Ext4 trumps Ext3 as soon as you start dealing with files larger then a few hundred MB. When you delete a big file under Ext3, it spends ages and ages playing cleanup. Delete a big file under Ext4 and it takes under a second.
Personally, I don't see a reason to use Ext4 yet on/boot and probably not the root partition either. When it comes to/boot and root, I prefer to stay very conservative and Ext4 is still a bit young (another 2 years and I'll be more comfortable with it). But I use ext4 all the time on data partitions.
Stealing your monitor is a symptom of deeper dysfunction. It would tell me that there is an endemic problem at the company that is not conducive to actually getting work done there.
Exactly. If they're willing to screw with a developer like that on an whim, then they're probably doing other stupid and trivial things instead of focusing on the real problems. It's a symptom of far deeper issues at the company and of management that has lost touch with reality in the trenches.
Why not a more popular distro not to disparage Cent.....
If you're doing Linux for business, there are only a few games in town.
RHEL - and CentOS / Scientific Linux are both based off of RHEL. Any book you pickup off the shelf that talks about RHEL 5/6 will apply to CentOS / Scientific Linux. Any software package out there which runs on RHEL 5/6 will run on CentOS. Which makes CentOS a very good distro for testing the waters and learning how Red Hat does things before you start paying for RHEL subscriptions. Red Hat is the big daddy in commercial / business Linux. Getting support from a local shop is a sure thing (as long as they have some Linux folks on staff).
Debian / Ubuntu - You can find books for it, but my impression over the years is that Canonical is really focused more on the desktop market then the server market. The primary reason for running a server on Ubuntu LTS would be because you're more used to how Debian does things then how RHEL does things. Big enough market share that a local Linux shop should know how to fix it.
SuSE - This used to be a great choice, until Novell bought them, did the patent deal with Microsoft, and things got really squirrel-like. I don't trust the company that owns SUSE, which makes it a non-choice for our business. It also doesn't seem to have as big of a market share as Debian/RHEL here in the USA, which makes getting support for it more difficult.
Slackware - very niche, very bare bones. But I'm not sure you could call up any local Linux shop and easily get support for it like you could with RHEL / Debian. Well, you probably would, as past certain things "Linux is Linux".
Gentoo - Very minimalistic, very powerful. But the distro maintainers have horrid QA and upgrades frequently break things. As much as I liked portage, and being able to strip a server down to minimal packages, Gentoo will be relegated to the hobbyist category until they get serious about keeping packages stable. (We started our Linux migration with a few Gentoo boxes w/ Xen, they've all been phased in favor of CentOS. I got tired of fixing things after upgrades.)
Remember, with 2008 R2 MS went down the ubuntu/redhat/bsd-ish route by allowing you to install a VERY basic server with almost nothing on it, then add packages one by one. A hyper-V cluster member will not have a full IE, full IIS, Office, Flash, etc. Nor will it likely have any file shares open.
If you're talking about the "Server Core" style in Windows Server 2008... it's extremely rudimentary. By default, there is no GUI shell in Server Core. No Desktop. No Start Menu. Just a command line.
Sounds wonderful right? Well, except that most packages that you would need to install on Server Core are still configured with MSCs or GUI widgets. So you have to learn arcane commands (worse then Linux) that bring up the GUI widget or MSC interface that lets you do the configuration. Configuration of the server is still kept in binary blobs instead of plain text files, so you can't choose to administer the server by tweaking config files with a text editor either. You're also still stuck using RDP to administer the server instead of being able to SSH in (and easily keep a log file of what you changed).
If you're considering deploying Server Core in any situation where you don't have a team of at least 3 people already, who are all technically adept, and you're going to be deploying at least a dozen of these things - then you're going to spend tons of time just figuring out how to configure something.
(We do Linux server admin via SSH and keep every config file versioned using FSVS. Server Core doesn't even come close to that power.)
Unstable Windows installations *these days* are almost always the fault of 3rd party software. I've had unstable Linux installs because of third party software. We're talking about virtualization, which does not require a great deal of third party software to mess things up.
Or hardware. Odds are pretty high that if you're experiencing frequent lockups that there's bad hardware or maybe bad drivers. And for infrequent lockups, I always start by looking closely at RAM / CPU. Maybe the system is OC'd too much. Maybe the RAM is not really what it says on the package and it can't handle the timings. Maybe the disk has a bad sector and is causing the system to wait while it attempts to re-read the sector a dozen times.
Start with MemTest86+, then try Prime95 in torture test mode for a day or two.
You can get ESX free as well, and there are plenty of tools to help manage it. Both VMware and Hyper-V charge for features as you move up the stack. Maybe VMware is more expensive for the advanced features (which may be because they support doing things Hyper-V can't yet), but...
Yeah, when we priced out VMWare a few years ago it was so expensive that it dwarfed the price of the actual server. Which makes it a no-go for small/medium businesses if you want the useful features like moving VMs between servers with no downtime. It's slick, but not worth the few thousand dollars per server price tag.
At that price, small/medium businesses have to go with other solutions like Xen/KVM.
Or FSVS, if you're willing to store extra copies of the actual binaries. (With the bonus that now all of your config files are version controlled and you can easily do a diff to see what changed.)
And even the old floppy disks relied solely on the good behavior of the host system not to write to the disk.
Sure, there was the notch that you either filled in or opened up, and there might have been circuitry inside the drive to detect that and actually prevent writes. But an attacker could have easily covered the lens (in 3.5" drives) or rewired the circuits on the drive (on 5.25" drives). Now, those two attacks require the complicity of the user - but now it might just be a JUMP instruction in the device driver that can easily be overwritten by malware.
This was always the paradox of ebooks. By every measure, ebooks should have the first thing that easily came to the computer. Files sizes were small and text was one of the first things reasonably conquered by computers. In the early days, sound cards were necessary to play music, video files were just goddamned intensive.... and yet as a medium, books came last after everything else.
Personally, I lay that at the feet of device design.
A paper book, requires no batteries, works in a wide range of light conditions, can be traded / loaned without technical issues (other then language), easy random access and bookmarking / highlighting / notes in the margin, etc.
Other things like music & movies/TV always required technology in some shape or form unless it was a live performance. They are also passive (watching a movie / TV show) or background activities (as music often is).
So for electronic books to take off, you need a device that is easy to use, easy to read, and has most of the same advantages as a dead-tree book. It turned out that cell phone / smart phone / Palm screens were too small or too expensive for most people. Not many people like to have a laptop in their lap or sit at a desk to read for long periods either. Very few electronic screens (other then pure B&W crystal LCDs from the 90s) were readable in bright sunlight. Or the devices that were designed as book readers were priced over $500 which is too expensive for most.
I picked up a Sony eReader about 3 years ago and immediately fell in love with it for fiction/leisure reading (cover-to-cover, very little random access). And device prices have dropped drastically since then which helps with mass market uptake. Now we just have to wait for the format wars to settle out and for a format to come to the foreground like MP3 did for audio. That probably be ePub which is the most open of the formats.
(Baen has been a good corporate citizen in all this. Decent prices, no DRM, and a variety of formats.)
Acronis True Image, Norton Ghost, or even NTFSClone. The first costs money, but is probably the easiest solution for a Windows laptop. I haven't used Ghost in forever and NTFSClone works well but can be a bit arcane.
All of them make it dead simple to swap out an older HD for a newer SSD. Put the old HD in an external USB case, install the new HD, clone from the old HD to the new HD. (Going from USB->Internal tends to be faster then Internal->USB. So do the drive swap first and use a boot CD to run Acronis/NTFSClone.)
(As a bonus you can put the old HD on the shelf as a long-term backup in case something happens. And Acronis is worth the money for the minimal hassle and being able to make images from a running system.)
- 2nd CPU core. I now consider this a must-have, even if the multi-core CPU is slower per-core then a single-core CPU. Every time I use a single-core machine, I'm reminded just how bad WinXP is at scheduling. (Ubuntu isn't as bad, but that 2nd core still really helps a lot.) It's also why I haven't picked up a netbook yet, dual-core is a must-have for me.
- Enough RAM to run everything you need to without swapping. For WinXP users, that's generally in the 1-2GB range. The office machines that we bought a few years ago all ended up with 2GB with room to improve them to ~3.5GB (32bit WinXP). For me personally, 1GB was way too tight and I'm much happier with 3GB where I can keep things open.
- SSD. For an older laptop, this can breathe even more life into an aging system if you do a lot of task swapping / file manipulation. My work laptop (3.5 years old, dual-core, 3GB, Thinkpad T61p) was on its last legs this year because the hard drive was just such a bottleneck. So I dropped a SSD in and suddenly it's a lot more pleasurable to use. Now that the 128GB SSDs are under $200, I don't ever plan on using an old-style rotational drive as the primary drive in any new systems. Use the SSD for the OS and primary data, use rotational drives for bulk storage. (The Thinkpad has the advantage that I can swap out the optical bay for a 2nd SATA drive.)
The summary claims that one rule is to pay for more RAM over better processor. That sounds like poor advice for at least three reasons: 1) RAM can usually be user-upgraded later, while the processor usually can't be; 2) RAM is cheaper than the processor; 3) some OEMs overcharge for RAM upgrades (cough, Apple). Plus, it is dubious to claim processors are usually fast enough for most people. All told, whoever offered that suggestion wasn't thinking very soundly.
That advice has been true for almost 20 years at this point.
As soon as the user runs past the physical memory limit in the machine and it starts to swap heavily, performance goes in the toilet and that super-fast CPU ends up sitting idle. I would always tell people to trade down to a slightly slower CPU and use the money on more RAM instead.
Manufacturers have *always* under-spec'd the base level of RAM in the system. Back in the NT4 days, it was not uncommon to find systems being sold with only 64MB of RAM (which is barely enough to boot and run 1 application). If you went with the default, you'd find the machine getting slow after you installed a few applications and started using the machine regularly. Upgrade that to 128MB or 192MB and now you had a machine that would run very well and last quite a few years.
WinXP (c2002) was typically sold with 128MB and sometimes 256MB of RAM. The user would be in their swap file from day two onward and wonder why the machine was so slow. Upgrading them to 512MB was not difficult (and 1GB was common by 2003-2004) and gave them the head room to run multiple applications at the same time.
We're only just starting to get to the point where memory has gotten cheap enough and the base level modules large enough (4GB is now common as a base install) that the operating system won't be starved for memory from day one. But I'm sure there are vendors out there still selling Win7 systems with 1GB or 2GB of RAM.
But the core issue remains - once you run out of available RAM and start swapping, that super fast CPU does you no good.
Keep in mind that once you go past about 1920x1200, DVI single-link cables and circuitry no longer work. DVI bandwidth is only capable of driving 2.75M pixels @ 60Hz. Which puts the limit up around 2098x1311 for a 16:10 display over DVI single-link. And a 1920x1200 @ 60Hz display is 2.3M pixels. The 2560x1600 is 4.1M pixels.
Enough to justify the cost difference? Probably not, but it's one factor that goes into the reason why prices take a jump upward once you get past 1920x1200@60Hz displays.
You can easily find 15" laptops with 1920x1080 displays now.
1680x1050 was popular a few years ago (such as my Thinkpad T61p).
And the price difference is often only $100.
You probably won't find 10-14" displays with that high of a resolution though, very few glass manufs make stuff with pixels smaller then 0.245mm, and the 0.27-0.285 is a popular size for desktop LCDs. And frankly, 1680x1050 on a 15" display (or even a 19") is a bit on the small side for normal work - I prefer that resolution on a 22" monitor.
(Large format LCDs start at around 0.5mm per pixel, and there's a few 0.31mm oddballs out there.)
Even with all of that, you're still going to end up with the user's profile getting infected. It won't (without a priv escalation) be able to infect the entire machine - but even up-to-date systems, with A/V, content filtering, whatever is susceptible to drive-by exploits (generally Flash / PDF / Javascript code in web pages) that the A/V software and filters don't yet know about.
So unless you're doing a whitelist of "sites you're allowed to browse" as well as whitelisting what applications are allowed to run at all on the machine as well as heavy GPO editing to lock the system down even farther... you're still going to seen infections.
When your provider sells you a connection of, say, 1.544mb/sec - thats megaBITS not megabytes. You need to divide by 8 to come up with your connection speed in bytes.
Dividing by 10 to calculate bytes is generally more accurate for net throughput numbers due to transport overhead (checksums, parity bits, packet headers, etc.).
It has gotten better over the years, overhead has gone down. The "divide by 10" was mostly during the dial-up age. Now it's probably only a 5-10% overhead.
But it's easier to just drop a digit when dividing by 10.
(And nobody complains much when a transfer finishes sooner then the hand calculation indicated.)
Yes, but the issue with the CRC on CDs is that often any damage is only detected *after* the RS encoding on the CD sectors has already failed. At which point, you're basically hosed unless you have a 2nd copy of the data that you know is good. Very few users will use CD/DVD testing software that is smart enough to check how much actual damage is being hidden by the RS coding on the disk itself.
The advantage of PAR, PAR2 and other parity-like systems is that they sit on top of that layer and the user can take action after the corruption is discovered. Plus you can tune the ratio of data:parity to match your paranoia level (mine is about 10% parity on top of the data).
In addition, optical media suffering from rot tend to fail linearly. First it will be a few sectors, then a few more, etc. Which gives you a recovery window of sorts where if you catch it early enough, before the damage exceeds the amount of parity data that you have, you can recover all of the data.
The ballpark I used back in the 56k days was an average of about 15MB/hr over dial-up. 33.6k was 12MB/hr and 56k was 20MB/hr, but you usually got speeds / throughput somewhere between those two numbers. In comparison, a T1 (1.5Mbps) can push about 500MB/hr in each direction.
PAR2 is still useful for archive media (CDs, DVDs, writable Blu-Ray, or removable disks). Since it doesn't care about the source file format, it works in more situations and the data can be left on the disk, usable as-is without decompression.
Combine PAR2 + dd_rescue and as long as the disk isn't completely non-functional, you have good odds of recovering data (if you have enough repair blocks). Setting aside 5-15% of the disk capacity for recovery data is not a big deal in most situations and will protect against most forms of loss where sectors go bad or bits get flipped.
The only downsides are (a) QuickPAR is not multi-threaded and (b) speed as modern CPUs can only calculate the recovery data at a rate of about 1 hour for 4GB of data with 20% redundancy and (c) the tools don't understand sub-directories. Unfortunately, the author of QuickPAR never managed to finish a PAR3 design which might have been a lot faster, or to fix the issues with sub-directory handling during repairs.
PKZIP multi-part was extremely rudimentary and you ended up with stupidity (due to the underlying design of a ZIP file) where you had to insert the last disk first.
We used to do backups to multiple floppies with PKZIP, it was not pretty.
7Z has all the shell integration that WinRar has (and has for a number of years). We switched away from WinZIP / WinRAR as soon as that happened. As a bonus, it's one less set of licenses that I have to track.
(Plus the 7Z package is open-source, so it's widely available.)
CD jukeboxes didn't really hit the ground until around 1990.
And in 1990, the basic multi-changer was a plastic cartridge type device that you loaded with 5/7/10 CDs and then inserted it into the unit. Very similar to the modern units that you install in the car. There were also designs around 1990 that let you put a few dozen (hundred?) into a unit that fit on your A/V shelf. Those systems cost an awful lot.
For a modern version that fits on an A/V shelf, look for the Sony CDP CX355, which holds 300 CDs. Of course, the carousel type units are still eating CDs, so nothing has changed in 20 years. There are also 100/200 disc models out there from other manufacturers.
(CDs were just getting popular by about 1986-1987. It took a few years after they came out in '82. The players in '86 were still very expensive.)
I keep all my music on a few 2.5" externals now, with a backup on a 3.5 that stays in a safe location. Optical discs no longer work for archival purposes when a 500GB drive (for $60) can hold over 100 DVD-R equivalents, and Blu-Ray media is still up around $0.50-$1.00 per GB.
In 2011, if you're still feeding 100Mbps to the desk for brand new installs, you're being incredibly cheap. 1Gbps ports are no longer that expensive. It's a difference of something like $10 vs $17 per port between 100Mbps and 1Gbps, and getting a decent 100Mbps switch is becoming more difficult. Hell, that statement was true going back as far as 2008 or 2009, when the lower end 24-port gigabit switches first dropped below $500. Not hard now to get a "smart" 48-port gigabit switch for about $800 ($17/port). It won't have all the management features of the high-end switches, but it also won't be a slouch.
As soon as you start moving gigabyte sized files around, you've just passed the point where 100Mbps still makes sense. Assuming you'll hit 70% of capacity, 100Mbps gives you 7 MB/s which means 146 seconds to transfer a GB worth of data. Switch to gigabit and even if you only manage 30MB/s, you've cut that 146 seconds down to 34 seconds.
(I generally see speeds in the 45-50 MB/s range on standard sized frame 1Gbps ethernet talking to SAMBA/CIFS. Which is a hell of a lot better then topping out at only 7-8 MB/s for 100Mbps ethernet. Jumbo frames would help a bit.)
In comparison, USB2 typically tops out at 20-30 MB/s, based on real world usage. If all you have is 100Mbps, then it's faster to copy the files to a USB2 device, walk them across the room, and plug them into the other system then to feed it over the network.
As soon as your users start touching files larger then 20MB, they're going to notice.
The password for an email account or website can be transmitted encrypted, so that even if someone intercepts the communication, they don't know the password. This may not *always* be the case, but its the intent of the systems design in most cases.
It could be, but probably 95% of all mail servers out there still fail to do SSL because the admins can't figure out SSL certificates. Or they use a simple self-signed cert, which is fairly useless at preventing MiTM attacks (you end up talking over an encrypted channel to your attacker).
Slashdot Mirror
Ext4 trumps Ext3 as soon as you start dealing with files larger then a few hundred MB. When you delete a big file under Ext3, it spends ages and ages playing cleanup. Delete a big file under Ext4 and it takes under a second.
/boot and probably not the root partition either. When it comes to /boot and root, I prefer to stay very conservative and Ext4 is still a bit young (another 2 years and I'll be more comfortable with it). But I use ext4 all the time on data partitions.
Personally, I don't see a reason to use Ext4 yet on
Stealing your monitor is a symptom of deeper dysfunction. It would tell me that there is an endemic problem at the company that is not conducive to actually getting work done there.
Exactly. If they're willing to screw with a developer like that on an whim, then they're probably doing other stupid and trivial things instead of focusing on the real problems. It's a symptom of far deeper issues at the company and of management that has lost touch with reality in the trenches.
Why not a more popular distro not to disparage Cent.....
If you're doing Linux for business, there are only a few games in town.
RHEL - and CentOS / Scientific Linux are both based off of RHEL. Any book you pickup off the shelf that talks about RHEL 5/6 will apply to CentOS / Scientific Linux. Any software package out there which runs on RHEL 5/6 will run on CentOS. Which makes CentOS a very good distro for testing the waters and learning how Red Hat does things before you start paying for RHEL subscriptions. Red Hat is the big daddy in commercial / business Linux. Getting support from a local shop is a sure thing (as long as they have some Linux folks on staff).
Debian / Ubuntu - You can find books for it, but my impression over the years is that Canonical is really focused more on the desktop market then the server market. The primary reason for running a server on Ubuntu LTS would be because you're more used to how Debian does things then how RHEL does things. Big enough market share that a local Linux shop should know how to fix it.
SuSE - This used to be a great choice, until Novell bought them, did the patent deal with Microsoft, and things got really squirrel-like. I don't trust the company that owns SUSE, which makes it a non-choice for our business. It also doesn't seem to have as big of a market share as Debian/RHEL here in the USA, which makes getting support for it more difficult.
Slackware - very niche, very bare bones. But I'm not sure you could call up any local Linux shop and easily get support for it like you could with RHEL / Debian. Well, you probably would, as past certain things "Linux is Linux".
Gentoo - Very minimalistic, very powerful. But the distro maintainers have horrid QA and upgrades frequently break things. As much as I liked portage, and being able to strip a server down to minimal packages, Gentoo will be relegated to the hobbyist category until they get serious about keeping packages stable. (We started our Linux migration with a few Gentoo boxes w/ Xen, they've all been phased in favor of CentOS. I got tired of fixing things after upgrades.)
Remember, with 2008 R2 MS went down the ubuntu/redhat/bsd-ish route by allowing you to install a VERY basic server with almost nothing on it, then add packages one by one. A hyper-V cluster member will not have a full IE, full IIS, Office, Flash, etc. Nor will it likely have any file shares open.
If you're talking about the "Server Core" style in Windows Server 2008... it's extremely rudimentary. By default, there is no GUI shell in Server Core. No Desktop. No Start Menu. Just a command line.
Sounds wonderful right? Well, except that most packages that you would need to install on Server Core are still configured with MSCs or GUI widgets. So you have to learn arcane commands (worse then Linux) that bring up the GUI widget or MSC interface that lets you do the configuration. Configuration of the server is still kept in binary blobs instead of plain text files, so you can't choose to administer the server by tweaking config files with a text editor either. You're also still stuck using RDP to administer the server instead of being able to SSH in (and easily keep a log file of what you changed).
If you're considering deploying Server Core in any situation where you don't have a team of at least 3 people already, who are all technically adept, and you're going to be deploying at least a dozen of these things - then you're going to spend tons of time just figuring out how to configure something.
(We do Linux server admin via SSH and keep every config file versioned using FSVS. Server Core doesn't even come close to that power.)
Unstable Windows installations *these days* are almost always the fault of 3rd party software. I've had unstable Linux installs because of third party software. We're talking about virtualization, which does not require a great deal of third party software to mess things up.
Or hardware. Odds are pretty high that if you're experiencing frequent lockups that there's bad hardware or maybe bad drivers. And for infrequent lockups, I always start by looking closely at RAM / CPU. Maybe the system is OC'd too much. Maybe the RAM is not really what it says on the package and it can't handle the timings. Maybe the disk has a bad sector and is causing the system to wait while it attempts to re-read the sector a dozen times.
Start with MemTest86+, then try Prime95 in torture test mode for a day or two.
You can get ESX free as well, and there are plenty of tools to help manage it. Both VMware and Hyper-V charge for features as you move up the stack. Maybe VMware is more expensive for the advanced features (which may be because they support doing things Hyper-V can't yet), but...
Yeah, when we priced out VMWare a few years ago it was so expensive that it dwarfed the price of the actual server. Which makes it a no-go for small/medium businesses if you want the useful features like moving VMs between servers with no downtime. It's slick, but not worth the few thousand dollars per server price tag.
At that price, small/medium businesses have to go with other solutions like Xen/KVM.
Or FSVS, if you're willing to store extra copies of the actual binaries. (With the bonus that now all of your config files are version controlled and you can easily do a diff to see what changed.)
And even the old floppy disks relied solely on the good behavior of the host system not to write to the disk.
Sure, there was the notch that you either filled in or opened up, and there might have been circuitry inside the drive to detect that and actually prevent writes. But an attacker could have easily covered the lens (in 3.5" drives) or rewired the circuits on the drive (on 5.25" drives). Now, those two attacks require the complicity of the user - but now it might just be a JUMP instruction in the device driver that can easily be overwritten by malware.
I would be more offended that the monitor was simply taken away, with no notice, no discussion, no attempt to pry it away by reason first.
That just smacks of "we don't care about you, chump".
If the company is so tight fisted that they can't splurge $150-$180 for another monitor, then it's time to move to another company.
This was always the paradox of ebooks. By every measure, ebooks should have the first thing that easily came to the computer. Files sizes were small and text was one of the first things reasonably conquered by computers. In the early days, sound cards were necessary to play music, video files were just goddamned intensive.... and yet as a medium, books came last after everything else.
Personally, I lay that at the feet of device design.
A paper book, requires no batteries, works in a wide range of light conditions, can be traded / loaned without technical issues (other then language), easy random access and bookmarking / highlighting / notes in the margin, etc.
Other things like music & movies/TV always required technology in some shape or form unless it was a live performance. They are also passive (watching a movie / TV show) or background activities (as music often is).
So for electronic books to take off, you need a device that is easy to use, easy to read, and has most of the same advantages as a dead-tree book. It turned out that cell phone / smart phone / Palm screens were too small or too expensive for most people. Not many people like to have a laptop in their lap or sit at a desk to read for long periods either. Very few electronic screens (other then pure B&W crystal LCDs from the 90s) were readable in bright sunlight. Or the devices that were designed as book readers were priced over $500 which is too expensive for most.
I picked up a Sony eReader about 3 years ago and immediately fell in love with it for fiction/leisure reading (cover-to-cover, very little random access). And device prices have dropped drastically since then which helps with mass market uptake. Now we just have to wait for the format wars to settle out and for a format to come to the foreground like MP3 did for audio. That probably be ePub which is the most open of the formats.
(Baen has been a good corporate citizen in all this. Decent prices, no DRM, and a variety of formats.)
Acronis True Image, Norton Ghost, or even NTFSClone. The first costs money, but is probably the easiest solution for a Windows laptop. I haven't used Ghost in forever and NTFSClone works well but can be a bit arcane.
All of them make it dead simple to swap out an older HD for a newer SSD. Put the old HD in an external USB case, install the new HD, clone from the old HD to the new HD. (Going from USB->Internal tends to be faster then Internal->USB. So do the drive swap first and use a boot CD to run Acronis/NTFSClone.)
(As a bonus you can put the old HD on the shelf as a long-term backup in case something happens. And Acronis is worth the money for the minimal hassle and being able to make images from a running system.)
The big (3) improvements:
- 2nd CPU core. I now consider this a must-have, even if the multi-core CPU is slower per-core then a single-core CPU. Every time I use a single-core machine, I'm reminded just how bad WinXP is at scheduling. (Ubuntu isn't as bad, but that 2nd core still really helps a lot.) It's also why I haven't picked up a netbook yet, dual-core is a must-have for me.
- Enough RAM to run everything you need to without swapping. For WinXP users, that's generally in the 1-2GB range. The office machines that we bought a few years ago all ended up with 2GB with room to improve them to ~3.5GB (32bit WinXP). For me personally, 1GB was way too tight and I'm much happier with 3GB where I can keep things open.
- SSD. For an older laptop, this can breathe even more life into an aging system if you do a lot of task swapping / file manipulation. My work laptop (3.5 years old, dual-core, 3GB, Thinkpad T61p) was on its last legs this year because the hard drive was just such a bottleneck. So I dropped a SSD in and suddenly it's a lot more pleasurable to use. Now that the 128GB SSDs are under $200, I don't ever plan on using an old-style rotational drive as the primary drive in any new systems. Use the SSD for the OS and primary data, use rotational drives for bulk storage. (The Thinkpad has the advantage that I can swap out the optical bay for a 2nd SATA drive.)
The summary claims that one rule is to pay for more RAM over better processor. That sounds like poor advice for at least three reasons: 1) RAM can usually be user-upgraded later, while the processor usually can't be; 2) RAM is cheaper than the processor; 3) some OEMs overcharge for RAM upgrades (cough, Apple). Plus, it is dubious to claim processors are usually fast enough for most people. All told, whoever offered that suggestion wasn't thinking very soundly.
That advice has been true for almost 20 years at this point.
As soon as the user runs past the physical memory limit in the machine and it starts to swap heavily, performance goes in the toilet and that super-fast CPU ends up sitting idle. I would always tell people to trade down to a slightly slower CPU and use the money on more RAM instead.
Manufacturers have *always* under-spec'd the base level of RAM in the system. Back in the NT4 days, it was not uncommon to find systems being sold with only 64MB of RAM (which is barely enough to boot and run 1 application). If you went with the default, you'd find the machine getting slow after you installed a few applications and started using the machine regularly. Upgrade that to 128MB or 192MB and now you had a machine that would run very well and last quite a few years.
WinXP (c2002) was typically sold with 128MB and sometimes 256MB of RAM. The user would be in their swap file from day two onward and wonder why the machine was so slow. Upgrading them to 512MB was not difficult (and 1GB was common by 2003-2004) and gave them the head room to run multiple applications at the same time.
We're only just starting to get to the point where memory has gotten cheap enough and the base level modules large enough (4GB is now common as a base install) that the operating system won't be starved for memory from day one. But I'm sure there are vendors out there still selling Win7 systems with 1GB or 2GB of RAM.
But the core issue remains - once you run out of available RAM and start swapping, that super fast CPU does you no good.
Keep in mind that once you go past about 1920x1200, DVI single-link cables and circuitry no longer work. DVI bandwidth is only capable of driving 2.75M pixels @ 60Hz. Which puts the limit up around 2098x1311 for a 16:10 display over DVI single-link. And a 1920x1200 @ 60Hz display is 2.3M pixels. The 2560x1600 is 4.1M pixels.
Enough to justify the cost difference? Probably not, but it's one factor that goes into the reason why prices take a jump upward once you get past 1920x1200@60Hz displays.
You can easily find 15" laptops with 1920x1080 displays now.
1680x1050 was popular a few years ago (such as my Thinkpad T61p).
And the price difference is often only $100.
You probably won't find 10-14" displays with that high of a resolution though, very few glass manufs make stuff with pixels smaller then 0.245mm, and the 0.27-0.285 is a popular size for desktop LCDs. And frankly, 1680x1050 on a 15" display (or even a 19") is a bit on the small side for normal work - I prefer that resolution on a 22" monitor.
(Large format LCDs start at around 0.5mm per pixel, and there's a few 0.31mm oddballs out there.)
Even with all of that, you're still going to end up with the user's profile getting infected. It won't (without a priv escalation) be able to infect the entire machine - but even up-to-date systems, with A/V, content filtering, whatever is susceptible to drive-by exploits (generally Flash / PDF / Javascript code in web pages) that the A/V software and filters don't yet know about.
So unless you're doing a whitelist of "sites you're allowed to browse" as well as whitelisting what applications are allowed to run at all on the machine as well as heavy GPO editing to lock the system down even farther... you're still going to seen infections.
When your provider sells you a connection of, say, 1.544mb/sec - thats megaBITS not megabytes. You need to divide by 8 to come up with your connection speed in bytes.
Dividing by 10 to calculate bytes is generally more accurate for net throughput numbers due to transport overhead (checksums, parity bits, packet headers, etc.).
It has gotten better over the years, overhead has gone down. The "divide by 10" was mostly during the dial-up age. Now it's probably only a 5-10% overhead.
But it's easier to just drop a digit when dividing by 10.
(And nobody complains much when a transfer finishes sooner then the hand calculation indicated.)
Yes, but the issue with the CRC on CDs is that often any damage is only detected *after* the RS encoding on the CD sectors has already failed. At which point, you're basically hosed unless you have a 2nd copy of the data that you know is good. Very few users will use CD/DVD testing software that is smart enough to check how much actual damage is being hidden by the RS coding on the disk itself.
The advantage of PAR, PAR2 and other parity-like systems is that they sit on top of that layer and the user can take action after the corruption is discovered. Plus you can tune the ratio of data:parity to match your paranoia level (mine is about 10% parity on top of the data).
In addition, optical media suffering from rot tend to fail linearly. First it will be a few sectors, then a few more, etc. Which gives you a recovery window of sorts where if you catch it early enough, before the damage exceeds the amount of parity data that you have, you can recover all of the data.
The ballpark I used back in the 56k days was an average of about 15MB/hr over dial-up. 33.6k was 12MB/hr and 56k was 20MB/hr, but you usually got speeds / throughput somewhere between those two numbers. In comparison, a T1 (1.5Mbps) can push about 500MB/hr in each direction.
PAR2 is still useful for archive media (CDs, DVDs, writable Blu-Ray, or removable disks). Since it doesn't care about the source file format, it works in more situations and the data can be left on the disk, usable as-is without decompression.
Combine PAR2 + dd_rescue and as long as the disk isn't completely non-functional, you have good odds of recovering data (if you have enough repair blocks). Setting aside 5-15% of the disk capacity for recovery data is not a big deal in most situations and will protect against most forms of loss where sectors go bad or bits get flipped.
The only downsides are (a) QuickPAR is not multi-threaded and (b) speed as modern CPUs can only calculate the recovery data at a rate of about 1 hour for 4GB of data with 20% redundancy and (c) the tools don't understand sub-directories. Unfortunately, the author of QuickPAR never managed to finish a PAR3 design which might have been a lot faster, or to fix the issues with sub-directory handling during repairs.
PKZIP multi-part was extremely rudimentary and you ended up with stupidity (due to the underlying design of a ZIP file) where you had to insert the last disk first.
We used to do backups to multiple floppies with PKZIP, it was not pretty.
7Z has all the shell integration that WinRar has (and has for a number of years). We switched away from WinZIP / WinRAR as soon as that happened. As a bonus, it's one less set of licenses that I have to track.
(Plus the 7Z package is open-source, so it's widely available.)
CD jukeboxes didn't really hit the ground until around 1990.
And in 1990, the basic multi-changer was a plastic cartridge type device that you loaded with 5/7/10 CDs and then inserted it into the unit. Very similar to the modern units that you install in the car. There were also designs around 1990 that let you put a few dozen (hundred?) into a unit that fit on your A/V shelf. Those systems cost an awful lot.
For a modern version that fits on an A/V shelf, look for the Sony CDP CX355, which holds 300 CDs. Of course, the carousel type units are still eating CDs, so nothing has changed in 20 years. There are also 100/200 disc models out there from other manufacturers.
(CDs were just getting popular by about 1986-1987. It took a few years after they came out in '82. The players in '86 were still very expensive.)
I keep all my music on a few 2.5" externals now, with a backup on a 3.5 that stays in a safe location. Optical discs no longer work for archival purposes when a 500GB drive (for $60) can hold over 100 DVD-R equivalents, and Blu-Ray media is still up around $0.50-$1.00 per GB.
In 2011, if you're still feeding 100Mbps to the desk for brand new installs, you're being incredibly cheap. 1Gbps ports are no longer that expensive. It's a difference of something like $10 vs $17 per port between 100Mbps and 1Gbps, and getting a decent 100Mbps switch is becoming more difficult. Hell, that statement was true going back as far as 2008 or 2009, when the lower end 24-port gigabit switches first dropped below $500. Not hard now to get a "smart" 48-port gigabit switch for about $800 ($17/port). It won't have all the management features of the high-end switches, but it also won't be a slouch.
As soon as you start moving gigabyte sized files around, you've just passed the point where 100Mbps still makes sense. Assuming you'll hit 70% of capacity, 100Mbps gives you 7 MB/s which means 146 seconds to transfer a GB worth of data. Switch to gigabit and even if you only manage 30MB/s, you've cut that 146 seconds down to 34 seconds.
(I generally see speeds in the 45-50 MB/s range on standard sized frame 1Gbps ethernet talking to SAMBA/CIFS. Which is a hell of a lot better then topping out at only 7-8 MB/s for 100Mbps ethernet. Jumbo frames would help a bit.)
In comparison, USB2 typically tops out at 20-30 MB/s, based on real world usage. If all you have is 100Mbps, then it's faster to copy the files to a USB2 device, walk them across the room, and plug them into the other system then to feed it over the network.
As soon as your users start touching files larger then 20MB, they're going to notice.
The password for an email account or website can be transmitted encrypted, so that even if someone intercepts the communication, they don't know the password. This may not *always* be the case, but its the intent of the systems design in most cases.
It could be, but probably 95% of all mail servers out there still fail to do SSL because the admins can't figure out SSL certificates. Or they use a simple self-signed cert, which is fairly useless at preventing MiTM attacks (you end up talking over an encrypted channel to your attacker).