i spoke too fast. just went back to check the seeing in DO and, buffer size was 500 and 1MB.
but i played with dearopus.net/ during one day (but had no energy to continue for now), so i restaured a backup of the configuration of DO who were around january 2013, maybe that's why no more copy freezing. i'll investigate a bit hardner now. so for now, buffer size is 500 and 1Mb.
Buffer size 500 has problem at my side, why not u try smaller setting?
yes i'll try and report it here
yes i'll try and report it here[/quote]
buffer size at 256kb crash the copy and the whole network
"network not available error 64", just copied a Mp4 of 254Mb, and crashed as soon as the copy started at 9Mb. it's now 3 minutes and the copy dialog does nothing is still "freezed" but the counter is continuing to disply.
(...)
ok the copy dialog just closed, and i'm trying to reach the administrator panel of my nas so i can make a log and send it to synology. the access came back only almost 1 minute after the copy dialog box closed on the DO side 
.
Your case seems complicated.
How about Windows explorer copying to NAS ? or DOS? If that also crash, then your system is having problem, could be PC side or NAS side.
It is easier if you have another PC / laptop to try, then can confirm if the problem is at PC or NAS. I have 4 laptops for testing. For all laptops, using DOPus to copy file to NAS has 50% performance impact.
I'd just like to add my 2c in here. I am also experiencing this issue, but with transfers from another Win7 PC. Explorer copies at ~115MB/sec all day, but the best I can get is ~100MB/sec with Opus and that's only for ~30sec, then it drops back to 60MB/sec or so.
I have tried various cache settings, but this is as good as I can get.
I, like the others, hate the idea of going back to explorer, it represents all that is wrong and distasteful in this world!! 
Surely this is something that can be fixed? If Explorer can do it, Opus can do it better!!
I did a lot of testing of network copy speed in Opus over gigabit ethernet recently, and I could not see any overall speed difference between Opus and Explorer.
There were differences in burst speed (which I think are just differences in how the different buffer sizes affect reported speed of the parts being measured rather than actual speed of the entire chain of operations), and also differences with both programs between how long each test took (some tests were much quicker than others, with the same program, but the high/low/average times were equal for both programs), but overall I could not find a significant difference.
If you set Opus to use a 1MB buffer size, that should give you the same characteristics as the Windows 7 version of Explorer when doing network transfers, although in my testing it made little to no difference. (That may not be the case with all disks, networks, antivirus, etc. of course. Some may treat DOpus.exe differently to Explorer.exe, for example. Some NAS are literally only tested/benchmarked against Explorer and some slight differences in the way Opus and other programs open files may trigger pathological performance issues. It shouldn't matter, though, because both Opus and Explorer are calling very high-level APIs that just take a buffer of data and write it to a file, and the rest is up to the OS and drivers. Assuming default non-buffered I/O settings, newer versions of Opus use separate threads to read and write the data when copying large files from a local drive to a network drive, so almost all of the time Opus spends is in waiting for Windows to read or write the data before it immediately asks Windows to read/write some more data.)
Also remember the factors in this post which can affect copy speed. e.g. Copying lots of small files, Opus will be slower than Explorer by default because Opus copies additional attributes that Explorer doesn't bother with.
And remember it's difficult to measure copy speed due to buffering and differences in how the speed is reported. You need to repeat the tests several times in a row (for each program), with the same files, to account for buffering or interference from other programs that may speed up or slow down each test. I did my tests by looking at the Network tab in Task Manager, and measuring the width of the graph for each file copy. (The high and low points on the graphs varied greatly but the widths of each operation -- i.e. the actual time taken to send the data -- were consistent between both programs.)
Hi,
I have the same problem with a NAS Synology, speed of transfer really really differs (50 mo/s for explorer, 5 mo/s for Dopus) between Explorer and DOPUS (win 10). It concerns both files, big and little. Really strange.
I'm using the latest release. I try some of the things on this post to solve this.
As well as what's in this post, there are a lot of posts on the Synology forum about speed issues (and speed differences with Explorer vs almost literally anything else) with some setups. It seems Synology don't always optimize things for the general case, only for Explorer's unusual way of transferring data, and different device models / firmwares bring back (similar) regressions that had already been fixed in others years ago, for some reason.
Try testing with another Windows machine as the server. If that is the same speed, it points to the NAS not handling some types of file transfer well, which will potentially affect almost everything that reads or writes from the drive that isn't Explorer. If it's also slow with Windows at both ends, it may point to the problem being elsewhere and not the NAS.
(Almost nothing else does the same thing Explorer does, unless they call the basic and fairly limited shell file copy APIs to copy files from A to B. Unfortunately, Microsoft changed the way Explorer works rather than adding similar changes to the general file read/write APIs.)
The buffer size and other settings can help with some devices, though. (Buffer-size performance problems can also be due to network hardware between the PC and NAS, and not always the NAS's fault. Similar with USB devices, where the controllers on some USB devices have pathological performance cases if the buffer sizes aren't what they assume them to be.
(And be careful how speed is measured. The best way is to look at the Network Performance graph in Task Manager and note how long data is being sent for. Don't rely on progress dialogs as, for example, Explorer closes its dialog and claims the data is sent long before it's actually completed. Buffering can also impact results a lot, if the same data is copied multiple times for tests. Virus scanners scanning data as it is read or written can also have a huge impact sometimes, and may treat different programs differently.)
I am reviving this topic as I am not satisfied with the solutions.
I am copying from my Windows computer (Windows 11 22H2) to a Mac Mini server on a flat network.
Using Directory Opus 12.15 x64 build 7102.
No NAS, just from one computer to another (through a switch). Don't ask me for the target firmware, operating system etc. That has nothing to do with it.
I am copying large files (> 1GB each).
Using Windows Explorer, I get 110MB/s copy speed.
Using Directory Opus, I only get 55MB/s - half of what Explorer does.
This is not only measured using the app dialog progress bars, but also confirmed with the Task Manager ethernet performance graph.
I tried difference copy_buffer_size values and different copy_nonbufferio_threshold values, but I cannot get Directory Opus to get more than 55MB/s.
I also want to comment on the old Jumbo Frames messages - for two computers to communicate using Jumbo Frames (eg any MTU other than 1500) both computers must have Jumbo Frames enabled (and set to the same MTU) and the switch/router/device in the middle must also support MTU of that set size. If any device along the network path does not support Jumbo Frames, or does not have it set to the same number, all devices will fall back to an MTU of 1500.
Directory Opus 13 will copy using the same mechanism Explorer uses, and should give the same speed you see in that (for better and worse; it depends on the hardware/network/etc. which is faster, but if you want the same speed as Explorer, that's the default now).
Okay, that's great to hear. I will download, install and test v13 this week.
Thanks Leo!
Installed v13 and yes, the issue is solved.
Directory Opus now gets the same speed as Explorer (~120MB/s) copying from my pc to my mac.
Thank you!
1 Like