Uhm, you do realise that Intel had one of the better ARM processors out there back in the PDA days, right? Do some reading on Intel Xscale CPUs. They were in several Palm/WinCE devices, and then Intel sold off the division.
I like the intersections that have countdown timers for pedestrians.:) Much easier to quickly glance at a number than to try to guess how long the "flashing hand" has been there and how much longer until the yellow light starts.:)
These are getting more common in the bigger cities, and even in the smaller ones around here. Very handy!
That's the same training (at least from the better schools, who knows what the fly-by-nights teach) on this side of the pond. It's not a training issue, it's a cultural one.
Just compare the differences in moped use between NA and Europe. Over here, a bike that weaves between cars to the red light gets honked at, cursed at, risks getting beat up, etc. In Europe, red lights mean "moped to the front", initial green light means "mopeds go first".:) It's crazy watching the bike traffic in Spain/Italy.
The way developers have been doing it for decades, using the de facto standard of major.minor.build version numbers.
IOW, JDK 5 uses 5.x.y numbers, JDK 6 uses 6.x.y numbers, JDK 7 uses 7.x.y numbers. X is incremented for feature releases and resets Y to 0. Y is incremented for security releases.
Simple. Easily parseable. And it the ways things used to be done before the whole stupid "my number needs to be bigger than your number, logic be damned" movement started.
Wife would pump at regular times throughout the morning, afternoon, evening, occassionally at night, and we'd use the appropriate bottle for the time of the feeding.
Our daughter wouldn't latch and wouldn't get anything from the breast. Every feeding was a bottle feeding. But every feeding was also breast milk.
It can be. Take your shirt off to provide warmth, sound of a heartbeat, and comfort. Cradle the baby the same way a mother would. Put the bottle nipple in approx the same place as the mother's would be.
They get all the same benefits, including the breast milk. There's very little difference between a mother nursing a baby and the father (or anyone else, really) feeding the baby breast milk from a bottle. You still get the same bonding time and experience, but now it's not "limited" to the mother.
We were in the hospital for a week, and she was basically in bed for another 2, and not fully-recovered for over a month!
A *LOT* of physical (and emotional, hormonal, mental, etc) happens during a pregnancy, and a lot of physical trauma occurs in child-birth. It's not "equivalent to day surgery", except maybe if it's the fourth or fifth child.
If you haven't been in the delivery room with a woman screaming in pain, and had to live with said women after the birth, you really are *NOT* qualified to comment.:)
Pretty flimsy limb considering breast pumps have been around for decades, and women expressing by hand into bottles has been around for centuries, allow anyone to feed the baby via bottle.
There's a total of 52 weeks parental leave; the first 8-ish have to be taken by the mother (which makes sense as she's just leaving the hospital), but the rest can be taken by either parent. It's only 60% of your normal pay, but it's better than nothing. And a lot of companies provide top-up benefits (we got 13 weeks top-up to 100%).
The 10" eeePCs were just about the perfect formfactor, although 11" would probably be perfect. It's too bad they haven't been updated in awhile and are running crappy Atom with even crappier Intel graphics. Always wanted to try one of the AMD-based versions.
Recommended, not required. There are plenty of folks running FreeBSD + ZFS on systems with 1 GB of RAM total, as well of plenty of folk running it on laptops/netbooks (slow disk, not a lot of RAM, 32-bit CPUs, etc).
Have a look in top output sometime. See all that RAM being used as "cache" and "buffer", that's your filesystem using up system RAM. I've seen it get into the multiple GBs, yet no one complains about that.
Yet, that is the equivalent of the ARC in ZFS. And, when the OS needs RAM for application usage, data in "cache" and "buffer" is flushed and used for the apps. Same thing on ZFS-using systems: if the OS needs RAM for apps, the ARC releases it.
Not really sure why this is an issue for people. At least ZFS lets you tune things, whereas you can't tune the amount of RAM used for "cache" or "buffer" in a 'normal' setup.
ZFS uses as much (or as little) RAM as you decide to let it use. The ARC is fully tunable.
By default, it uses "all but 1 MB" of RAM, but releases RAM as needed when the OS tells it to. However, you can set a max as low as you want (although you can't disable it).
Have 64 GB of RAM, but need 32 GB for the DB? Then set the arc_max tunable to 32 GB (or even less).
Open-source ZFS is mainly developed by Illumos, with contributions from FreeBSD, Linux, Joyent, Nexenta, and many others.
Many new features have hit OSS ZFS that aren't in Oracle ZFS: - LZ4 compression - feature flags - delayed deletion of snapshots - many many many bug fixes - many many many optimisations - bunch of other stuff I can't recall
If you use OSS ZFS with feature flags enabled, the version number shows as 5000.
It's nowhere near "5 GB of ARC per 1 TB of disk". That meme really needs to stop.
We have 48 GB of RAM, 40 GB reserved for ARC, no L2ARC, running with dedupe enabled on a server with just under 30 TB of raw storage, 22 TB in use. Runs fine. It does rsync backups every night of 60-odd remote servers. Starts at 5pm, ends around 2am. Then does a ZFS send to an off-site system in just over an hour. Only has 15 disks in 3 raidz2 vdevs.
The other backups boxes have 64 GB of RAM, but just under 50 TB of disk.
The off-site box has only 128 GB of RAM for just shy of 100 TB of disk.
The correct math is approx 1 GB of RAM per unique TB of data. The more duplicated data you have, the less ARC space you need to hold the dedup tables.
Maybe not on servers. But we install the nVidia binary drivers on approx 3000 Debian-based diskless clients in our district. Have been for almost a decade now.
I love how everytime someone "solves" your "not ready for production" issue you dream up a new one.
Look, if you, specifically, don't want to install ZFS on your particular Linux systems doesn't mean it's not "production ready" for the rest of us.
Why haven't you moved to ZFS, then?;) Has all the features you want for a backups server (which is where we currently use it):
- near instantaneous snapshots - near unlimited snapshots - no need to worry about sizes - data and metadata checksums - multiple levels of parity (single, double, triple) - n-way mirroring - compression (lzjb, gzip, lz4) - online dedupe (although you need gobs of RAM) - and more!
If you don't like Oracle, run it on Nexenta/Illumos-based distros. If you don't like Solaris, then run it on FreeBSD. If you really don't like OSes that work, you can even run it on Linux.:)
We're using it for backups (just under 200 remote systems rsync'd to a ZFS server; zfs send/recv to an off-site replica) with compression and dedupe (just over 5:1 disk savings). Separate filesystems for each remote site (toying with idea of separate filesystems for each server), snapshots made every morning, going back almost 2 years already.
Using 3x 24-drive SuperMicro chassis for the "local" backups, and a 45-drive JBOD plugged into a head unit for the off-site replica.
A lot of Btrfs development comes from RedHat. And there are other sources of patches as well. Plus, the code is part of the Linux kernel, meaning Oracle can't close-source it.
Slashdot Mirror
Uhm, you do realise that Intel had one of the better ARM processors out there back in the PDA days, right? Do some reading on Intel Xscale CPUs. They were in several Palm/WinCE devices, and then Intel sold off the division.
Now that is some smart thinking!
I like the intersections that have countdown timers for pedestrians. :) Much easier to quickly glance at a number than to try to guess how long the "flashing hand" has been there and how much longer until the yellow light starts. :)
These are getting more common in the bigger cities, and even in the smaller ones around here. Very handy!
That's the same training (at least from the better schools, who knows what the fly-by-nights teach) on this side of the pond. It's not a training issue, it's a cultural one.
Just compare the differences in moped use between NA and Europe. Over here, a bike that weaves between cars to the red light gets honked at, cursed at, risks getting beat up, etc. In Europe, red lights mean "moped to the front", initial green light means "mopeds go first". :) It's crazy watching the bike traffic in Spain/Italy.
I understand math, I understand 2x=100% faster, etc.
But "within 2x native speed" is nonsensical and non-mathematical.
Considering even you got it wrong ... I wonder who needs to go to school now ...
Considering everyone who's replied to my post has gotten it wrong, my point stands. ;)
The way developers have been doing it for decades, using the de facto standard of major.minor.build version numbers.
IOW, JDK 5 uses 5.x.y numbers, JDK 6 uses 6.x.y numbers, JDK 7 uses 7.x.y numbers. X is incremented for feature releases and resets Y to 0. Y is incremented for security releases.
Simple. Easily parseable. And it the ways things used to be done before the whole stupid "my number needs to be bigger than your number, logic be damned" movement started.
Correct ... for the distro version.
The kernel version is 3.whatever.
Two completely separate numbers, incremented separately. Just like Windows OS versions and kernel versions.
What the hell does "within 2x of native speed" mean? Is it 2x faster? Is it 2x slower? Is it 200% faster/slower? Is it 50% slower?
If only there were a marketing-to-English dictionary, or perhaps a school we could send people to ...
If the system needs RAM, it's released from the ARC and made available to the OS/apps.
If the system needs RAM, it's released from the page cache and made available to the OS/apps.
How's it any different?
[i]There's no "buffer" in a 'normal' setup[/i]
Oh? Then that "buffer" stat for RAM in top output must be magical and not really exist on every Linux system out there?
It's not like she pumped once a day. Yeesh!
Wife would pump at regular times throughout the morning, afternoon, evening, occassionally at night, and we'd use the appropriate bottle for the time of the feeding.
Our daughter wouldn't latch and wouldn't get anything from the breast. Every feeding was a bottle feeding. But every feeding was also breast milk.
It can be. Take your shirt off to provide warmth, sound of a heartbeat, and comfort. Cradle the baby the same way a mother would. Put the bottle nipple in approx the same place as the mother's would be.
They get all the same benefits, including the breast milk. There's very little difference between a mother nursing a baby and the father (or anyone else, really) feeding the baby breast milk from a bottle. You still get the same bonding time and experience, but now it's not "limited" to the mother.
Especially if its a first pregnancy!
We were in the hospital for a week, and she was basically in bed for another 2, and not fully-recovered for over a month!
A *LOT* of physical (and emotional, hormonal, mental, etc) happens during a pregnancy, and a lot of physical trauma occurs in child-birth. It's not "equivalent to day surgery", except maybe if it's the fourth or fifth child.
If you haven't been in the delivery room with a woman screaming in pain, and had to live with said women after the birth, you really are *NOT* qualified to comment. :)
Pretty flimsy limb considering breast pumps have been around for decades, and women expressing by hand into bottles has been around for centuries, allow anyone to feed the baby via bottle.
You never heard of a breast pump and bottles?
We breast fed our daughter for over 6 months, yet I (the father) did most of the feedings, especially the night-time ones.
How? My wife would pump throughout the day, we'd store it up in the fridge, and I'd use it as needed.
Not really rocket-science here folks.
Canada does to.
There's a total of 52 weeks parental leave; the first 8-ish have to be taken by the mother (which makes sense as she's just leaving the hospital), but the rest can be taken by either parent. It's only 60% of your normal pay, but it's better than nothing. And a lot of companies provide top-up benefits (we got 13 weeks top-up to 100%).
The 10" eeePCs were just about the perfect formfactor, although 11" would probably be perfect. It's too bad they haven't been updated in awhile and are running crappy Atom with even crappier Intel graphics. Always wanted to try one of the AMD-based versions.
All the Intel-based Chromebooks are just rebadged Windows laptops. The only difference is the BIOS/EFI firmware needed to boot ChromeOS.
Anandtech's review of the Intel-based Chromebooks covers this in detail, even giving you model numbers to compare.
And it's because these are crappy rebadged uber-low-end Windows laptops that they make such crappy Chromebooks.
If you actually want a Chromebook, then get the Samsung one using the Cortex-A15 chipset. At least then you'll get all-day battery life.
Recommended, not required. There are plenty of folks running FreeBSD + ZFS on systems with 1 GB of RAM total, as well of plenty of folk running it on laptops/netbooks (slow disk, not a lot of RAM, 32-bit CPUs, etc).
Have a look in top output sometime. See all that RAM being used as "cache" and "buffer", that's your filesystem using up system RAM. I've seen it get into the multiple GBs, yet no one complains about that.
Yet, that is the equivalent of the ARC in ZFS. And, when the OS needs RAM for application usage, data in "cache" and "buffer" is flushed and used for the apps. Same thing on ZFS-using systems: if the OS needs RAM for apps, the ARC releases it.
Not really sure why this is an issue for people. At least ZFS lets you tune things, whereas you can't tune the amount of RAM used for "cache" or "buffer" in a 'normal' setup.
ZFS uses as much (or as little) RAM as you decide to let it use. The ARC is fully tunable.
By default, it uses "all but 1 MB" of RAM, but releases RAM as needed when the OS tells it to. However, you can set a max as low as you want (although you can't disable it).
Have 64 GB of RAM, but need 32 GB for the DB? Then set the arc_max tunable to 32 GB (or even less).
ZFS was forked years ago with ZFSv28.
Oracle ZFS is ZFSv33 or thereabouts.
Open-source ZFS is mainly developed by Illumos, with contributions from FreeBSD, Linux, Joyent, Nexenta, and many others.
Many new features have hit OSS ZFS that aren't in Oracle ZFS:
- LZ4 compression
- feature flags
- delayed deletion of snapshots
- many many many bug fixes
- many many many optimisations
- bunch of other stuff I can't recall
If you use OSS ZFS with feature flags enabled, the version number shows as 5000.
It's nowhere near "5 GB of ARC per 1 TB of disk". That meme really needs to stop.
We have 48 GB of RAM, 40 GB reserved for ARC, no L2ARC, running with dedupe enabled on a server with just under 30 TB of raw storage, 22 TB in use. Runs fine. It does rsync backups every night of 60-odd remote servers. Starts at 5pm, ends around 2am. Then does a ZFS send to an off-site system in just over an hour. Only has 15 disks in 3 raidz2 vdevs.
The other backups boxes have 64 GB of RAM, but just under 50 TB of disk.
The off-site box has only 128 GB of RAM for just shy of 100 TB of disk.
The correct math is approx 1 GB of RAM per unique TB of data. The more duplicated data you have, the less ARC space you need to hold the dedup tables.
Maybe not on servers. But we install the nVidia binary drivers on approx 3000 Debian-based diskless clients in our district. Have been for almost a decade now.
I love how everytime someone "solves" your "not ready for production" issue you dream up a new one.
Look, if you, specifically, don't want to install ZFS on your particular Linux systems doesn't mean it's not "production ready" for the rest of us.
Why haven't you moved to ZFS, then? ;) Has all the features you want for a backups server (which is where we currently use it):
- near instantaneous snapshots
- near unlimited snapshots
- no need to worry about sizes
- data and metadata checksums
- multiple levels of parity (single, double, triple)
- n-way mirroring
- compression (lzjb, gzip, lz4)
- online dedupe (although you need gobs of RAM)
- and more!
If you don't like Oracle, run it on Nexenta/Illumos-based distros. If you don't like Solaris, then run it on FreeBSD. If you really don't like OSes that work, you can even run it on Linux. :)
We're using it for backups (just under 200 remote systems rsync'd to a ZFS server; zfs send/recv to an off-site replica) with compression and dedupe (just over 5:1 disk savings). Separate filesystems for each remote site (toying with idea of separate filesystems for each server), snapshots made every morning, going back almost 2 years already.
Using 3x 24-drive SuperMicro chassis for the "local" backups, and a 45-drive JBOD plugged into a head unit for the off-site replica.
A lot of Btrfs development comes from RedHat. And there are other sources of patches as well. Plus, the code is part of the Linux kernel, meaning Oracle can't close-source it.
IOW, there's nothing to worry about here.