The frequency and volume of them is somewhat annoying though. They seem to blithely assume that everyone is on a superfast internet connection.
IBM's OS/2 came quite close to the ideal early on. It could terminate errant processes with extreme prejudice without any side effects. There was a 16550 emulator for buffered RS232 emulator written for it that ran in real time.
It was still potentially vulnerable but you really had to work hard. Windoze 3 was pretty much like a Swiss cheese by comparison.
All OS's and applications need to be updated as vulnerabilities are found. The big problem with Mickeysoft was that they put far too many things into the highest privileged ring to make games go faster.
The ship it and be damned approach to product development...
Unix systems are typically a fair bit more robust than Windoze.
However for a more intuitive display a with the ability to display all the individual parameters I would suggest installing a tool like crystal disk info:
formatting link
(that should take you straight to the download page)
Worth checking then with those particular drives, since any background access windows makes to the drive will cause it to spin up again.
Thanks. The first one says OK for my drive, the second says FALSE under PredictFailure
I spent much of yesterday reinstalling apps on the C: drive and removing stuff I don't use from D:. Adobe Elements V9 took ages to remove. Soooo many files, and I rarely use it anyway. The big one is yet to come, Lightroom V3.6 which I am nervous of because I cannot download another copy. I still have the 3.6 installer on the E: drive, but I can't remember if it needed to download more stuff from Adobe. The worst case scenario would be to reinstall from the V3 upgrade DVD.
Since C+D are tightly linked via the registry, it seems pointless to have a separate partition for applications. Currently D: is 41Gbytes and only 4Gb was used, and many of the Apps are ancient and date back to Win98.
Shrink the partition down, and clone it across and put it on the same SSD ?
There's got to be a better answer.
And any time you start a destructive set of edits, make a backup first. Then if you discover you don't like how things are going, just start over using a restoration for the purpose.
On occasion, I've been damn lucky on backups. One day, on a whim, I did a backup of a Windows 7. And ten minutes later, I managed to destroy that C: - totally ruined, CHKDSK would not touch it. And at the time, there was no good, logical reason to make that backup. I was pretty happy after the restore completed.
Well I am trying to create a system image backup of C+D using Windows 7 backup in Control Panel (the way I usually do it).
The first issue was windows got stuck saying "looking for backup devices". Normally I do data backups to my WD MyCloud device which is network connected but I only power it up to do backups. Once powered on and connected, lo and behold Win10 then noticed that I had plugged my LaCie external USB2 mains-powered disk (320 Gb), but in the past I have noticed that it takes ages to create a system image. This time is no exception. Task manager shows that it is only managing about 1MB/sec to write to the external USB2 drive !!??.
Resource monitor shows a process called wbengine which is managing to read a bit more. How come so slow ?. Resource monitor also shows a response time of about
1,300 millisec for the process writing to the I: drive, the external USB2 device. This is worse than even USB1 could manage, so where is the hold up ?. I am using a grey USBA->USBB cable that probably came with my Epson scanner bought in 1998. Could this be the problem ?.
I have Sandisk Ultra USB3 64GB thumb drive. Maybe I'll try and create another system image onto this, plugged directly into one of the USB2 backplane M/B ports next time.
If you have a bunch of USB cables handy, you might notice some have transparent outer insulation, and you can "see" the shield foil underneath. These are shielded cables, likely intended for USB2 work at
480 Mbit/sec rates. I consider this an excellent habit, as it allows customers to verify the cable is shielded.
Shielded cables are "fat" compared to keyboard cables, so just the OD of the cable is a hint.
They also used to make "keyboard cables". These have four conductors but no shield, and work at 1.5Mbit/sec rates. Whereas USB1.1 non-keyboard, runs at 12Mbit/sec and typically transfers backups at 1 megabyte per second (about 2/3rds of the cable bitrate).
You should not use the keyboard cables any more. They can be retired. They really have no suitable high frequency characteristic to speak of. Whereas the USB2 shielded cables are "nicer". And the five pin diff interface portion of the nine pin USB3, those are "excellent" in terms of high frequency characteristic.
You can try benching the WD MyCloud with this if you want.
formatting link
That is the Free Version. The free version does read-only benchmarks. The paid version supports both read and write and you must be a lot more careful when using it. The free version has an upper limit on the size of the disk, and the author of the program never expected 18TB drives to exist :-)
Still, it's an excellent utility for quick tests.
HDTune is not entirely "insulated" from OS effects, and Windows 10 has a bad habit of noticing you're trying to get work done, and it will start doing really weird maintenance in the middle of your benchmark run. Not every "spike" in the output plot, is coming from the disk drive, and some of the bench runs are being ruined by the OS. The very "smoothest" runs come from Windows 2000.
The benchmarks have different shapes. On a hard drive, the outer diameter of the platter is fastest, and that's on the left of the diagram. As long as the cable or bus connection is not the limiting factor, you can see the curved shape of the hard drive speed characteristic.
formatting link
SSDs don't have a platter, and the transfer rate is constant over the surface. Their plot is a flat line, with the exception being uninitialized TLC or QLC drives, where the line is flat, but "raised" in some parts.
If the cable is slowing things up, the plot will be a flat line at a really slow rate. A 200MB/sec hard drive on a 30MB/sec USB2 cable, gives a 30MB/sec flat line.
An SSD on the end of a slow cabling scheme, you could not tell from the shape (both good and bad plots are flat), but the magnitude of the speed (200MB/sec versus 30MB/sec) tells you how things are going.
If you're using a keyboard cable, the benchmark will be flat with a lot of spikes in it. A "noisy" flat line at a guess. I don't even have any keyboard cables of the old kind here, to test with, so I can't make any sample plots.
I only have one USB keyboard and the cable is connected inside it, with only the standard USB 'A' connector at the computer end. The cable I am using has a USB'A' connector at one end and the 'square' ?USB'B' connector at the other end. It has a cable which, using a domestic tape measure seems to have a diameter of about 5mm. Seems to be almost identical to the fixed keyboard USB cable. I.e. both are very flexible.
It is 8PM now. I'll leave it running until midnight and if it is still running I'll abort it.
There seem to be a lot of comments about slowww image backups using Win7 backup on Win10. If Microsoft no longer support it, what do they expect users to do ?. Is this part of their Win11 strategy to get everyone locked into the Microsoft store and cloud backups ?.
I think I will buy a copy of Macrium reflect 8 (home). It looks like a good product.
You can download the free version and that will do image backups of a partition to a file....
Out of curiosity I just did a test to see if it imposed any speed limitations. I told it to image a drive with just under 290GB on it to a file on another drive:
formatting link
It seems to be romping through at about 250 MB/sec which is about half the sequential write speed of the target drive (it's an SSD but limited by the SATA 3 interface). The source is a NVMe drive that can read at about 2GB/sec
(this is win 10 pro running on Hyper V in bare metal hypervisor mode - so you lose a bit of raw IO speed)
Interestingly, looking at task manager it shows the source drive pegged at 100% load, and the destination at only 25% - so probably doing lots of smaller random reads on the source.
I left my backup running and went to bed at 00:30 and it was finished before 06:30. Still can't understand where the bottleneck was. Average of 1 MByte/sec writing to a USB2 disk is crap. It 'only' used to take about 45 mins with Win7.
Windows has some strange ideas about "cache choking" on the System Write cache, that gives slower results than WinXP. I quite typically find hard disk drives that could do 150MB/sec or 200MB/sec, only doing 100MB/sec thanks to Windows.
There are different kinds of Windows backups, and for a backup of critical partitions ("enough so Windows boots, minimum"), the container is .vhd or .vhdx. These are placed in folders, where the permissions do not allow easy access.
Only the clusters with file content are captured.
If cluster 1,2,3,5,7,20,21,22,23 are occupied, this is four read operations, and the head smoothly and sequentially goes over the source partition. There should not be a lot of clunking noises. Restoral should be similarly sequential and quiet. (When you hear clunking sometimes, it's NTFS TXF doing that, perhaps twice a second.)
Windows 7 uses .vhd and has a 2TB datafill per .vhd limitation. Windows 10 uses .vhdx containers and those can have much more than the 2TB datafill limit. But .vhdx is also a more obscure format, and in my computer room, is a "dead end".
The permissions on the backup folder are meant to annoy, but with the right approach you can get in there for a look.
VHD files can be "attached" in Windows, in diskmgmt.msc . If you need random access to a backup file, you can have it. While Microsoft tries to make it so that access to VHDX requires turning on HyperV in Windows 10, there are likely to be attachment methods you can use anyway. Even 7ZIP can access a VHD at least, but I don't know if it does VHDX. You can kinda tell, from my lack of experience with VHDX, that I am studiously ignoring them and giving them the cold shoulder :-)
I used to use the Windows backup at first, but it just isn't tidy enough for my tastes and is a very basic (disaster protection) method.
With Macrium, I can dump the entire machine (four disk drives) as a single 2.6TB .mrimg file. But for reliability reasons, I don't recommend doing it that way, and a partition-at-a-time is another way to give surgical access to content. Some partitions have incompressible content, and wasting compression time on them is pointless. By managing the partitions one at a time, they can be tailored according to characteristic - if they're compressible, maybe compressing them later makes sense. Compressing the
2.6TB file, would cost a pound of electricity, so at some point, what you do does matter :-) If you had an Epyc running 7ZIP, it might burn 280W all day long, compressing your archive. The room gets warm :-) Great in winter. Other times, not so much.
VHD
+-----------------------------+------------------------ - - - | allocated in 2MB chunks | allocated in 2MB chunks +-----------------------------+------------------------ - - - 0 1TB
^ ^ | | +--- a single cluster +---- portions of partition which don't captured, wastes contain content, need not be a 2MB chunk captured. The diagram here, implies 999GB of white space, took no storage space in the .vhd file.
It might pay some savings, to do a "consolidate" defrag before backup. Now, most people do not care about details, so a lot of this fine detail can be ignored. The VHD container has sufficient basic efficiency, to make it a worthwhile choice for storage. Fiddling with the source partition, is like jamming down the lid on a steamer trunk - a battle of will power against irresistible forces. For some people, the urge to fiddle is just as powerful as that steamer trunk.
I can do a backup with "dd" disk dump, from any OS, but like the steamer trunk analogy, the waste of human labour hours on such a project, makes doing this into "bar bet" material. You can start by zeroing the white space with "sdelete.exe", then later, an archive compression utility can compress away the zeroed sections. With Windows Backup or Macrium Backup, the failure to capture white space, means white space needs no special treatment, and you can leave your copy of "sdelete.exe" at home :-) The VHD and the .mrimg only allocate file space, when the content being copied is "real, and important".
System Image Windows 10 creates .vhdx files, one per partition System Image Windows 8 creates .vhdx files, one per partition WinXP "disk2vhd 1.64" creates .vhd files, one per disk drive
Win10 A steady 73MB/sec, no matter what. No resource is the bottleneck, like activity was throttled by OS Yet output drive (SSD on USB3) reads "92% activity" at 73MB/sec
Win8 130MB/sec backup of HDD 227MB/sec backup of SSD Output disk reads "running at 60% of speed"
WinXP 163MB/sec backup of HDD 200MB/sec backup of SSD
Win10 seems to have had some "loser code" added to it.
I was testing in W10-1909 (SSD is not kept up-to-date, a waste of time). Most of the other Win10 are up-to-date, but I have to leave some for comparison testing.
I did not get any 1MB/sec behavior on the USB3-connected output SSD. But neither was the rate all that impressive.
Strangely, the WinXP test was less smooth than I was expecting, and I don't understand why, unless it was the "Paged Pool" issue my copy of Windows XP has got. It's not normal for WinXP to do that many writes, without getting "wobbly". Not everyone sees that effect. I blame it on VPC2007 and VirtualBox being installed.
And Macrium, running on any of those, should manage at least 100MB/sec. Macrium is limited to some extent, by its need to do a checksum when preparing the output file, and the checksum limits how fast it can go on cheesy processors. You'd need a 11900K to see a tripling of the speed.
Thanks. I think I will download the free version of Macrium just out of interest and do another backup to the same disk, then repeat it to a Sandisk Ultra 64G USB3 thumb drive plugged directly into one of the USB2 backplane connectors. I don't have any USB3 ports.
For desktops, you can get a USB3 card to add USB3 ports to the back of the machine. This takes external disk speed from USB2 30MB/sec up to the limits imposed by the hard drive itself. (First generation USB3 disk enclosures might have managed 200MB/sec, and it's possible later ones do better than that.)
But the Windows 7 Backup software offered by Windows 10, was not using the speed of the port. It's possible an earlier version of Windows 10, that software ran faster.
When I do backups here, I do them to SATA-connected drives. Which takes things like USB2 out of the picture.
If your output drive is limited to 30MB/sec, then Macrium ReflectFree cannot go faster than that to the USB2 connected drive. You'd hope it could do better than 1MB/sec though.
HomeOwnersHub website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.