Berkeley CSUA MOTD:2009:March:29 Sunday <Monday>
Berkeley CSUA MOTD
2009/3/29-4/2 [Politics/Domestic/Crime, Politics/Foreign/Europe] UID:52765 Activity:low
3/29    Good news, Yoo to be investigated for violating International law:
        \- I'm actually happier to see Alberto, Feith, and Addington on
           the list ahead of Yoo.
2009/3/29-4/3 [Uncategorized] UID:52766 Activity:kinda low
3/29    Swensen hater:
        \_ The invisible hand is taking care of this.
           \_ What are you talking about?
              \_ Wealth is being taken from the weak hands and properly
                 redeployed to those who will make better use of it. This
                 is the way capitalism is supposed to function.
2009/3/29-4/3 [Politics/Domestic/Election, Politics/Domestic/President/Bush] UID:52767 Activity:kinda low
3/29    Yah, that's not fascist or anything:
        GM CEO resigns at Obama's behest
        \_ You're right, it's not!
        \_ No, it's not. Again, you're confusing fascist with national-
           ization. Since GM took bailout money from the govt., I don't
           see why there's outrage when the govt. demands the CEO step
           down. If GM doesn't like it, they're still free to refund the
           money given them and fend for themselves.
        \_ Actually this is kind of fascist. But I can live with it. -O voter
           \_ huh?  what is so fascist about this?  Wagner step down per
              it's share holder's request.  What is wrong with that?
              \_ In spite of popular usage fascist is not a synonym for
                 "anything I don't like." Government and industry being in
                 bed together is pretty fascist, especially since the
                 government is dictating to GM how they should run their
                 business. As I said before, I don't really have a problem
                 with it, since GM obviously is failing with its current
2009/3/29-4/3 [Computer/HW/Laptop, Science/GlobalWarming] UID:52768 Activity:high
3/29    "Leaving computers on overnight = $2.8 billion a year"
        \_ Not good for hardware to power it up and down all the time. I
           always leave all my computers on all the time, except for
           laptops which I allow to sleep (but still be powered).
           \_ How is this the case for desktops but not laptops?  I don't see
              how turning something off at the end of the day would cause a
              problem.  Seems like less wear and tear.
              \_ It's more wear and tear to power cycle and lubricants get
                 cold. Your system will fail more quickly if you turn it
                 off every night in my experience.
                 How laptops are different:
                 1. Their h/w is better designed for low or no power.
                 2. They are much better about sleep/wakeup than desktops,
                    in part because of 1.
                 3. If a laptop dies it's not usually a big deal because
                    they are not expected up 100% of the time anyway.
                 \_ Here you go. Welcome to the 21st century.
                    \_ From a link found in one of your links:
"[Temperature cycling] is a well-established failure mechanism and a
stress on components," McCredie pointed out. "What it really comes down to
is all these things -- chips soldered on modules, soldered on boards and
connectors -- that expand and contract when they heat and cool.... When
they all contract and expand at different rates, they can fail. That's
ultimately the bad thing with power cycling," he said.
--Brad McCredie, an IBM fellow for the Systems and Technology Group
                        You can say "Cycling power on a sick system is
                        going to bring attention to latent component
                        weaknesses that go unnoticed in operation" which
                        maximizes "server availability", but that's an odd
                        way of looking at it. If you don't power down the
                        system then you won't have the failure and your
                        availability will be greater. I have gone through
                        just a few dozen exercises where equipment in an
                        entire data center was powered down and equipment
                        always fails at a much higher rate when that
                        happens. Whether it's just highlighting a latent
                        sickness is semantics. If you don't power it off,
                        then it won't fail, even if the power off is not
                        the root cause.
                        \_ Well, if you want to cling onto any comment
                           as a way to rationalize your practices, that's
                                \_ why give this guy a pass?  Fuck that!
                                   this is our planet too.  --ecoterr'rst
                           The way I synthesize everything is if
                           you're not power-cycling a lot, you'll very
                           likely be fine for the useful life of the
                           product. It's a tradeoff between perceived
                           higher failure rates, which may go up "negligibly"
                           with occasional power cycling, versus the cost
                           to your pocketbook and the environment of running
                           hardware 24/7 even when it's not being used.
                           There are multiple comments that suggest that
                           turning off your equipment at night (or whenever
                           it won't be used for an extended period of time)
                           is the better tradeoff, but of course you
                           conveniently ignore those points.
                           P.S. A datacenter is a very different use case
                           than a typical home or work desktop. Although
                           powering down parts of a datacenter would still
                           appear to be much more a logistical problem than
                           a hardware failure problem.
                           \_ Different use case, but basically same
                              components. If they fail measurably (not
                              "negligibly") in that case then they will
                              fail at your home, too. Most people probably
                              do not notice because a 1/N failure rate
                              means you will likely be fine, but when the
                              number of machines is 1000N it's a noticeable
                              issue. I always count on a power outage to
                              result in dead hardware. You can make the
                              argument that a 5-10% chance of failure (say)
                              over the lifetime of the system is low, but
                              I'd rather just keep my system up. The key
                              point here is not that I am against powering
                              off equipment, but realize it comes with a
                              non-negligible risk of failure. Given my
                              experiences I just keep my equipment powered
                              on. As far as being environmentally aware,
                              maybe you should talk to the people in the 55
                              story buildings who leave the lights on all
                              night. Go anywhere in SF, LA, or NYC
                              to see that lots of lights are not on sensors.
                              Those "spectacular city views" are wasting a
                              lot of $$$.
                              \_ well, if you have control of the lights
                                 in the skyscrapers, by all means, turn
                                 them off when not in use.  If you don't,
                                 you should make wise choices about the
                                 things you do control.  -tom
                                 \_ I gave up on convincing him. Guy just
                                    doesn't want to do it.
                                    \_ I like my h/w to not fail. Maybe you
                                       don't mind if it does. You can
                                       pretend the h/w won't fail, the OS
                                       will always boot fine, and the s/w
                                       will start up fine every time if you
                                       want to but there is evidence to the
                                       contrary. It's not a case of comparing
                                       my $500 machine to the $8/month bill
                                       and saying that I can replace the
                                       machine with the savings over the life
                                       of the machine. Having my machine
                                       fail is *BAD* and causes me grief.
                                       If you turn yours off every day then
                                       more power to you. My experience has
                                       shown that IBM guy (who *you* ignore)
                                       is correct and power cycling systems
                                       results in failures. You know that
                                       server in the corner that no one knows
                                       what it does exactly but is VERY
                                       IMPORTANT and the guy who built the
                                       s/w that runs on it - and only on
                                       that very h/w - left 15 years ago?
                                       DON'T TURN IT OFF OR REBOOT IT.
                                       Now, if you understand why that is a
                                       bad idea extend it to your own system
                                       because the same physics are at work.
                                       Now if a h/w failure is meaningless to
                                       you because you have great backups and
                                       no need for quick recovery and can buy
                                       another machine and so on then go ahead
                                       and power down, but that's analyzing
                                       risk/reward which is not the same as
                                       saying there is no problem with
                                       powering your system down. There is.
                                       \_ You speak as if hardware won't fail
                                          if you don't turn it off. It will;
                                          it's just a matter of time. And if
                                          your plan for handling the failure
                                          of a VERY IMPORTANT system is to
                                          not have it fail in the first place
                                          by not turning it off, you're just
                                          stupid. Although I've seen people
                                          have just such "plans", but then,
                                          newsflash, there are stupid people
                                          out there.
                                          \_ It will fail more rapidly if you
                                             keep turning it on and off every
                                             day, whether that's because
                                             the components are stressed
                                             or just because latent failures
                                             materialize. I know HP said
                                             that the h/w won't, but I
                                             doubt they tested systems
                                             under real world situations
                                             like under a desk with a case
                                             full of dust bunnies. My own
                                             experience has shown that h/w
                                             will fail when powered down
                                             and I dread it whenever
                                             electrical work has to be done
                                             because it means something is
                                             going to die. Sure, sometimes
                                             h/w dies anyway but it's
                                             *guaranteed* something will
                                             die when everything has been
                                             powered off. I have seen this
                                             many times over 12 years now.
Berkeley CSUA MOTD:2009:March:29 Sunday <Monday>