RAID0 - cache

Hi guys, from a long email i will start this one: (the last one wasn=B4=
t send)

think about this:

/dev/sda (sata disk 1000gb size, latency of 1ms to move head, 100mb/s
to write/read)
/dev/sdb (ssd 1gb size, latency of <100us to find what flash and
position, 500mb/s to write/read)
ram memory (8gb size, with 1gb not used)
/dev/mdxxx (the new device i=B4m calling it as raid0-cache)

how raid0-cache work?


how to create raid?
tell what the sequencial read cache estrategy:
memory(tell fixed memory size) - /dev/sdb - /dev/sda - another device
(maybe just 2 is suficiente /dev/sdb and /dev/sda)...


how to read?
first, read from the first read device (memory)
if information not there -> read from second read device (/dev/sdb),
update information at first read device, return the read value
if information not there -> read from third read device (/dev/sda),
update information at second and first read device, return the read
value

how to write?
flush space from first read device to second read device
write to first read device, return OK

background thread:
sync from first write device (memory) to second write device (/dev/sdb)
flush from second write device (/dev/sdb) to third write device (/dev/s=
da)

startup:
using cache:
flush everythink from first write device (memory) to second write
device (/dev/sdb)
flush everythink from second write device (/dev/sdb) to second write
device (/dev/sda)
since memory is volatille, at startup it can=B4be flushed...
without cache:
make first write device unusable (memory)
make second write device unusable
here we can have write problems, since a wrong power off, can lost
information from cache (first and second write device)



anyidea who can make it?
thanks
--
Roberto Spadim
Spadim Technology / SPAEmpresarial
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" i=
n
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: RAID0 - cache

On Wed, 19 Jan 2011 18:13:23 -0200 Roberto Spadim r>
wrote:

> Hi guys, from a long email i will start this one: (the last one wasn=B4=
t send)
>=20
> think about this:
>=20
> /dev/sda (sata disk 1000gb size, latency of 1ms to move head, 100mb/s
> to write/read)
> /dev/sdb (ssd 1gb size, latency of <100us to find what flash and
> position, 500mb/s to write/read)
> ram memory (8gb size, with 1gb not used)
> /dev/mdxxx (the new device i=B4m calling it as raid0-cache)
>=20
> how raid0-cache work?
>=20
>=20
> how to create raid?
> tell what the sequencial read cache estrategy:
> memory(tell fixed memory size) - /dev/sdb - /dev/sda - another device
> (maybe just 2 is suficiente /dev/sdb and /dev/sda)...
>=20
>=20
> how to read?
> first, read from the first read device (memory)
> if information not there -> read from second read device (/dev/sdb),
> update information at first read device, return the read value
> if information not there -> read from third read device (/dev/sda),
> update information at second and first read device, return the read
> value
>=20
> how to write?
> flush space from first read device to second read device
> write to first read device, return OK
>=20
> background thread:
> sync from first write device (memory) to second write device (/dev/sd=
b)
> flush from second write device (/dev/sdb) to third write device (/dev=
/sda)
>=20
> startup:
> using cache:
> flush everythink from first write device (memory) to second write
> device (/dev/sdb)
> flush everythink from second write device (/dev/sdb) to second write
> device (/dev/sda)
> since memory is volatille, at startup it can=B4be flushed...
> without cache:
> make first write device unusable (memory)
> make second write device unusable
> here we can have write problems, since a wrong power off, can lost
> information from cache (first and second write device)
>=20
>=20
>=20
> anyidea who can make it?

"Hey, I've got this clever idea, I wish someone would implement it" alm=
ost
always results in ... nothing.
People with the ability to implement are much more interested in implem=
enting
their own ideas.

But see also "bcache".

NeilBrown

> thanks
> --
> Roberto Spadim
> Spadim Technology / SPAEmpresarial
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid"=
in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe linux-raid" i=
n
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: RAID0 - cache

there=B4s a forum (linux kernel) for ideas?

2011/1/19 NeilBrown :
> On Wed, 19 Jan 2011 18:13:23 -0200 Roberto Spadim br>
> wrote:
>
>> Hi guys, from a long email i will start this one: (the last one wasn=
=B4t send)
>>
>> think about this:
>>
>> /dev/sda (sata disk 1000gb size, latency of 1ms to move head, 100mb/=
s
>> to write/read)
>> /dev/sdb (ssd 1gb size, latency of <100us to find what flash and
>> position, 500mb/s to write/read)
>> ram memory (8gb size, with 1gb not used)
>> /dev/mdxxx (the new device i=B4m calling it as raid0-cache)
>>
>> how raid0-cache work?
>>
>>
>> how to create raid?
>> tell what the sequencial read cache estrategy:
>> memory(tell fixed memory size) - /dev/sdb - /dev/sda - another devic=
e
>> (maybe just 2 is suficiente /dev/sdb and /dev/sda)...
>>
>>
>> how to read?
>> first, read from the first read device (memory)
>> if information not there -> read from second read device (/dev/sdb),
>> update information at first read device, return the read value
>> if information not there -> read from third read device (/dev/sda),
>> update information at second and first read device, return the read
>> value
>>
>> how to write?
>> flush space from first read device to second read device
>> write to first read device, return OK
>>
>> background thread:
>> sync from first write device (memory) to second write device (/dev/s=
db)
>> flush from second write device (/dev/sdb) to third write device (/de=
v/sda)
>>
>> startup:
>> using cache:
>> =A0flush everythink from first write device (memory) to second write
>> device (/dev/sdb)
>> =A0flush everythink from second write device (/dev/sdb) to second wr=
ite
>> device (/dev/sda)
>> =A0since memory is volatille, at startup it can=B4be flushed...
>> without cache:
>> =A0make first write device unusable (memory)
>> =A0make second write device unusable
>> =A0here we can have write problems, since a wrong power off, can los=
t
>> information from cache (first and second write device)
>>
>>
>>
>> anyidea who can make it?
>
> "Hey, I've got this clever idea, I wish someone would implement it" a=
lmost
> always results in ... nothing.
> People with the ability to implement are much more interested in impl=
ementing
> their own ideas.
>
> But see also "bcache".
>
> NeilBrown
>
>> thanks
>> --
>> Roberto Spadim
>> Spadim Technology / SPAEmpresarial
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-raid=
" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at =A0http://vger.kernel.org/majordomo-info.html
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid"=
in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at =A0http://vger.kernel.org/majordomo-info.html
>



--=20
Roberto Spadim
Spadim Technology / SPAEmpresarial
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" i=
n
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: RAID0 - cache

very nice i think that bcache is the solution
another question (ok i know...) could we improve read/write performace
on raid1 for ssd? round robin is better than near head (since ssd
don=B4t have head) i think there=B4s a patch for it just test it and pu=
t
at linux kernel (ok it will have a time to be on stable linux kernel)
thanks neil!

2011/1/19 Roberto Spadim :
> there=B4s a forum (linux kernel) for ideas?
>
> 2011/1/19 NeilBrown :
>> On Wed, 19 Jan 2011 18:13:23 -0200 Roberto Spadim m.br>
>> wrote:
>>
>>> Hi guys, from a long email i will start this one: (the last one was=
n=B4t send)
>>>
>>> think about this:
>>>
>>> /dev/sda (sata disk 1000gb size, latency of 1ms to move head, 100mb=
/s
>>> to write/read)
>>> /dev/sdb (ssd 1gb size, latency of <100us to find what flash and
>>> position, 500mb/s to write/read)
>>> ram memory (8gb size, with 1gb not used)
>>> /dev/mdxxx (the new device i=B4m calling it as raid0-cache)
>>>
>>> how raid0-cache work?
>>>
>>>
>>> how to create raid?
>>> tell what the sequencial read cache estrategy:
>>> memory(tell fixed memory size) - /dev/sdb - /dev/sda - another devi=
ce
>>> (maybe just 2 is suficiente /dev/sdb and /dev/sda)...
>>>
>>>
>>> how to read?
>>> first, read from the first read device (memory)
>>> if information not there -> read from second read device (/dev/sdb)=
,
>>> update information at first read device, return the read value
>>> if information not there -> read from third read device (/dev/sda),
>>> update information at second and first read device, return the read
>>> value
>>>
>>> how to write?
>>> flush space from first read device to second read device
>>> write to first read device, return OK
>>>
>>> background thread:
>>> sync from first write device (memory) to second write device (/dev/=
sdb)
>>> flush from second write device (/dev/sdb) to third write device (/d=
ev/sda)
>>>
>>> startup:
>>> using cache:
>>> =A0flush everythink from first write device (memory) to second writ=
e
>>> device (/dev/sdb)
>>> =A0flush everythink from second write device (/dev/sdb) to second w=
rite
>>> device (/dev/sda)
>>> =A0since memory is volatille, at startup it can=B4be flushed...
>>> without cache:
>>> =A0make first write device unusable (memory)
>>> =A0make second write device unusable
>>> =A0here we can have write problems, since a wrong power off, can lo=
st
>>> information from cache (first and second write device)
>>>
>>>
>>>
>>> anyidea who can make it?
>>
>> "Hey, I've got this clever idea, I wish someone would implement it" =
almost
>> always results in ... nothing.
>> People with the ability to implement are much more interested in imp=
lementing
>> their own ideas.
>>
>> But see also "bcache".
>>
>> NeilBrown
>>
>>> thanks
>>> --
>>> Roberto Spadim
>>> Spadim Technology / SPAEmpresarial
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-rai=
d" in
>>> the body of a message to majordomo@vger.kernel.org
>>> More majordomo info at =A0http://vger.kernel.org/majordomo-info.htm=
l
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-raid=
" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at =A0http://vger.kernel.org/majordomo-info.html
>>
>
>
>
> --
> Roberto Spadim
> Spadim Technology / SPAEmpresarial
>



--=20
Roberto Spadim
Spadim Technology / SPAEmpresarial
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" i=
n
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html