Thanks for the advice, raztoki.
Quote:
Originally Posted by raztoki
Outside of that, I do see that you micromanage downloads as in force start over and above maxsimdl setting to ensure that downloads are underway even though other tasks (CRC/extraction) are still taking place within the slot. To me you have created the high IO loads with doing this, as the more tasks you start the slower over all performance will be (classic example is starting 2+ move function on same traditional harddrive, your head on harddive is moving disk in order to read and seek, if you queue task performance stays the highest, and next tasks starts as that finishes).
|
I don't know if you've read all the earlier posts in this thread, but we already came to this conclusion, just without the hard drive example you used.
Quote:
Originally Posted by madcowpig
If you could've seen the whole picture, you would've also seen that I had nine more forced downloads running, with ten connections each, as well as a four file, 3.4 GB extraction process running, while watching a 1080p video on YouTube, at the same time as the checksum. Of course, a checksum by itself, is very fast.
|
Quote:
Originally Posted by raztoki
As you're already micromanaging maybe it would be best if you just disable the auto checksumming and extraction tasks? You can do that within JDownloader context menu when IO loads are reduced and|or within external programs if you prefer. Disabling these tasks will give you the fastest download speed!
|
Jiaz already suggested this, but I'd rather keep it automatic. The thing is, this wasn't slowing down my desired speed, which was 12 MB/s. See if I replace the nine downloads with ten connections each, in my example above, with one file from MEGA, I could download with 12 MB/s, since MEGA has unlimited download speed, I only need one slot available to reach max speed. The issue is with openload.co, they limit speed on the connection level, depending on how busy their server is, so on certain files, even twenty connections wouldn't get me to 12 MB/s. I have to use multiple downloads to get to 12MB/s and because the queued checksums were holding up the available slots, I was not able to reach that speed, using simultaneous downloads. This is why the new advanced options feature, Jiaz suggested, would help me in this particular case. You're right though, this would help if I was trying to max out my speed, but I'm not. All I'm running is one checksum, one extraction and simultaneous downloads capped at 12 MB/s, which my hard drive can handle, but JD does not let me do this unless I manually select forced downloads. I don't care if checksums and extraction are slow during this process, since I can still achieve, 12 MB/s.
Quote:
Originally Posted by raztoki
My further advice would on a fast internet connection to download to multiple harddrives, this load balances system (womanboard/harddrive/cpu/etc) bottlenecks, and you can achieve significant performance gains at little cost (you don't have to buy new hardware, just slight change in behaviour).
|
This, I have no idea about. If you could, please elaborate.
You say,
Quote:
Originally Posted by raztoki
download to multiple harddrives
|
and then you say,
Quote:
Originally Posted by raztoki
(you don't have to buy new hardware, just slight change in behaviour)
|
so do I or don't I, have to buy another hard drive, to get this to work?
Jiaz's Event Scripter idea should work in theory, but I don't know how to write scripts. I want to set it so that all packages in download list have to finish, before hash checks start. Then all hash checks have to finish, before it starts to extract all files in download list, one package at a time.
Thanks again, for this speed up suggestion.
P.S. There is a tiny problem on this board. When you spell the words, m.o.t.h.e.r or m.o.m (Without the periods.), it shows the word, woman. This happened to you raztoki, when you tried to spell the word, m.o.t.h.e.rboard (Without the periods.), it showed the word, womanboard. The same thing happened to me, while private messaging.