Home Server Projects
The best servers are those that don't appear to be servers at all.
Tuesday 30 June 2009, 2:13 AM
Windows 7 on a Read-only Flash Drive?
Considering that the price of a 4GB USB flash drive has been as low as 5 dollars on close-out specials, financially it wouldn't make sense UNLESS Microsoft decides to go into the Flash RAM business. The full Win 7 Ultra install DVD is 3.9GB large. Even stripped down it will be better than 2 GB so 4GB sounds like the right size for the Win7 USB Flash drive.
The price per Win 7 Starter license I was seeing for netbooks was around $3 each license in volume. That would mean the price of USB flash drives need to be much lower than that to make it possible to do economically.
Then think about what it would take to make a mass assembly line operation to clone the image onto USB flash drives thousands or millions of times! I suppose you could use something like Ghost to make 128 copies at a time through a massively wicked USB "router". The time it would take to plug in and out all the USB flash drives would be tremendous!
Actually its been done sort of by SanDisk. They have a number of U3-based USB flash drives that have a "read-only" partition that plays like a CDR. When its up and running, the USB flash drive looks like 2 drives, a CDR and a HDD. It has some software on it that makes Windows XP Pro work better when it comes time to unmount the USB flash drive. Actually fairly nice.
I don't know how they are doing it. There is likely some sort of JTAG software interface hidden in "plain sight" connected to the USB signal pins. With special commands to special hidden registers, part of the drive is marked as read only.
If you delete the partitions you kill the U3 interface. If you are careful, the read-write partition can be formatted in NTFS and it coexists nicely with the U3 CD-ROM type partition.
I could see SanDisk turning out millions of Win7 USB flash disks on the U3 read-only "CDR"-like partition for Microsoft. But to do it for a price the netbook owners will be willing to pay, hmmm, I don't know about that.
I think most likely Microsoft is going to make the netbook owners that want it on their old netbooks eat the price. It probably won't be available in anything other than download ISO form. Buy your own USB drive. Buy the download on Shop Microsoft. Copy the image over to the Flash. Use an updated tool based on ufdprep.exe to prep the flash drive. It will have a file converter program that will open enough of the image to mount a bootstrap loader on the netbook and then mount and install the ISO onto the system hard drive or internal flash drive or SSD. (Sounds like a job for a really small install of Linux!)
Don't expect this function above to be available until late in the year until after the retailers get an idea how netbook sales with Win 7 pre-installed are doing. Microsoft will think of the USB based install to be a loss-leader, a way to get users hooked on Windows 7 so they'll buy it pre-installed on their next computer or as shrink-wrap for their old desktops or regular laptops. That's what the Win7 RC is by the way, a form of digital "crack" to get the neighborhood geeks all hooked. Call it Windows 7 Express!
Its prettier than XP Pro but not faster. It is definitely better than Vista on substantial hardware, like a dual core CPU with 4 GB of RAM.
I have no clue how well it will work stripped down enough to work on a netbook. I've already had my taste of "Windows-7-crack". Would I buy it?
No.
The price per Win 7 Starter license I was seeing for netbooks was around $3 each license in volume. That would mean the price of USB flash drives need to be much lower than that to make it possible to do economically.
Then think about what it would take to make a mass assembly line operation to clone the image onto USB flash drives thousands or millions of times! I suppose you could use something like Ghost to make 128 copies at a time through a massively wicked USB "router". The time it would take to plug in and out all the USB flash drives would be tremendous!
Actually its been done sort of by SanDisk. They have a number of U3-based USB flash drives that have a "read-only" partition that plays like a CDR. When its up and running, the USB flash drive looks like 2 drives, a CDR and a HDD. It has some software on it that makes Windows XP Pro work better when it comes time to unmount the USB flash drive. Actually fairly nice.
I don't know how they are doing it. There is likely some sort of JTAG software interface hidden in "plain sight" connected to the USB signal pins. With special commands to special hidden registers, part of the drive is marked as read only.
If you delete the partitions you kill the U3 interface. If you are careful, the read-write partition can be formatted in NTFS and it coexists nicely with the U3 CD-ROM type partition.
I could see SanDisk turning out millions of Win7 USB flash disks on the U3 read-only "CDR"-like partition for Microsoft. But to do it for a price the netbook owners will be willing to pay, hmmm, I don't know about that.
I think most likely Microsoft is going to make the netbook owners that want it on their old netbooks eat the price. It probably won't be available in anything other than download ISO form. Buy your own USB drive. Buy the download on Shop Microsoft. Copy the image over to the Flash. Use an updated tool based on ufdprep.exe to prep the flash drive. It will have a file converter program that will open enough of the image to mount a bootstrap loader on the netbook and then mount and install the ISO onto the system hard drive or internal flash drive or SSD. (Sounds like a job for a really small install of Linux!)
Don't expect this function above to be available until late in the year until after the retailers get an idea how netbook sales with Win 7 pre-installed are doing. Microsoft will think of the USB based install to be a loss-leader, a way to get users hooked on Windows 7 so they'll buy it pre-installed on their next computer or as shrink-wrap for their old desktops or regular laptops. That's what the Win7 RC is by the way, a form of digital "crack" to get the neighborhood geeks all hooked. Call it Windows 7 Express!
Its prettier than XP Pro but not faster. It is definitely better than Vista on substantial hardware, like a dual core CPU with 4 GB of RAM.
I have no clue how well it will work stripped down enough to work on a netbook. I've already had my taste of "Windows-7-crack". Would I buy it?
No.
Sunday 28 June 2009, 8:13 PM
Constricted Internet Connections
The MJ event and the Iranian and Chinese governments' choke-holds on its citizen's Internet connections demonstrate in two different ways how vulnerable network endpoint access really is for the entire world's Internet users. The death of a single individual managed to do in some measure worldwide what paranoid governments attempt to achieve within their own borders. To the user within those borders the effect of remote server congestion is the same as local government deep packet inspection.
No Internet server is directly connected to an end-user, there is a lot of intervening hardware and software impeding the flow of data. Users at the mercy of their government's control of the network will experience varying degrees of responsiveness from various remote servers. Between governments and commercial vendors controlling or regulating Internet access and unusual events bringing large sections of the network to a crawl or a complete halt, there isn't a truly open, full-speed network connection for any Internet user in every server situation.
Compounding the troubles on the network itself, the existence of zombies in parasitic botnets that siphon away even more computing cycles on infected systems contribute to the data flow throttling. Even non-infected systems are affected due to the need to run malware protection software and the effects of DDOS attacks on popular network servers. Anti-malware running in any of the server computers or routers in the data paths that make up the network connection serving the end-user likewise impede the data flow.
For typical Internet users, the result is varying speeds of access to differently sourced servers. For any given connection speed to an ISP there will be an constriction of data that can vary due to:
where the data is sourced;
the connection management policy on the server (does it impose throttling, limit the number of connections from a single IP, etc);
the content of the data (in deep packet inspection) ;
the paranoia of the local government;
commercial interests of the ISP itself (for instance, does the data displayed come with ISP generated advertising?);
is the connection being dynamically shuffled through different paths by routers? (ie a tenuous routing or using something like TOR will cause more traffic delays);
the effects of anti-malware software hosted at the ISP necessary to prevent the user from complaining;
and finally: the effects of the anti-malware software on the user's computer.
From an engineering perspective this suggests the twin concepts of network-access “impedance” and “data device impedance”. A low network access impedance would be when the data servers are providing data at the highest speed possible up the limit of the user's connection speed. If the connection never matches the connection speed then there is the effect of the “network access impedance”. Most of the network access impedance causes are out of control of the User.
Within the User's computer, the speed at which the data can be utilized would be a measure of how fast the data downloaded is displayed on screen, played through the speakers, or saved to disk. The concept could be applied to broadband, other wireless and wired network connections. This impedance could be different for every connection or connection type established by the user but is likely to be similar for most connections.
Some would argue that “network impedance” is simply a measure of system bandwidth. The problem with labeling it as “bandwidth” does not take into account the fact that the User might not be paying for a fully unlimited speed connection or that connected servers might be deliberately throttling speed. Most ISPs offer slower connections at discounts. Users on a budget will buy slower service. In addition, the data sources or the equipment in between the User and the server might be deliberately delaying the data stream. Presumably two systems with the same data device impedance would have similar network service quality when accessing the Internet through the exact same ISP connection.
A measure of specific system impedance with a known set of applications running and anti-malware measures running might give a more useful figure of merit, the “data device impedance”, to the potential Customer/User. The ability of the device drivers, the design of the radio components in the wireless Ethernet adapter and the operating system to deliver the desired data product to the User will probably be the major components of this “data device impedance”. A good chip set with efficient drivers will have a low impedance.
The release of the Intel Atom provoked this line of reasoning because it is designed to be specifically a single threaded processor. The ATOM question is; “How does the processor fare when compared to multi-threaded processors when running Internet related applications?” A number of reviewers were comparing the ATOM to (in one case a quad core CPU !) low power Celerons, VIA CPUs etc. Even in cases where the CPU speed was the same or nearly so, the tests did not really get into the operating mode of the intended device for the ATOM, a pda or netbook processor.
In areas of electrical engineering, the most efficient transfer of power occurs when the impedance of the source (Internet connection) and the impedance of the power sink (netbook or desktop computer) match. This might be stretching the analogy too far. However it stands to reason that a User intending to buy a netbook for an networking environment he's most likely to encounter could save quite a bit of cash by buying enough CPU and OS capacity to match his probable networking environment and not more. Especially true if the networking is less than optimum. Along with CPU speed, a “data device impedance” might be a useful characteristic.
No Internet server is directly connected to an end-user, there is a lot of intervening hardware and software impeding the flow of data. Users at the mercy of their government's control of the network will experience varying degrees of responsiveness from various remote servers. Between governments and commercial vendors controlling or regulating Internet access and unusual events bringing large sections of the network to a crawl or a complete halt, there isn't a truly open, full-speed network connection for any Internet user in every server situation.
Compounding the troubles on the network itself, the existence of zombies in parasitic botnets that siphon away even more computing cycles on infected systems contribute to the data flow throttling. Even non-infected systems are affected due to the need to run malware protection software and the effects of DDOS attacks on popular network servers. Anti-malware running in any of the server computers or routers in the data paths that make up the network connection serving the end-user likewise impede the data flow.
For typical Internet users, the result is varying speeds of access to differently sourced servers. For any given connection speed to an ISP there will be an constriction of data that can vary due to:
where the data is sourced;
the connection management policy on the server (does it impose throttling, limit the number of connections from a single IP, etc);
the content of the data (in deep packet inspection) ;
the paranoia of the local government;
commercial interests of the ISP itself (for instance, does the data displayed come with ISP generated advertising?);
is the connection being dynamically shuffled through different paths by routers? (ie a tenuous routing or using something like TOR will cause more traffic delays);
the effects of anti-malware software hosted at the ISP necessary to prevent the user from complaining;
and finally: the effects of the anti-malware software on the user's computer.
From an engineering perspective this suggests the twin concepts of network-access “impedance” and “data device impedance”. A low network access impedance would be when the data servers are providing data at the highest speed possible up the limit of the user's connection speed. If the connection never matches the connection speed then there is the effect of the “network access impedance”. Most of the network access impedance causes are out of control of the User.
Within the User's computer, the speed at which the data can be utilized would be a measure of how fast the data downloaded is displayed on screen, played through the speakers, or saved to disk. The concept could be applied to broadband, other wireless and wired network connections. This impedance could be different for every connection or connection type established by the user but is likely to be similar for most connections.
Some would argue that “network impedance” is simply a measure of system bandwidth. The problem with labeling it as “bandwidth” does not take into account the fact that the User might not be paying for a fully unlimited speed connection or that connected servers might be deliberately throttling speed. Most ISPs offer slower connections at discounts. Users on a budget will buy slower service. In addition, the data sources or the equipment in between the User and the server might be deliberately delaying the data stream. Presumably two systems with the same data device impedance would have similar network service quality when accessing the Internet through the exact same ISP connection.
A measure of specific system impedance with a known set of applications running and anti-malware measures running might give a more useful figure of merit, the “data device impedance”, to the potential Customer/User. The ability of the device drivers, the design of the radio components in the wireless Ethernet adapter and the operating system to deliver the desired data product to the User will probably be the major components of this “data device impedance”. A good chip set with efficient drivers will have a low impedance.
The release of the Intel Atom provoked this line of reasoning because it is designed to be specifically a single threaded processor. The ATOM question is; “How does the processor fare when compared to multi-threaded processors when running Internet related applications?” A number of reviewers were comparing the ATOM to (in one case a quad core CPU !) low power Celerons, VIA CPUs etc. Even in cases where the CPU speed was the same or nearly so, the tests did not really get into the operating mode of the intended device for the ATOM, a pda or netbook processor.
In areas of electrical engineering, the most efficient transfer of power occurs when the impedance of the source (Internet connection) and the impedance of the power sink (netbook or desktop computer) match. This might be stretching the analogy too far. However it stands to reason that a User intending to buy a netbook for an networking environment he's most likely to encounter could save quite a bit of cash by buying enough CPU and OS capacity to match his probable networking environment and not more. Especially true if the networking is less than optimum. Along with CPU speed, a “data device impedance” might be a useful characteristic.
Wednesday 24 June 2009, 4:57 AM
Compact Hot Flashes
No not a new medical condition.
Well I spent the day at home yesterday working on a code project for a system release I'm about to do. It was a nice change of pace from having to ride home @ 5 or 6 or even 7PM in a hot car for at least the first 15 minutes or so. Hot meaning the freaking flesh frying 150 degrees the seats get up to in the wonderful 95 plus days we've been having for the last three weeks! Its so nice now that SUMMER is here for real, not!
I've done some experiments with CPU boards running in 120 degree heat. I have an oven at work that can be controlled plus/minus 1 degree F or about ½ degree C up to 150F. Turns out that not many CPU boards continue to run with crappy consumer grade Compact Flash drives as the hard drive media. There really is a reason to pay extra for good Industrial temp range Compact Flash drives. Obviously this is work related experimentation. (But cleverly, it applies to my home projects as well!)
I'm trying to see if there are some high-quality consumer Compact Flash drives that might be able to take Industrial temperature ranges especially on the top end. Of course there are other issues that are intruding into this test.
Determining whether or not the NAND or NOR memory cells are single level (SLC) or multi-level cells (MLC). As much as I'd like to have the expanded size of the MLC based CF drives, the scary part is the ECC algorithms used in the MLC drives might trash executable files with certain bit errors but not bother JPGs or PNGs files at all. Meaning they really are good media for media. Eh hemmm.
Most of the Compact Flashes out there play lip service to the specifications setup by SanDisk, the company that invented the form-factor and the interface specifications. With a good socket and short leads to the IDE interface, some of them seem to run fine at ATA33 speed, very few to be sure. Most seem to just barely make ATA speed or ISA bus speed for the real slow ones.
Another stinky trick I've seen is that the CF parts will work at 5 Volts to full speed but not at 3.3V. Really tacky. They are supposed to work at both voltages interchangeably.
At home the CF will be the drive for my solar energy data collection/management system.
The problem with calling up my favorite Industrial parts vendor is that the CF drives I know will work cost $90 to $150 depending on the size of the drive. I'm angling for at least 2GB so I can put a full desktop on it to start with. Eventually it will have an honest command line Linux server install.
I found some old low-power 500 MHz P3 CPUs chips that I would like to try in the board. The current CPU runs at 1GHz and runs at about 25 watts. The low power ones are supposed to be good for 12 watts and max out at 500 MHz, so we'll see.
So I was wondering if I could leave the salvaged SBC I've talked about before outside in the hot weather. Part of the process was trying to install a Linux distro on a Compact Flash and letting it cook in the lovely spring weather we've been having lately. Obviously I wasn't fast enough and the weather has gone way past “lovely”. Well I get to do the worst case test first. I was thinking about running it off a battery and a solar panel in the eventual installation. For the test, I'll run it off an AC/DC power supply and hope BOTH of them survive the heat.
Well I spent the day at home yesterday working on a code project for a system release I'm about to do. It was a nice change of pace from having to ride home @ 5 or 6 or even 7PM in a hot car for at least the first 15 minutes or so. Hot meaning the freaking flesh frying 150 degrees the seats get up to in the wonderful 95 plus days we've been having for the last three weeks! Its so nice now that SUMMER is here for real, not!
I've done some experiments with CPU boards running in 120 degree heat. I have an oven at work that can be controlled plus/minus 1 degree F or about ½ degree C up to 150F. Turns out that not many CPU boards continue to run with crappy consumer grade Compact Flash drives as the hard drive media. There really is a reason to pay extra for good Industrial temp range Compact Flash drives. Obviously this is work related experimentation. (But cleverly, it applies to my home projects as well!)
I'm trying to see if there are some high-quality consumer Compact Flash drives that might be able to take Industrial temperature ranges especially on the top end. Of course there are other issues that are intruding into this test.
Determining whether or not the NAND or NOR memory cells are single level (SLC) or multi-level cells (MLC). As much as I'd like to have the expanded size of the MLC based CF drives, the scary part is the ECC algorithms used in the MLC drives might trash executable files with certain bit errors but not bother JPGs or PNGs files at all. Meaning they really are good media for media. Eh hemmm.
Most of the Compact Flashes out there play lip service to the specifications setup by SanDisk, the company that invented the form-factor and the interface specifications. With a good socket and short leads to the IDE interface, some of them seem to run fine at ATA33 speed, very few to be sure. Most seem to just barely make ATA speed or ISA bus speed for the real slow ones.
Another stinky trick I've seen is that the CF parts will work at 5 Volts to full speed but not at 3.3V. Really tacky. They are supposed to work at both voltages interchangeably.
At home the CF will be the drive for my solar energy data collection/management system.
The problem with calling up my favorite Industrial parts vendor is that the CF drives I know will work cost $90 to $150 depending on the size of the drive. I'm angling for at least 2GB so I can put a full desktop on it to start with. Eventually it will have an honest command line Linux server install.
I found some old low-power 500 MHz P3 CPUs chips that I would like to try in the board. The current CPU runs at 1GHz and runs at about 25 watts. The low power ones are supposed to be good for 12 watts and max out at 500 MHz, so we'll see.
So I was wondering if I could leave the salvaged SBC I've talked about before outside in the hot weather. Part of the process was trying to install a Linux distro on a Compact Flash and letting it cook in the lovely spring weather we've been having lately. Obviously I wasn't fast enough and the weather has gone way past “lovely”. Well I get to do the worst case test first. I was thinking about running it off a battery and a solar panel in the eventual installation. For the test, I'll run it off an AC/DC power supply and hope BOTH of them survive the heat.
Sunday 21 June 2009, 6:39 PM
eASS (European Alphabet Soup Syndrome)
I used to think that NASA and the US Government was bad about 3 and 4 letter acronyms, well the EU has caught up and gone past. I tried reading a couple of "white papers" on this very site and they were essentially unintelligible and useless.
NASA at least publishes a glossary with definitions of every acronym used in a paper.
I went looking for enlightenment by clicking on the tags and was soon awash in a sea of even worse examples of IT double-speak and gobbledegook.
Obviously the "mainframe priesthood in white lab coats" is back.
When language is used to hide or obscure meaning, there is likely a bureaucrat mangling it to his own purpose.
What do you call a bureaucrat with a bunch of computers? An IT Admin.
NASA at least publishes a glossary with definitions of every acronym used in a paper.
I went looking for enlightenment by clicking on the tags and was soon awash in a sea of even worse examples of IT double-speak and gobbledegook.
Obviously the "mainframe priesthood in white lab coats" is back.
When language is used to hide or obscure meaning, there is likely a bureaucrat mangling it to his own purpose.
What do you call a bureaucrat with a bunch of computers? An IT Admin.
Tuesday 16 June 2009, 3:22 AM
Clonezilla, here to stay, not fade away!*
Now that I've downloaded it and started testing it, I'm really impressed with Clonezilla. Running as a “Linux live CD” iso image just over 100 megabytes large, one method touted is to operate it on a USB drive pen. Since all the test-victims in-house have CDROM/DVD Rom drives, I tested it purely as a CDROM image and it works exactly as advertised. Since I have tried booting a Linux “Live CDROM” off a USB drive before, I didn't think it would be much of an issue. But I will be testing it that way soon.
There are multiple file formats supported. FAT, FAT32, NTFS, ext2 and ext3 all were listed and that sold me right off. Drive-to-drive, partition-to-image, drive-to-image, multi-cast and peer-to-peer-casting and probably a couple other methods that I missed are all possible.
I tried booting up the CDROM image on 4 different computers with 4 different NICs and all netted up without any issues. I used a Linksys router as a DHCP server for IP addresses et al and it worked each time without a hitch. It also accepts multi-cast DHCP addresses from a Linux server.
Debbietoo served as the server/target of the Clonezilla distro (http://www.clonezilla.org/) booted up on both paqman (old P3 Compaq Enpro with Ubuntu 9.04) and the Windows XP Pro box (P4) and got viable clones out of both of them. There is a script that gets made as part of the process that can be used for the same cloning operation at another time, but I haven't tried copying it from the RAM image to debbietoo for future use.
I've been using the Clonezilla CDROM iso as the boot disk to make the clones manually. The best trick would be to host the iso in a separate partition on each of the systems you want to clone and have a way to trigger a reboot into that partition at a specified time. Cron and at on the Linux and Windows systems would work for the scheduling. The rest could be handled by writing or renaming some files to cause them to be executed on the reboot. As long as you have enough drive space for the images, this could give you a way to do a full “backup” without user intervention required.
Automatic clones of system drives has been something I've wanted to do for some time now. It just never percolated up to the top of the “to-do” stack. Its at the top now. My son's Windows XP Pro box has become practically unusable AGAIN. I've spent a lot of time copying and cleaning drive space. His computer is not going back onto the Internet before I get a clone of the clean system so I can just “drop” it back onto the C drive and reboot. Clonezilla looks a lot like the tool I need to do it.
The interface is not as “clean” as Norton Ghost but it also doesn't have the dozens of command line options of Norton that allow for scripting but make it almost impossible to use right away. Most of the options are text menu driven so it is easier to use.
In the compression department, it doesn't use a proprietary scheme so single files can be extracted using the gzip or other zip type utility.
The image itself becomes a folder with files inside it. The bulk of the image becomes a single compressed file, or a series of sections sized by user's choice in the compression choice made by the user. Some are heavily compressed but slow but one option is no compression at all, which of course is the fastest. Going from Debian 4.0 on a Celeron server 2.4 GHz (debbietoo) to a Windows XP Home edition bare metal restore on a P4 1.2 GHz took an hour and a half over an Ethernet 100TX copper connection. The original image was over 33GB.
II haven't tested the entire feature set yet of Clonezilla but already its a keeper.
* a rip from Buddy Holly
There are multiple file formats supported. FAT, FAT32, NTFS, ext2 and ext3 all were listed and that sold me right off. Drive-to-drive, partition-to-image, drive-to-image, multi-cast and peer-to-peer-casting and probably a couple other methods that I missed are all possible.
I tried booting up the CDROM image on 4 different computers with 4 different NICs and all netted up without any issues. I used a Linksys router as a DHCP server for IP addresses et al and it worked each time without a hitch. It also accepts multi-cast DHCP addresses from a Linux server.
Debbietoo served as the server/target of the Clonezilla distro (http://www.clonezilla.org/) booted up on both paqman (old P3 Compaq Enpro with Ubuntu 9.04) and the Windows XP Pro box (P4) and got viable clones out of both of them. There is a script that gets made as part of the process that can be used for the same cloning operation at another time, but I haven't tried copying it from the RAM image to debbietoo for future use.
I've been using the Clonezilla CDROM iso as the boot disk to make the clones manually. The best trick would be to host the iso in a separate partition on each of the systems you want to clone and have a way to trigger a reboot into that partition at a specified time. Cron and at on the Linux and Windows systems would work for the scheduling. The rest could be handled by writing or renaming some files to cause them to be executed on the reboot. As long as you have enough drive space for the images, this could give you a way to do a full “backup” without user intervention required.
Automatic clones of system drives has been something I've wanted to do for some time now. It just never percolated up to the top of the “to-do” stack. Its at the top now. My son's Windows XP Pro box has become practically unusable AGAIN. I've spent a lot of time copying and cleaning drive space. His computer is not going back onto the Internet before I get a clone of the clean system so I can just “drop” it back onto the C drive and reboot. Clonezilla looks a lot like the tool I need to do it.
The interface is not as “clean” as Norton Ghost but it also doesn't have the dozens of command line options of Norton that allow for scripting but make it almost impossible to use right away. Most of the options are text menu driven so it is easier to use.
In the compression department, it doesn't use a proprietary scheme so single files can be extracted using the gzip or other zip type utility.
The image itself becomes a folder with files inside it. The bulk of the image becomes a single compressed file, or a series of sections sized by user's choice in the compression choice made by the user. Some are heavily compressed but slow but one option is no compression at all, which of course is the fastest. Going from Debian 4.0 on a Celeron server 2.4 GHz (debbietoo) to a Windows XP Home edition bare metal restore on a P4 1.2 GHz took an hour and a half over an Ethernet 100TX copper connection. The original image was over 33GB.
II haven't tested the entire feature set yet of Clonezilla but already its a keeper.
* a rip from Buddy Holly
Previous
