SABnzbd Forums

Ever since I upgraded my (Synology) Sabnzbd to 1.1.0 [8abcf08] (upgraded from 1.0.3), download-performance seems to have been cut in half. I used to download at ca. 4.8 Mbps, and now it's never higher than ca. 2.6 Mbps.

I have 2 servers at priority 0, both with 8 connections defined. Lowering those, are adding more, didn't make a real difference.

In 1.1.0 we chose tuning parameters that turn out to be non-optimal for some low-CPU systems.
In 1.2.0, we'll change that and also look at optimizations that benefit such systems.

shypike wrote:In 1.1.0 we chose tuning parameters that turn out to be non-optimal for some low-CPU systems.
In 1.2.0, we'll change that and also look at optimizations that benefit such systems.

Thanks.

My system is hardly low-CPU, though. It's Synology DSM 6 inside a VM, with 4x i7 980X virtual CPU's, and 4G RAM. CPU usage, as a result, is therefore not an issue here (repair, for instance, takes around 17% CPU, at the most). It's just the download speed which has dropped significantly (downloading itself takes only ca. 6% CPU or less).

I even briefly reniced the process (to -n1), but that made no visible difference.

EDIT: P.S. Is there a manual way to undo the new 'tuning parameters'?

Only if you can get to the python code, can you?

On my system, commenting out the following code in sabnzbd/downloader.py around line 487 (first introduced in this commit) causes download speed to go up significantly:

safihre wrote:Only if you can get to the python code, can you?

Sure, I can get to it just fine. Please, tell me what to do!

jcfp wrote:On my system, commenting out the following code in sabnzbd/downloader.py around line 487 (first introduced in this commit) causes download speed to go up significantly:

Code: Select all

# Why check so often when so few things happend?
if len(readkeys) >= 8 and len(read) < len(readkeys)/4:
    time.sleep(0.05)

Thanks, will do!

Right, so clearly this optimization works on Windows and Mac, but not on Linux.
Guess will have to make that conditional.

asarian wrote:

jcfp wrote:On my system, commenting out the following code in sabnzbd/downloader.py around line 487 (first introduced in this commit) causes download speed to go up significantly:

Code: Select all

# Why check so often when so few things happend?
if len(readkeys) >= 8 and len(read) < len(readkeys)/4:
    time.sleep(0.05)

Thanks, will do!

Yep, this worked immediately. Thanks again.

jcfp wrote:On my system, commenting out the following code in sabnzbd/downloader.py around line 487 (first introduced in this commit) causes download speed to go up significantly:

Code: Select all

# Why check so often when so few things happend?
if len(readkeys) >= 8 and len(read) < len(readkeys)/4:
    time.sleep(0.05)

Thank you so much, took my speed from 5.3MB/s back up to what it was before the update 7.9MB/s