[Arny Krueger]

"Scott Dorsey" wrote in message
...


No, he's not paging or swapping to the flash drive. That is instant
death.

http://dansdata.com/flashswap.htm

"If your Flash device is "4Gb" with a formatted capacity of 3900Mb, and you
do nothing but write to it as fast as you can - at, say, 30Mb/s - you'll
still only be able to replace its entire contents every 130 seconds. At that
rate, it'll take you 150 days to hit 100,000 cycles."

From from instant death, I'd say.

[nospam[_3_]]

In article , Les Cargill
wrote:

enough to burn through some SSDs in short order. If you can afford SSD for
swap/paging, you can afford enough RAM to make I/O to swap or paging
unnecessary (in many cases).

That is true.

it's not true. ssd may be more expensive than a similar capacity hard
drive, but it's not *that* expensive unless you get really huge
capacities. 128-256 gig ssd are very affordable.

Not the point - it would be even faster with no backing store at all.

which is almost never an option.

A quick google shows 16 GB of DDR3 for roughly $100.

and a quick google shows 128 gig ssd for under $100, which makes for an
incredibly nice speed boost for an older laptop, especially one which
may be limited to 4 gig of ram or less.

I have 4 or 8 GB
RAM on this machine,

don't you know which one it is?

and I never see it use more than 1.5 GB.

then you don't do much with it.

[nospam[_3_]]

In article , Floyd L. Davidson
wrote:

enough to burn through some SSDs in short order. If you can afford SSD for
swap/paging, you can afford enough RAM to make I/O to swap or paging
unnecessary (in many cases).

That is true.

it's not true. ssd may be more expensive than a similar capacity hard
drive, but it's not *that* expensive unless you get really huge
capacities. 128-256 gig ssd are very affordable.

You say first that it is not true... and then recite exactly that it
is in fact true. Logic escapes you?

comprehension escapes you?

[nospam[_3_]]

In article , Floyd L. Davidson
wrote:

Any OS that cannot do that is abjectly insecure.

There's no correlation between being able to boot and run from read-only
media and security.

You are clueless.

If the boot device is an R/O device, there is no such
thing as compromising the booted OS. Think about it...

oh yes there is. you won't be able to compromise the contents of the
actual read-only drive (and who cares), but you can compromise the
*booted* *system* itself once it's up and running.

think about it.

That's not the point.

it's exactly the point.

The point absolutely was that you can't compromise
the booted OS.

which is bull****.

And that means that if it gets compromised after it is booted
the cure is quick and simple: reboot.

which can only be done *after* you notice it has been compromised, and
that might be after quite a bit of damage has been done by the
intruder.

worse, once you do reboot the system, it can get compromised again
because you can't patch the exploit since it's on read only media.

[nospam[_3_]]

In article , Floyd L. Davidson
wrote:

Doesn't that make it a bit difficult to add kernel extensions (drivers)?

It makes it impossible to durably add anything, since any changes require
a write to non-volatile media.

That has nothing to do with adding kernel extensions. The in memory kernel
is modified by loadable modules. The kernel on disk is not modified at
all.

and where do those loadable modules go?

what's on the read-only disk can't be modified, which means you can't
add anything to it. the *only* way you can add an extension or anything
else is by writing to a volume, somewhere on the system.

You need to learn how it works and stop posting nonsense.

after you.

The module to be loaded need never appear on a disk assessible to the
system.

it *must* be accessible, or it can't be loaded.

It could, as an example, exist only in RAM and be downloaded
via a network.

then it's accessible.

it's just not directly attached with a firewire/usb/sata cable.

In fact there need not even be an r/o boot device in
that sense, because it can be loaded from another system via a network.

and what if a module is required for network access?

[nospam[_3_]]

In article , Floyd L. Davidson
wrote:

Doesn't that make it a bit difficult to add kernel extensions
(drivers)?

Drivers and other kernel extensions are linked dynamically in
RAM, not on disk (which indeed is the way it was accomplished
up through the 1980's).

but to add them, you must have a writeable volume.

No, you must have a *readable* volume.

no, you must have a *writeable* volume. how do you add something to a
volume that is only readable? you *can't*. it is *not* possible. there
has be a way to *write* the new kernel extensions or whatever else
you're adding to the system.

The question was not about adding to the boot volume, it was about
adding to the kernel. The kernel, after it is booted, can and typically
is modified.

it can, but then you'd need to do that *each* time you boot. nobody is
going to put up with that ****.

[nospam[_3_]]

In article , Floyd L. Davidson
wrote:

it is definitely rare. who boots off a dvd as part of their day to day
computer usage? almost nobody.

Who boots in day to day computer usage?

who said anything about booting every single day?

So you admit your point was non-sequitur to begin with.

not at all.

the point is that a typical computer is almost always booted off a hard
drive, except in the rare case when a new operating system is being
installed to the hard drive from a dvd, and even that is going away
since operating systems are now dowloadable, no dvds needed at all.

That is true for what *you* do. But it is not necessarily true for
what many computers do.

actually it is true for what many, if not most, computers do.

The typical firewall does not do that.

who cares? firewalls don't have ssd, which is what started this thread.

nevertheless, many people do shut down at night and boot in the morning
for reasons i don't understand, mainly windows users.

I rebooted my computers last week due to an extended
power outage. It was the first time in more than a year
that any of them had been rebooted.

and you no doubt booted off a hard drive, not a dvd.

But that still misses the point. Whether you or I commonly
boot our desktops from a hard disk is not the point in question.

right, it's what the masses do, and they almost universally boot off
hard drives or ssds. *millions* of pcs and macs ship with a hard drive
or ssd.

Note that I did *not* boot my router from a hard disk!

routers, firewalls, even microwave ovens (which has a cpu/ram) are not
relevant to this discussion.

it's about ssd versus hard drives and which one is more likely to fail
first (and that's the hard drive).

With the single exception of my WIFI router that gets
rebooted about once every two weeks. It of course boots
from the r/o device you say is so rare as to never be
used.

you clearly need a better wifi router. anything that needs to be
rebooted every two weeks has some serious problems.

You clearly don't have a clue do you.

i'm not the one with a router than needs to be rebooted twice a month.

i have several wifi routers and they all run 24/7 and i don't remember
the last time they were rebooted. it's been years.

So. That has no significance to the discussion.

see above. you need more reliable wifi routers.

The point still remains that there are multimillions of
systems out there being boot *every day* from r/o devices.

and multi-billions that aren't.

That has no significance to the discussion. It's is not
a question of whether virtually all systems do this or
that, or even if most do this or that.

not only is it significant, it's the entire point.

Try using a little logic and to be focused on the
discussion.

after you.

[nospam[_3_]]

In article , Robert Coe
wrote:

: In some cases, a memory requirement can be addressed by either flash or
: hard drive. In the case of a Sound Devices 744T, I chose to stick with the hard
: drive. My personal "feeling", which I can't substantiate with any hard info,
: is that some hard drives, selected for both make and particular model, and
: handled and mounted to avoid mechanical shock, are more reliable than flash
: alternatives.

Well, that's an entertaining thought, but hard drives in CF format have been
tried. And my recollection (though I never owned one) is that they were
expensive and quite fragile.

microdrives were definitely fragile, however, they're not relevant to
this thread.

[nospam[_3_]]

In article , Eric Stevens
wrote:

It's
gotten worse as Microsoft increasingly attempts to hide the hard disk's
contents from the user. (I want to put things where I think they should go,
and be able to find them.)

Hear hear!

who cares where files are? computers can do a *much* more effective job
at managing where the user's data lives.

apps such as lightroom or aperture pretty much eliminate managing files
and folders, and it's *so* much more productive.

you don't know which sectors a file is on disk, and it doesn't matter.
the hard drive controller manages it for you and does a better job.

[nospam[_3_]]

In article , Mxsmanic
wrote:

it's not true. ssd may be more expensive than a similar capacity hard
drive, but it's not *that* expensive unless you get really huge
capacities. 128-256 gig ssd are very affordable.

If the system requires substantially more backing store or swap than it has
physical RAM, it will run slowly no matter what type of device is used for the
backing store or swap.

not necessarily true. it depends on what you're doing.

[nospam[_3_]]

In article , Mxsmanic
wrote:

you absolutely want swap on ssd because it's significantly faster than
hd. stop worrying about wearing out ssd. you'll be wanting to replace
the computer with a better model long before an ssd wears out.

On production systems, you don't replace anything with a "better model" unless
the workload changes in a way that is no longer served by the existing model.
Replacing SSDs that wear out due to write failure with high physical writes
can be prohibitively expensive.

people replace their computers every few years and there are ssds with
*ten* year warranties. how many people use ten year old computers?

[nospam[_3_]]

In article , Floyd L. Davidson
wrote:

An operating system can be compromised just as easily when booted from a
read-only device as in other cases. Remember, the OS code in memory is not
read-only.

Oh, really? So just any user can write to it?

it doesn't have to be 'any user'. the fact is that ram is writeable and
a running system can be compromised, and if the boot volume is read
only, you *can't* fix it.

[Neil Gould]

Trevor wrote:
"Neil Gould" wrote in message
...
Files from audio apps can be very large and get re-written
frequently, so it would be difficult, if not impossible to
"guarantee" that
such operations would not happen, which is why I asked.

Al the good programs I have used give you setup options on where to
place temp files, data files, backup files, autosave files, multiple
scratch disk usage etc.
There will always be poorly written programs that don't, your mission
is to find something else if you don't like how they behave.

I won't go so far as to say that an app that doesn't give you those setup
options is poorly written for several reasons. For one, the user may not
know where the optimal location of such files should be, for example if some
temp files are on the same drive (regardless of partition) as the primary
data, it can affect performance.

IMO, the functionality of an app takes precedence over such things. I don't
care if I have "ultimate control" over the minutia if the use of the app for
its primary purpose is less efficient.

--
best regards,

Neil

[Arny Krueger]

"Mxsmanic" wrote in message
...
nospam writes:

not when there isn't a second drive available, such as in a laptop,
which is exactly where you are most likely to find ssd. plus, putting
virtual memory swap on an hd is dumb. it's much better to keep it on
the much faster ssd.

Swap and paging files receive a tremendous amount of I/O in most systems,
enough to burn through some SSDs in short order. If you can afford SSD for
swap/paging, you can afford enough RAM to make I/O to swap or paging
unnecessary (in many cases).

Most PC systems lack the duty cycle it would take to burn though the 100,000
or so changes that a modern SSD supports in any given memory location.

If a SSD memory location gets over-used, the data is automatically
relocated.

At this point in time, it is senseless to buy a SSD that is less than 128 GB
because it corresponds to the price of a typical hard drive.

Other than us freaks who edit video, audio or collect high rez photographs
from professional-grade cameras, we'll probably use less than 100 GB over
the life of a computer.

How many hours a day is a not typical but "high use" PC actively used? One?
Two? Three hours?

Most of the time the PC is being "actively used", it is actually just
sitting there, waiting on its user.

[Gary Eickmeier]

I THINK I have just edited some 3D material and produced it out to a BlueRay
disc. I did it in Cyberlink Powerdirector, because the Vegas couldn't do 3D.
It had to be done on a Blue Ray disc or else the 3D wouldn't work, i.e. on a
DVD in any format no 3D, but on a BRD it came though.

Just need to get to the office and try it on our 3D DVD player.

Gary Eickmeier

[Arny Krueger]

"nospam" wrote in message
...
In article , Floyd L. Davidson
wrote:

enough to burn through some SSDs in short order. If you can afford SSD
for
swap/paging, you can afford enough RAM to make I/O to swap or paging
unnecessary (in many cases).

That is true.

it's not true. ssd may be more expensive than a similar capacity hard
drive, but it's not *that* expensive unless you get really huge
capacities. 128-256 gig ssd are very affordable.

128 GB runs about $90 these days - the price of a typical laptop or desktop
hard drive.

[Arny Krueger]

"Mxsmanic" wrote in message
...
nospam writes:

it's not true. ssd may be more expensive than a similar capacity hard
drive, but it's not *that* expensive unless you get really huge
capacities. 128-256 gig ssd are very affordable.

If the system requires substantially more backing store or swap than it
has
physical RAM, it will run slowly no matter what type of device is used for
the
backing store or swap.

The difference that a SSD makes over my personal baseline - a 1 GB 7200 rpm
hard drive, is pretty impressive.

I just built a laptop that booted so fast that its resume/suspend feature
was superfluous. It did either in about the same amount of time - 20
seconds. My new 5.1 receiver takes longer to power up to the point where it
is making sound!

[William Sommerwerck]

128 GB runs about $90 these days -- the price
of a typical laptop or desktop hard drive.

My first gigabyte hard drive -- a SCSI device -- cost about $900 20+ years
ago.

An SSD costs about eight to ten times as much as a mechanical hard drive.
But $90 is not a lot a money for mass storage in a machine costing over
$1000.

[Mxsmanic]

Floyd L. Davidson writes:

False. The program code segment is *never* written to a
"swap file" on any modern OS.

Sure it is, if it has modified itself, which many programs do. Operating
systems do not generally mandate pure procedure code; on some hardware
platforms, they can't, and on others, forcing pure procedure greatly reduces
the flexibility of the operating system.

Of course all such segments today can be "shared or reused".

Sometimes they can, sometimes they cannot. The OS should provide for "dirty"
procedure segments if a process requests it. It really doesn't make much
difference to the OS, anyway, since swapping a procedure segment isn't really
any different from swapping any other segment.

So there is not "always physical I/O".

If there is no physical I/O, there has been no swap. When a process is
swapped, there is always a write to disk. Since the average process has at
least one or more dirty data segments, there is always something that must be
written in a swap.

You do not appear to have any experience with kernel code.

I've written kernel code.

What else can an OS do if an app requests more memory than
exists?

It can deny the request. It happens regularly with applications that use very
large amounts of memory. The application can then decide for itself how it
would like to handle the denial.

It can simply return an error, and in that case any sane program has no
choice but to exit.

Not true. Many applications that use a large amount of memory will handle
denials and continue to operate with degraded performance. Photoshop is one
application that has traditionally been very good about dealing with RAM
limitations.

Or the OS can kill some other process to obtain the memory requested.

I don't know of any OS that does this. The OS cannot know which user
applications can be safely killed and which cannot. It will simply deny the
memory request.

Or it can simply lose sanity, and crash.

I don't know of any OS that does this, either.

Yes. And of course modern OS's are the ones being discussed.

Modern or not, it depends on the circumstances. More disk cache is not always
a good decision.

And that is why large databases typically manage their
own filesystems, and do not use disk caching.

Large files are not always managed by DBMS. In fact, all too often, they
aren't.

Gee whiz, Sherlock, nice that you know that. But since
nobody said they were, what's yoour point. Regardless
of the fact that paging is what is being done, it is
still called a swap device.

No, it's more frequently called a backing store, although nomenclature varies
with the OS, and with which college classes the speaker has taken.

The whole point is that most processes have significant
code and data that need not be kept in RAM.

That is not necessarily true. It depends on the program.

The data is paged to the swap device, the program code is merely
released and marked as available from the original
binary file.

As a general rule, nothing is paged out unless there are things that need to
be paged in and there are no free pages. Paging is extremely expensive and is
resorted to only when limited main memory requires it.

We aren't talking about legacy systems or inferior systems.

I know. But modern and superior systems work the same way.

[Mxsmanic]

nospam writes:

not necessarily true. it depends on what you're doing.

True, but the risk of thrashing rises very rapidly beyond a certain threshold,
and the OS cannot know in advance what that threshold might be.

Overall, a system that is low on memory will degrade smoothly up to this
threshold, and then will dramatically degrade to almost a standstill.
Essentially any system that pages or swaps is potentially vulnerable to it.

[Mxsmanic]

Arny Krueger writes:

The difference that a SSD makes over my personal baseline - a 1 GB 7200 rpm
hard drive, is pretty impressive.

That is to be expected: access time on the SSD is much lower. If SSDs didn't
have the weakness of limited writes (and high price, although prices will come
down), they'd be a great replacement for hard disks. But some uses of
disk/SSD, such as paging and swapping and certain server-style applications,
can potentially burn through a SSDs life expectancy very rapidly (months,
weeks, or days).

I just built a laptop that booted so fast that its resume/suspend feature
was superfluous. It did either in about the same amount of time - 20
seconds. My new 5.1 receiver takes longer to power up to the point where it
is making sound!

Do you have swap/paging files on the SSD?

[Mxsmanic]

Floyd L. Davidson writes:

And nobody in their right mind, given the price of RAM
today, operates any computer in that manner.

Most computers are operated that way. It's not usually cost-effective to add
enough RAM to hold all processes at all times, even when it is physically
possible.

Disk space may be cheap, but so is RAM today.

A terabyte of RAM is much more expensive than a terabyte of disk, and that is
likely to remain true indefinitely.

[Mxsmanic]

Arny Krueger writes:

Most PC systems lack the duty cycle it would take to burn though the 100,000
or so changes that a modern SSD supports in any given memory location.

Except for things like paging files. Some server applications can also do huge
amounts of physical I/O that put a strain even on conventional hard drives.

Desktops are pretty safe except for paging and swap files, and if there is
truly a huge amount of RAM on the system and processes don't overflow too
much, paging and swap may reduced as well, although I'm sure there is
unnecessary activity in every case.

Most of the time the PC is being "actively used", it is actually just
sitting there, waiting on its user.

Yes. But at the same time, the primary source of delay on a typical desktop is
disk I/O. That's why replacing a conventional disk with SSD can speed up the
desktop so dramatically.

In an ideal world, a system would have nothing but terabytes of RAM, and it
would be high-speed RAM that runs as fast as processor cache. Such a system
would be so fast that it would scarcely be recognizable as a computer--the
computer's reaction would be instantaneous, all the time, for everything.

[Mxsmanic]

Arny Krueger writes:

"If your Flash device is "4Gb" with a formatted capacity of 3900Mb, and you
do nothing but write to it as fast as you can - at, say, 30Mb/s - you'll
still only be able to replace its entire contents every 130 seconds. At that
rate, it'll take you 150 days to hit 100,000 cycles."

From from instant death, I'd say.

Depends. If you have multiple computers, the 150-day delay is divided by their
number. So if you have 50 machines in the office, you're going to be replacing
a SSD every 72 hours.

[Mxsmanic]

nospam writes:

people replace their computers every few years and there are ssds with
*ten* year warranties. how many people use ten year old computers?

I do. I use a computer for as long as it lasts or until I need something
faster, whichever comes first. I've been using essentially the same
applications for many years, so the life expectancy of the hardware is the
limit factor, and it's often well beyond ten years if it's of good quality.

[Mxsmanic]

nospam writes:

you don't know which sectors a file is on disk, and it doesn't matter.
the hard drive controller manages it for you and does a better job.

The controller doesn't have enough information or intelligence to optimize
sector placement in every case.

[Mxsmanic]

Floyd L. Davidson writes:

"The module to be loaded need never appear on a disk assessible to the
system."

Emphasis added for those with a low attention span.

Emphasizing it does not prevent it from being self-contradictory.

A loadable object is always accessible, otherwise it is not loadable.

In that case, it is accessible to the system, isn't it?

The disk it is on is not.

You haven't answered my question. It is accessible, isn't it? Which undermines
your own previous statement.

No it isn't.

The very vast majority of servers and desktops do not dynamically load
anything from a network. One key reason is that networks are unbelievably
slow. Another is that they are unreliable, as well as their contents. Still
another is the security risk that this represents. And most of the time there
is simply no advantage to it.

So?

So most people use conventional desktops, and indirectly servers, neither of
which are generally implementing these exotic configurations.

[Mxsmanic]

Floyd L. Davidson writes:

Mxsmanic wrote:
Floyd L. Davidson writes:

Mxsmanic wrote:
isw writes:

Doesn't that make it a bit difficult to add kernel extensions (drivers)?

It makes it impossible to durably add anything, since any changes require a
write to non-volatile media.

That has nothing to do with adding kernel extensions. The in memory kernel
is modified by loadable modules. The kernel on disk is not modified at all.

Thank you for explaining my point.

When was that ever *your* point?

In the statement you backquoted just before your own reply.

[Mxsmanic]

Floyd L. Davidson writes:

Bull**** kid. It doesn't take "an extremely long time".

Without a pre-determined configuration, a system must poll for discovery of
any devices that require drivers. Since this often involves waiting for
timeouts, it is extremely time-consuming and can prolong the boot process by
many minutes, and sometimes much more. If anything hangs the discovery
procedure, the machine may not finish booting at all.

[Sean Conolly]

"Soundhaspriority" wrote in message
...


"Sean Conolly" wrote in message
...
[snip]
SLC drives in the same application should last over 10 years. I sure hope
so because we just bought six of them at $3K each, discounted!

Sean

Sean,
What do you think of the Tom's Hardware article? One of the conclusions
is
"Despite SLC-based drives accounting for only a fraction of the NAND
market, we have much more data on SLC-based SSDs than we do on those using
MLC technology. Even though our data set is one-twentieth the size of
previous studies on hard drives, our information starts to suggest that
SLC-based SSDs are no more reliable than SAS and SATA hard drives. "

http://www.tomshardware.com/reviews/...te,2923-3.html


Interesting but not that surprising, really. The application that I'm using
them for will be for database buffer caching, basically as slower RAM
instead of a faster disk. If one fails it will be relatively benign to
normal operations.

It's actually not the best solution to the problem, but it seems to be the
best within the organizational parameters I've been given. (Translated: any
time there's money to be spent, there's people lining up to tell you how to
spend it)

Sean

[Mxsmanic]

Floyd L. Davidson writes:

Mxsmanic wrote:

If you want to add them across boots, you need a writable, non-volatile
device.

That's bull****.

You either load your modules and then write the modified OS to a writable
device, which can then be loaded directly on subsequent boots, or you load
your modules anew for each and every boot. There are no other options.

And if you don't have a writable device, your only option is to load the
modules for every boot, which means that you cannot retain them across boots.

[Mxsmanic]

Floyd L. Davidson writes:

So?

Reloading large parts of the OS at every boot is very time-consuming.

Typically a modern desktop system might load anything
from a dozen to a many dozens of modules at every boot.
The machine that I'm writing this article on has 31
kernel modules loaded as I'm writing.

Do you have any writable devices on the system?

[Mxsmanic]

Floyd L. Davidson writes:

Oh, really? So just any user can write to it?

Sometimes. That's how systems get compromised. Some operating systems are more
secure than others.

[nospam[_3_]]

In article , Arny
Krueger wrote:

not when there isn't a second drive available, such as in a laptop,
which is exactly where you are most likely to find ssd. plus, putting
virtual memory swap on an hd is dumb. it's much better to keep it on
the much faster ssd.

Swap and paging files receive a tremendous amount of I/O in most systems,
enough to burn through some SSDs in short order. If you can afford SSD for
swap/paging, you can afford enough RAM to make I/O to swap or paging
unnecessary (in many cases).

Most PC systems lack the duty cycle it would take to burn though the 100,000
or so changes that a modern SSD supports in any given memory location.

If a SSD memory location gets over-used, the data is automatically
relocated.

At this point in time, it is senseless to buy a SSD that is less than 128 GB
because it corresponds to the price of a typical hard drive.

true.

Other than us freaks who edit video, audio or collect high rez photographs
from professional-grade cameras, we'll probably use less than 100 GB over
the life of a computer.

keep large assets on a hard drive and still use ssd for the system and
applications.

How many hours a day is a not typical but "high use" PC actively used? One?
Two? Three hours?

Most of the time the PC is being "actively used", it is actually just
sitting there, waiting on its user.

true.

[nospam[_3_]]

In article , William Sommerwerck
wrote:

128 GB runs about $90 these days -- the price
of a typical laptop or desktop hard drive.

My first gigabyte hard drive -- a SCSI device -- cost about $900 20+ years
ago.

so what? what matters is what it costs today.

An SSD costs about eight to ten times as much as a mechanical hard drive.
But $90 is not a lot a money for mass storage in a machine costing over
$1000.

nonsense.

128 gig ssd are under $100. where can one get a similar capacity hd for
$10, new? you can't.

[nospam[_3_]]

In article , Arny
Krueger wrote:

The difference that a SSD makes over my personal baseline - a 1 GB 7200 rpm
hard drive, is pretty impressive.

I just built a laptop that booted so fast that its resume/suspend feature
was superfluous. It did either in about the same amount of time - 20
seconds. My new 5.1 receiver takes longer to power up to the point where it
is making sound!

20 seconds to suspend/resume???

it should be instant. something is wrong. even with a hard drive it's a
second or two.

[nospam[_3_]]

In article , Mxsmanic
wrote:

The difference that a SSD makes over my personal baseline - a 1 GB 7200 rpm
hard drive, is pretty impressive.

That is to be expected: access time on the SSD is much lower. If SSDs didn't
have the weakness of limited writes (and high price, although prices will come
down), they'd be a great replacement for hard disks.

they are a great replacement for hard disks. the only issue is price
which really only affects the larger sizes. 128-256g ssd is very
affordable.

But some uses of
disk/SSD, such as paging and swapping and certain server-style applications,
can potentially burn through a SSDs life expectancy very rapidly (months,
weeks, or days).

nonsense.

I just built a laptop that booted so fast that its resume/suspend feature
was superfluous. It did either in about the same amount of time - 20
seconds. My new 5.1 receiver takes longer to power up to the point where it
is making sound!

Do you have swap/paging files on the SSD?

that's where it should be. why would you want swap on a slower hard
drive?

[nospam[_3_]]

In article , Mxsmanic
wrote:

Or the OS can kill some other process to obtain the memory requested.

I don't know of any OS that does this.

mac os x 10.7 and later and ios 4.0 and later will quit idle apps and
resume then as needed, often to exactly where it was when quit (depends
on the app), so to the user, they don't even notice.

The OS cannot know which user
applications can be safely killed and which cannot. It will simply deny the
memory request.

nonsense. of course it can.

Or it can simply lose sanity, and crash.

I don't know of any OS that does this, either.

you've never seen an os crash?

[nospam[_3_]]

In article , Mxsmanic
wrote:

people replace their computers every few years and there are ssds with
*ten* year warranties. how many people use ten year old computers?

I do. I use a computer for as long as it lasts or until I need something
faster, whichever comes first. I've been using essentially the same
applications for many years, so the life expectancy of the hardware is the
limit factor, and it's often well beyond ten years if it's of good quality.

you're an outlier.

very, very few people are using ten year old systems.

[nospam[_3_]]

In article , Mxsmanic
wrote:

you don't know which sectors a file is on disk, and it doesn't matter.
the hard drive controller manages it for you and does a better job.

The controller doesn't have enough information or intelligence to optimize
sector placement in every case.

nonsense. of course it does.

the point is, users don't care where on the hard drive the file
actually is, nor should they care what folder it's in. as long as they
can access their data, they're happy.