> On Oct 30, 2014, at 3:06 PM, Jeppe Øland <[email protected]> wrote:
> 
> On Thu, Oct 30, 2014 at 8:33 AM, Jim Thompson <[email protected]> wrote:
>>> On the other hand, I tend to distrust manufacturers that shipped
>>> completely unreliable drives without any thought.
>>> Kingston/OCZ/Crucial are all in this boat for me.
>> 
>> I’m sure I’ve been burned at least as badly by these, and others, and I
>> still buy from them.
> 
> What can you do? The speed increase from SSDs in a PC means its almost
> impossible to go back to an HDD.
> And in a firewall/appliance, the benefits from no moving parts/lower
> power/heat/noise is hard to ignore.
> 
>>> As for Nano, I thought it mounted almost everything as RO and only
>>> changed settings to write down settings changes, and RRD databases etc
>>> on reboots?
>> 
>> I think I’ve already responded to this.
>> 
>> nano is a > 10 year old “solution” to the problems that existed at the time.
>> http://markmail.org/message/rxe4xfpmdwva7q3e
>> 
>> That doesn’t mean it’s a bad solution, but though it’s author is a brilliant
>> individual, he obviously didn’t envision SSD in 2004.
> 
> Are you saying the "nano" release only covers the boot-slices?

See how the there are three partitions in the below?  Observe the sizes 
(“922257 sectors”) of the first two.

$ file pfSense-2.1.5-RELEASE-1g-amd64-nanobsd-20140825-0744.img 
pfSense-2.1.5-RELEASE-1g-amd64-nanobsd-20140825-0744.img: x86 boot sector; 
partition 1: ID=0xa5, active, starthead 1, startsector 63, 922257 sectors; 
partition 2: ID=0xa5, starthead 1, startsector 922383, 922257 sectors; 
partition 3: ID=0xa5, starthead 0, startsector 1844640, 102816 sectors, code 
offset 0x31

> I thought the nano/embedded versions also write less to the disk.
> I don't have a full install handy to check, but the nano install
> definitely mounts the drive RO, and all runtime stuff (/var, /tmp) is
> run out of RAM disks.

Yes, and I am aware of the differences with the “nano” builds.

CF devices don’t have the same type of sophisticated wear-leveling and virtual 
block remapping that modern SSDs and eMMC devices have.

Yes, I am saying that compression (which on a modern 64-bit Intel / AMD CPU is 
way faster than disk I/O (yes, even to a SSD)) and making those sectors 
available to the drive has
potentially far greater impact than the crippled nature of the “nano/embedded” 
version.

We’re not changing this for pfSense software version 2.2, but you can bet 
$CURRENCY to $SNACK_FOOD that it’s being evaluated and tested for something 
subsequent.  


_______________________________________________
List mailing list
[email protected]
https://lists.pfsense.org/mailman/listinfo/list

Reply via email to