Convert btrfs software code to ASIC

Convert btrfs software code to ASIC

[Le Nguyen Tran]

Hi,

I am Nguyen. I am not a software development engineer but an IC (chip)
development engineer. I have a plan to develop an IC controller for
Network Attached Storage (NAS). The main idea is converting software
code into hardware implementation. Because the chip is customized for
NAS, its performance is high, and its cost is lower than using micro
processor like Atom or Xeon (for servers).

I plan to use btrfs as the file system specification for my NAS. The
main point is that I need to understand the btrfs sofware code in
order to covert them into hardware implementation. I am wandering if
any of you can help me. If we can make the chip in a good shape, we
can start up a company and have our own business.

If you are interested in my idea and have further questions, please
send me an email: lntran.de@gmail.com

Thanks.

Nguyen.

Re: Convert btrfs software code to ASIC

[Fajar A. Nugraha]

On Mon, May 19, 2014 at 3:40 PM, Le Nguyen Tran <lntran.de@gmail.com> wrote:
> Hi,
>
> I am Nguyen. I am not a software development engineer but an IC (chip)
> development engineer. I have a plan to develop an IC controller for
> Network Attached Storage (NAS). The main idea is converting software
> code into hardware implementation. Because the chip is customized for
> NAS, its performance is high, and its cost is lower than using micro
> processor like Atom or Xeon (for servers).
>
> I plan to use btrfs as the file system specification for my NAS. The
> main point is that I need to understand the btrfs sofware code in
> order to covert them into hardware implementation. I am wandering if
> any of you can help me. If we can make the chip in a good shape, we
> can start up a company and have our own business.

I'm not sure if that's a good idea.

AFAIK btrfs depends a lot on other linux subsystems (e.g. vfs, block,
etc). Rather than converting/reimplementing everything, if your aim is
lower cost, you might have easier time using something like a mediatek
SOC (the ones used on smartphones) and run a custom-built linux with
btrfs support on it.

For documentation,
https://btrfs.wiki.kernel.org/index.php/Main_Page#Developer_documentation
is probably the best place to start

-- 
Fajar

Re: Convert btrfs software code to ASIC

[Le Nguyen Tran]

Hi Nugraha,

Thank you so much for your information. Frankly speaking, no one can
confirm a new start-up idea works or not. The probability of failure
is always high. However, the benefit if it works is also very high.

I do not plan to exactly replicate the C source code. There are always
some techniques in ASIC design to implement which are not the same as
in software (less flexible but faster).

The main advantages of my proposed chip are:
- Very high performance: Performance of ASIC chip is normally more
than 10x higher than performance of processors because processor run
only 1-4 instructions sequentially. That is very suitable for server
when there are many requests from users.
- Low-cost: In side the chip, we can customized for our function only.
In my plan, we do not need cache (which covers a very large area), and
we can use low cost technology 0.18um.
- Low-power: Processors run instructions sequentially and access
memory ( or cache). As a result, they consume much more power than
ASIC chip (also can be 10x higher).

Actually ARM processors like mediatek cannot be comparable with ASIC
chip. However, as I mentioned, it is just my draft idea. I still to
work more to verify my idea.

Thanks.

Nguyen.

On 5/19/14, Fajar A. Nugraha <list@fajar.net> wrote:
> On Mon, May 19, 2014 at 3:40 PM, Le Nguyen Tran <lntran.de@gmail.com>
> wrote:
>> Hi,
>>
>> I am Nguyen. I am not a software development engineer but an IC (chip)
>> development engineer. I have a plan to develop an IC controller for
>> Network Attached Storage (NAS). The main idea is converting software
>> code into hardware implementation. Because the chip is customized for
>> NAS, its performance is high, and its cost is lower than using micro
>> processor like Atom or Xeon (for servers).
>>
>> I plan to use btrfs as the file system specification for my NAS. The
>> main point is that I need to understand the btrfs sofware code in
>> order to covert them into hardware implementation. I am wandering if
>> any of you can help me. If we can make the chip in a good shape, we
>> can start up a company and have our own business.
>
> I'm not sure if that's a good idea.
>
> AFAIK btrfs depends a lot on other linux subsystems (e.g. vfs, block,
> etc). Rather than converting/reimplementing everything, if your aim is
> lower cost, you might have easier time using something like a mediatek
> SOC (the ones used on smartphones) and run a custom-built linux with
> btrfs support on it.
>
> For documentation,
> https://btrfs.wiki.kernel.org/index.php/Main_Page#Developer_documentation
> is probably the best place to start
>
> --
> Fajar
>

RE: Convert btrfs software code to ASIC

[Paul Jones]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset="utf-8", Size: 3803 bytes --]

Hi Nguyen,
 Perhaps a better idea would be to use a low-cost low-power som module to run Linux and btrfs code, and use an FPGA/ASIC to offload compression/encryption/checksums and to possibly act as a raid controller. Since btrfs will be under heavy development for the foreseeable future I doubt it would be a good idea to lock it into silicon. Using this approach the mature technologies can be hardware accelerated, and the software parts are available for easy upgrades.
It also significantly reduces risk for your project, and VCs like that sort of thing!

Regards,
Paul.

-----Original Message-----
From: linux-btrfs-owner@vger.kernel.org [mailto:linux-btrfs-owner@vger.kernel.org] On Behalf Of Le Nguyen Tran
Sent: Monday, 19 May 2014 9:07 PM
To: Fajar A. Nugraha
Cc: linux-btrfs
Subject: Re: Convert btrfs software code to ASIC

Hi Nugraha,

Thank you so much for your information. Frankly speaking, no one can confirm a new start-up idea works or not. The probability of failure is always high. However, the benefit if it works is also very high.

I do not plan to exactly replicate the C source code. There are always some techniques in ASIC design to implement which are not the same as in software (less flexible but faster).

The main advantages of my proposed chip are:
- Very high performance: Performance of ASIC chip is normally more than 10x higher than performance of processors because processor run only 1-4 instructions sequentially. That is very suitable for server when there are many requests from users.
- Low-cost: In side the chip, we can customized for our function only.
In my plan, we do not need cache (which covers a very large area), and we can use low cost technology 0.18um.
- Low-power: Processors run instructions sequentially and access memory ( or cache). As a result, they consume much more power than ASIC chip (also can be 10x higher).

Actually ARM processors like mediatek cannot be comparable with ASIC chip. However, as I mentioned, it is just my draft idea. I still to work more to verify my idea.

Thanks.

Nguyen.

On 5/19/14, Fajar A. Nugraha <list@fajar.net> wrote:
> On Mon, May 19, 2014 at 3:40 PM, Le Nguyen Tran <lntran.de@gmail.com>
> wrote:
>> Hi,
>>
>> I am Nguyen. I am not a software development engineer but an IC 
>> (chip) development engineer. I have a plan to develop an IC 
>> controller for Network Attached Storage (NAS). The main idea is 
>> converting software code into hardware implementation. Because the 
>> chip is customized for NAS, its performance is high, and its cost is 
>> lower than using micro processor like Atom or Xeon (for servers).
>>
>> I plan to use btrfs as the file system specification for my NAS. The 
>> main point is that I need to understand the btrfs sofware code in 
>> order to covert them into hardware implementation. I am wandering if 
>> any of you can help me. If we can make the chip in a good shape, we 
>> can start up a company and have our own business.
>
> I'm not sure if that's a good idea.
>
> AFAIK btrfs depends a lot on other linux subsystems (e.g. vfs, block, 
> etc). Rather than converting/reimplementing everything, if your aim is 
> lower cost, you might have easier time using something like a mediatek 
> SOC (the ones used on smartphones) and run a custom-built linux with 
> btrfs support on it.
>
> For documentation,
> https://btrfs.wiki.kernel.org/index.php/Main_Page#Developer_documentat
> ion
> is probably the best place to start
>
> --
> Fajar
>
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at  http://vger.kernel.org/majordomo-info.html
ÿôèº{.nÇ+‰·Ÿ®‰­†+%ŠËÿ±éݶ\x17¥Šwÿº{.nÇ+‰·¥Š{±ý»k~ÏâžØ^n‡r¡ö¦zË\x1aëh™¨è­Ú&£ûàz¿äz¹Þ—ú+€Ê+zf£¢·hšˆ§~†­†Ûiÿÿïêÿ‘êçz_è®\x0fæj:+v‰¨þ)ߣøm

Re: Convert btrfs software code to ASIC

[Duncan]

Paul Jones posted on Mon, 19 May 2014 12:24:53 +0000 as excerpted:

>> On Mon, May 19, 2014 at 3:40 PM, Le Nguyen Tran <lntran.de@gmail.com>
>> wrote:
>>>
>>> I have a plan to develop an IC controller for Network Attached
>>> Storage (NAS). The main idea is converting software code into
>>> hardware implementation.
>>>
>>> I plan to use btrfs as the file system specification for my NAS.

> Perhaps a better idea would be to use a low-cost low-power som module
> to run Linux and btrfs code, and use an FPGA/ASIC to offload
> compression/encryption/checksums and to possibly act as a raid
> controller. Since btrfs will be under heavy development for the
> foreseeable future I doubt it would be a good idea to lock it into
> silicon. Using this approach the mature technologies can be hardware
> accelerated, and the software parts are available for easy upgrades.
> It also significantly reduces risk for your project, and VCs like that
> sort of thing!

This is a very good idea and what I was about to suggest.  Certainly, 
btrfs is still not fully stable, and I really would hate to see the 
current implementation etched in silicon at this time.  However, a hybrid 
approach where the mature bits such as (de-/)compression/checksums/
encryption are hardware etched/accelerated while the more general and 
still developing code is deployed as upgradeable firmware on a system-on-
module sounds like a very good idea indeed, particularly if that firmware 
is deployed as a user-modifiable/replaceable free-as-in-freedom kernel in 
keeping with the spirit of the GPL under which the Linux kernel and thus 
btrfs are written.

In other words... I doubt very much that any list regular here familiar 
with the continuing flow of bugs we see, as well as the roadmapped but 
not yet implemented features that people wanting a hardware 
implementation would certainly be interested in, would find the idea of a 
hardware implementation of anything like current code anything but 
nightmare material. =:^\  Maybe in a couple years... but even then, 
upgradeable firmware with critical mature bits offloaded for hardware 
acceleration sounds like a far better idea.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

Re: Convert btrfs software code to ASIC

[Le Nguyen Tran]

Hi Paul,

Thank you for your advice. Actually, I currently have ideas to
implement database management (like list, tree), and dynamic memory
allocation in hardware to accelerate the file system operations. I
still do not have a clear picture about which part is implemented by
processor (as your advice) and which part is accelerated by hardware.
I now need to understand the operation of btrfs source code to
determine. I hope that one of you can help me and if it work, we can
start-up our own business.

Thanks.

Nguyen.

On 5/19/14, Paul Jones <paul@pauljones.id.au> wrote:
> Hi Nguyen,
>  Perhaps a better idea would be to use a low-cost low-power som module to
> run Linux and btrfs code, and use an FPGA/ASIC to offload
> compression/encryption/checksums and to possibly act as a raid controller.
> Since btrfs will be under heavy development for the foreseeable future I
> doubt it would be a good idea to lock it into silicon. Using this approach
> the mature technologies can be hardware accelerated, and the software parts
> are available for easy upgrades.
> It also significantly reduces risk for your project, and VCs like that sort
> of thing!
>
> Regards,
> Paul.
>
> -----Original Message-----
> From: linux-btrfs-owner@vger.kernel.org
> [mailto:linux-btrfs-owner@vger.kernel.org] On Behalf Of Le Nguyen Tran
> Sent: Monday, 19 May 2014 9:07 PM
> To: Fajar A. Nugraha
> Cc: linux-btrfs
> Subject: Re: Convert btrfs software code to ASIC
>
> Hi Nugraha,
>
> Thank you so much for your information. Frankly speaking, no one can confirm
> a new start-up idea works or not. The probability of failure is always high.
> However, the benefit if it works is also very high.
>
> I do not plan to exactly replicate the C source code. There are always some
> techniques in ASIC design to implement which are not the same as in software
> (less flexible but faster).
>
> The main advantages of my proposed chip are:
> - Very high performance: Performance of ASIC chip is normally more than 10x
> higher than performance of processors because processor run only 1-4
> instructions sequentially. That is very suitable for server when there are
> many requests from users.
> - Low-cost: In side the chip, we can customized for our function only.
> In my plan, we do not need cache (which covers a very large area), and we
> can use low cost technology 0.18um.
> - Low-power: Processors run instructions sequentially and access memory ( or
> cache). As a result, they consume much more power than ASIC chip (also can
> be 10x higher).
>
> Actually ARM processors like mediatek cannot be comparable with ASIC chip.
> However, as I mentioned, it is just my draft idea. I still to work more to
> verify my idea.
>
> Thanks.
>
> Nguyen.
>
> On 5/19/14, Fajar A. Nugraha <list@fajar.net> wrote:
>> On Mon, May 19, 2014 at 3:40 PM, Le Nguyen Tran <lntran.de@gmail.com>
>> wrote:
>>> Hi,
>>>
>>> I am Nguyen. I am not a software development engineer but an IC
>>> (chip) development engineer. I have a plan to develop an IC
>>> controller for Network Attached Storage (NAS). The main idea is
>>> converting software code into hardware implementation. Because the
>>> chip is customized for NAS, its performance is high, and its cost is
>>> lower than using micro processor like Atom or Xeon (for servers).
>>>
>>> I plan to use btrfs as the file system specification for my NAS. The
>>> main point is that I need to understand the btrfs sofware code in
>>> order to covert them into hardware implementation. I am wandering if
>>> any of you can help me. If we can make the chip in a good shape, we
>>> can start up a company and have our own business.
>>
>> I'm not sure if that's a good idea.
>>
>> AFAIK btrfs depends a lot on other linux subsystems (e.g. vfs, block,
>> etc). Rather than converting/reimplementing everything, if your aim is
>> lower cost, you might have easier time using something like a mediatek
>> SOC (the ones used on smartphones) and run a custom-built linux with
>> btrfs support on it.
>>
>> For documentation,
>> https://btrfs.wiki.kernel.org/index.php/Main_Page#Developer_documentat
>> ion
>> is probably the best place to start
>>
>> --
>> Fajar
>>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org More majordomo info at
> http://vger.kernel.org/majordomo-info.html
>

Re: Convert btrfs software code to ASIC

[Fajar A. Nugraha]

On Mon, May 19, 2014 at 8:09 PM, Le Nguyen Tran <lntran.de@gmail.com> wrote:
> I now need to understand the operation of btrfs source code to
> determine. I hope that one of you can help me


Have you read the wiki link?

-- 
Fajar