems to me that fio
already satisfies our requirements for data integrity testing,
contingent on Jens' answers to my questions above. Also, wouldn't
these features be sufficient for data retention testing as well? If
you write do_verify=0 and then 3 months later read do_verify=1,
wouldn't that be enough?

Juan

On Wed, Jul 31, 2013 at 7:25 PM, Jens Axboe <axboe@kernel.dk> wrote:
> On Wed, Jul 31 2013, Grant Grundler wrote:
>> On Wed, Jul 31, 2013 at 2:23 PM, Jens Axboe <axboe@kernel.dk> wrote:
>> ...
>> >> o log writes (save a map) so we can repeatedly verify writes at a
>> >> later date (weeks or months later)
>> >
>> > One approach that I have started favoring is instead providing coverage
>> > through the use of an lfsr.
>>
>> "Log File Structured <something>?" Expand that and I can better assess
>> if I think this is a good idea or not.
>
> Linear feedback shift registers. Basically a way to generate a "random"
> sequence of numbers that are guarenteed not to repeat until the cycle is
> repeated. Then you never have to do on-the-side tracking to avoid
> overlaps or overwrites.
>
> For verify, you simple re-set the seed to the value when the sequence
> started to hit all of the same blocks again. Or variations around that
> theme, if you only want some of them.
>
>> I'm not too picky about the implementation that meets the requirement.
>> I understand maps or sparse maps can get very awkward to handle for
>> devices with more than a few billion blocks. ("Can plz haz 4k blox?"
>> :)
>
> Indeed, fio has the "axmap" to track it otherwise, which gets very close
> to 1 bit per block without having the pathological behavior when the map
> gets near full.
>
>>
>> > This negates the need for dedicated tracking
>> > map or log. Fio supports this through random_generator=lfsr. The one
>> > thing that does not YET work is lfsr and multiple block sizes. Should be
>> > doable through layered use of multiple lfsrs. Something for your intern
>> > to tackle?
>>
>> Yes. That would be fair game.
>
> Excellent!
>
>> >> o provide some "bread crumbs" for debugging when data is NOT correct.
>> >>    (Not available typically will result in reported errors)
>> >
>> > So lets say that one of the fio verify modes was augmented to include a
>> > time stamp (say the meta mode, or could even be added to all the verify
>> > modes), that could be part of the bread crumbs and aid in judging
>> > retention of the data.
>>
>> I want four pieces of data for bread crumbs:
>> o timestamp
>
> Don't have that, trivial to add.
>
>> o LBA written (e.g. if it's a partition, that means the offset into
>> the partition)
>
> Got it.
>
>> o magic number for that test run (think of it as a GUID - verifies the
>> block was written by fio)
>
> Got it, but it's a fio generic magic. We could add a specific magic as
> well, would be trivial.
>
>> o generation number in the case that we rewrite an LBA - so we can
>> detect stale data
>
> Don't have that, trivial to add.
>
>> I haven't checked if fio provides all four of those.
>>
>> BTW, to eventually support adding "trim/discard command" testing into
>> the mix, we would need to know when a block is explicitly unmapped and
>> should be all zeros if we attempt to read it.
>
> This possibly again could be done without on-the-side tracking, if we
> used a separate lfsr to generate the read/write/trim part. This would
> keep the memory foot print down. Fio does support trim already.
>
>> >> o be done in < 6 weeks by a full time intern. :)
>> >
>> > Depends on the quality of the intern :-)
>>
>> Juan is capable enough. He's also extremely persistent. So I think
>> yes, he can get this done.
>
> Excellent!
>
>> >> Questions:
>> >> 1) You know anyone else developing data integrity/retention testing with fio?
>> >
>> > I know of lots of people/companies using it for data integrity testing
>> > (I even work for one of them), but not data retention. So I think that
>> > would be a very interesting feature to add.
>>
>> Good. thanks!
>>
>> My review a few months ago of fio docs didn't give me the impression
>> the data integrity checking was providing enough bread crumbs for good
>> debugging or able to detect stale data. But my memory isn't very good
>> and I could be wrong or just out of date.
>
> Depends on your use case. Fio checksums the verify header separate. If
> that is good, we can check the actual data. If that is not good, we can
> recreate the original content and compare with what is on disk. That
> gives you a pretty good idea of what was destroyed and how. But it does
> not have the required bits for real retention testing, like timestamp
> and/or sequence. That could be added to the verify_header structure, or
> it could be a specific part of eg the meta verify. The latter has the
> offset written already, for instance.
>
>> >> 3) You have a preference on how this might be implemented if (a) we
>> >> used code from OR (b) integrated this functionality into fio?
>> >
>> > I think the the data retention aspect should be integrated into the
>> > verify modes. The fio verification modes checksum both the stored
>> > header, as well as the actual contents. There might be additional
>> > tracking required on the side for retention, to be able to pass some
>> > interesting info on where we seem to fall off a cliff.
>>
>> Ok. Let me clarify the requirement for data retention: I wanted
>> "verify" to be an option to the "read" workload mix. So not
>> necessarily all data that gets written will get verified "during" the
>> write workload. The reason is performance statistics need to be as
>> consistent as possible without "verify" in a mixed read/write
>> workload.
>
> Fio already supports that. Simply do the write workload with
> do_verify=0, then do a similar read workload with do_verify=1 and the
> same verify checksum etc settings.
>
>> To verify everything that was written or trimmed, we can invoke fio
>> again (think autotest invoking fio twice per test run) to check for
>> retention. And then invoke fio many more times while the device is
>> getting baked in a thermal chamber.
>
> Trim verification can be done if the device supports persistent and
> guaranteed zero return on a completed trim. trim_verify_zero. If that
> isn't set, trimmed regions are simply ignored for a verify.
>
>> > I'll be happy to work with you guys on this, both on the initial
>> > design phase and the final integration into fio.
>>
>> Awesome - thank you!
>>
>> Design phase?! :) This is design phase. :)
>
> Agree :-)
>
> --
> Jens Axboe
>

Re: FIO and Storage Data Integrity testing

[Grant Grundler]

On Fri, Aug 2, 2013 at 2:21 PM, Juan Casse <jcasse@google.com> wrote:
...
> Grant, if fio currently has lfsr and the breadcrumbs and can do writes
> with do_verify=0 and reads with do_verify=1, it seems to me that fio
> already satisfies our requirements for data integrity testing,

That's almost true. :)

> contingent on Jens' answers to my questions above. Also, wouldn't
> these features be sufficient for data retention testing as well?

Yes, they will be once you implement them. E.g. numberio is not a
generation number. Given the LFSR seed, one could determine the
generation number. So perhaps add the LSFR seed value to the data
structure so that gets written with each test run many many times too.
Later, when we validate the drive, we can determine generation number
and confirm the value is correct that was read from media. All of this
needs to be implemented.

Carefully re-read Jen's replies. Some of his answers are "speculative"
(e.g. "could do this" vs "does this").

> If you write do_verify=0 and then 3 months later read do_verify=1,
> wouldn't that be enough?

Correct - That's the goal. It's not reality yet. :)

cheers,
grant