speeding up disk-based laptop by moving /usr to flash?

[gaf-mNDKBlG2WHs at public.gmane.org (Jerry Feldman)]

I just want to elaborate a bit further. Once you install a new battery,
you will be able to suspend so that you are booting your system less
frequently.

On 12/02/2009 07:51 AM, Jerry Feldman wrote:
> I don't think that would affect performance, and may even hurt
> performance. Remember that when Linux loads a program or a shared
> library, the location of that program or shared object effectively
> becomes part of swap. The program and its dependent shared libraries ar=
e
> mapped into memory, but only loaded into pages when needed.
>
> Boot time is really a few stages, first, the kernel is loaded and singl=
e
> user mode is started, so a few shared libraries are needed. Then the
> transition into multi-user mode. Lots of things are done at this point,=

> and there are some bottlenecks.
>
> Also remember that the amount of available RAM affects performance
> probably more than anything else.
>
> What I would do is to look at possibly replacing the X41 since prices
> are now very low. Most new laptops today are built for Windows 7 which
> requires more memory so those laptops have more memory builtin.
>
> But, first, analyze your boot time by looking at the time each step tak=
es.
>
> If you want to keep the X41, what I would do would be to get a new
> battery, and if need be a new hard drive, and maximize the memory.
>
> On 12/02/2009 12:19 AM, Brendan Kidwell wrote:
>  =20
>> At the last BLU meeting I was thinking out loud about buying netbook, =
but
>> I'm leaning away from that now. I think someone mentioned here "the ke=
yboard
>> is just too small," and I'm starting to think the lack of 4:3 aspect r=
atio
>> netbooks is really a dealbreaker for me. I need my 768 lines.
>>
>> I currently use Ubuntu 9.10 on a Thinkpad X41 Tablet.
>>
>> I've thought about replacing my hard drive with an SSD device, but I'm=

>> reading a lot of FUD that's giving me... well FUD. It seems that the c=
heap
>> SSD devices are unreliable and of unknown origin (a lot like CD and DV=
D
>> blanks) and unfortunately have a tendency to NOT report read and write=

>> errors. Eek! Don't put your data there. And premium SSD devices are ju=
st too
>> expensive to justify for a home user.
>>
>> So here's a possible plan I conceived tonight: My goals are to maximiz=
e the
>> perceived performance of my Thinkpad as well as maximizing battery lif=
e.
>>
>> 1. Buy a fresh battery. (No question about this step; the current one =
is
>> weak and years old.)
>> 2. Keep the 60GB parallel ATA (IDE) hard drive and keep most of my stu=
ff on
>> it.
>> 3. Install a 16GB "133x" (20MB - 25MB per second) CF card in a PCMCIA-=
CF
>> adapter in my PCMCIA slot. -- I never use the slot for anything else.
>> 4. Move /usr to the CF card and mount it from there with "no access ti=
me
>> recording" switch. Maybe also /var and /etc.
>> 5. Maybe try putting /tmp in a RAM disk. (See what uses /tmp first and=

>> determine if it's worth carving out the block of RAM.)
>>
>> Most of the parts of the OS that don't live in /usr and /etc and /var =
look
>> like they're small enough that moving them off the disk isn't worth it=
, and
>> at this point I think I want to keep my data on the disk.
>>
>> I'm guessing this will improve my boot time as well as startup time fo=
r
>> applications by virtual of the fact that flash memory generally has mu=
ch
>> faster seek time than a disk.
>>
>> My concerns are:
>>
>> - Will keeping a CF card in a PCMCIA adapter online consume a lot of
>> additional power?
>> - Will the maximum throughput of the CF card be so slow as to overshad=
ow the
>> expected reduction in seek time?
>>
>> Has anyone ever tried anything like this?
>>
>> Would I get different performance from using an SD card in the SD slot=
? From
>> what I read tonight it seems the "Secure Digital" interface has even m=
ore
>> limited bandwidth than a typical CF card.
>>  =20
>>    =20

--=20
Jerry Feldman <gaf-mNDKBlG2WHs at public.gmane.org>
Boston Linux and Unix
PGP key id: 537C5846
PGP Key fingerprint: 3D1B 8377 A3C0 A5F2 ECBB  CA3B 4607 4319 537C 5846