General Discussion Need help? Have a problem? Let us help you. Bug reports and feature requests should be made using the Bug Tracker or Feature Tracker |
01-20-2011, 11:46 AM
|
#1
|
Member
FlashFXP Beta Tester
Join Date: Oct 2006
Posts: 38
|
SSL file transfer CPU usage?
Hi.
I am running FlashFXP version 4.0.0 build 1526, and today I noticed something weird.
I downloaded some files from a FTP server using "secure file transfers".
I noticed that my CPU fan started to run faster than normal, and when I checked the CPU usage, FlashFXP used between 40 and 50 % CPU. And that is on my 3 GHz Core 2 Duo.
I then tried to download the exact same file using FalshFXP 3.7.9 build 1348, and here the CPU usage stays below 2 %.
With the 3.7.9 version I can't really tell the difference between "secure file transfers" and normal transfers. It is around 1-2 % CPU usage.
But with 4.0.0 build 1526 it is a total different story.
I am running Windows XP Pro with SP3 in case that matters.
I would really like a solution, as I use "secure file transfers" a lot.
I have no idea how long the issue has been there. I just noticed it today, but it may have been there for a while, and I just didn't notice the CPU load.
Best regards
rco133
|
|
|
01-21-2011, 04:42 PM
|
#2
|
FlashFXP Developer
FlashFXP Administrator ioFTPD Beta Tester
Join Date: Oct 2001
Posts: 8,012
|
Hello,
At what speed are the files being transfered? In 3.7.9 high speed transfers were being throttled so the code was reworked to provide faster throughput.
Also what is your receive TCP/IP Buffer size in Preferences dialog under Connection. It should be set to default, if you upgraded from 3.7.9 it might be 32KB.
|
|
|
01-22-2011, 03:39 AM
|
#3
|
Member
FlashFXP Beta Tester
Join Date: Oct 2006
Posts: 38
|
Hi Bigstar.
I have a 10 Mbit line, and both versions of FlashFXP maxes out the line when I download.
In 3.7.9 build 1348 the TCP/IP buffersizes are Send 64 KB and Receive is 128 KB.
In 4.0.0 build 1525, which I am using now and also has the same issue, the TCP/IP buffersizes are Send 64 KB and Receive is 128 KB.
I have tried default and auto for both Send and Receive, but the CPU load is still between 40 and 50 %.
I have also tried to ise the SSL files from the old version, but that doesn't change anything either.
rco133
|
|
|
01-22-2011, 12:46 PM
|
#4
|
FlashFXP Developer
FlashFXP Administrator ioFTPD Beta Tester
Join Date: Oct 2001
Posts: 8,012
|
Thanks for the information.
So roughly 1.25MB/s transfer speed?
I will investigate this issue further and get back to you.
|
|
|
01-22-2011, 01:50 PM
|
#5
|
Member
FlashFXP Beta Tester
Join Date: Oct 2006
Posts: 38
|
Quote:
Originally Posted by bigstar
Thanks for the information.
So roughly 1.25MB/s transfer speed?
I will investigate this issue further and get back to you.
|
Yes. It is just above 1 MB/s.
rco133
|
|
|
04-02-2011, 07:48 PM
|
#6
|
Junior Member
FlashFXP Beta Tester
Join Date: Nov 2006
Posts: 26
|
I just wanted to bump this thread. I experience high CPU usage when downloading with SSL (AUTH TLS) too. I have a 3GHz Quad core CPU. Eyeballing the task manager with 3 instances of FlashFXP downloading at 100-300kb/s, each instance takes up about 15% (half core) and may spike to 25% (full core) CPU usage. Additionally, the "System" process is over 10% usage. If I close out of all instances of FlashFXP, it drops to almost nothing. I am using the latest build v4 build 1548.
Any updates on this? Why does v3 use so much less CPU power? I like using a lot of instances but this kills system performance with v4.
|
|
|
04-02-2011, 08:10 PM
|
#7
|
Junior Member
FlashFXP Beta Tester
Join Date: Nov 2006
Posts: 26
|
I see there are numerous threads like this one. I took the suggestion of setting the receive window size to the default of 32kb. After that, using the same 3 queues, instances are running on average at 0% CPU usage. System process is also at 0%.
It is unfortunate that the higher receive window uses so much CPU in v4. I was able to drastically increase download speeds with the high receive window from certain locations. I was using a 512kb window originally.
Any news on this would be nice!
|
|
|
04-03-2011, 08:56 AM
|
#8
|
FlashFXP Developer
FlashFXP Administrator ioFTPD Beta Tester
Join Date: Oct 2001
Posts: 8,012
|
The TCP/IP buffer size adjustment logic was flawed in v3.x and it didn't work as designed, there were so many issues that prevented this feature from working as intended. And because of this its impossible to compare this feature in v3.x to v4.x.
If your TCP/IP buffer is too big/small it will result in high CPU usage. Which is why with v4.0 we made the default value "(default)" which is the OS default, rather than a fixed size of 32kb.
While FlashFXP offers a wide range of TCP/IP buffer sizes most are unrealistic, On my gigabit lan 128KB is enough to reach over 100MB/s, however all networks are different.
At 1MB/s default would probably be the most practical value but there are other factors as well, You mentioned you're using Windows XP its difficult to compare the TCP/IP stack in XP to Windows 7 as they are quite different, it also depends on how your TCP/IP stack is configured, such as using one of the many tweak tools.
Every setup is unique because there are so many different factors involved.
Hardware, Operating system, Firewall, anti-virus software, etc
|
|
|
04-03-2011, 09:21 AM
|
#9
|
FlashFXP Developer
FlashFXP Administrator ioFTPD Beta Tester
Join Date: Oct 2001
Posts: 8,012
|
In addition one of the biggest factors that can effect performance is the network card.
Some network cards let you configure the "Max IRQ per Second" and from my experience this value is usually set to a value that provides average performance with a typical range of 1000 to 30000.
The default might be 5,000 or 10,000, however increasing this value can improve performance as much as 10-30% depending on the quality of the network card/motherboard and how well the operating system can handle the IRQ requests.
|
|
|
04-03-2011, 09:35 AM
|
#10
|
Junior Member
FlashFXP Beta Tester
Join Date: Nov 2006
Posts: 26
|
Thanks, good info. For the record, I'm running Win7 Ultimate 64bit.
|
|
|
02-24-2012, 04:28 PM
|
#11
|
Senior Member
FlashFXP Beta Tester
Join Date: Jul 2005
Posts: 106
|
interesting thread, while im experiencing similar issue in latest beta 1719 ... let me check network card and for that setting and update drivers aswell... in my case the AV scanner mcafee bugs with 50% if running fxp transfer ssl'ed.
|
|
|
02-24-2012, 04:41 PM
|
#12
|
FlashFXP Developer
FlashFXP Administrator ioFTPD Beta Tester
Join Date: Oct 2001
Posts: 8,012
|
This thread is specific to direct ssl downloads and uploads..
If you're performing a direct ssl fxp transfer then none of this would apply to you.
|
|
|
02-24-2012, 04:43 PM
|
#13
|
Senior Member
FlashFXP Beta Tester
Join Date: Jul 2005
Posts: 106
|
check my recent bug , it issues on ssl/non-ssl fxp and downloads , i go for NIC driver update, let's see. not sure what mcafee mcshield.exe bugging to 50% if ffxp transfer a file. yoiu have experience on AV products and analysing your ffxp ?
|
|
|
02-24-2012, 04:46 PM
|
#14
|
Senior Member
FlashFXP Beta Tester
Join Date: Jul 2005
Posts: 106
|
Quote:
Originally Posted by bigstar
In addition one of the biggest factors that can effect performance is the network card.
Some network cards let you configure the "Max IRQ per Second" and from my experience this value is usually set to a value that provides average performance with a typical range of 1000 to 30000.
The default might be 5,000 or 10,000, however increasing this value can improve performance as much as 10-30% depending on the quality of the network card/motherboard and how well the operating system can handle the IRQ requests.
|
my marvell yukon's driver were from 2010 , having now january 2012 drivers, i have that default value on 5000 , perhaps i play around witrh the values lets see if this helps ,
let's focus on my other thread, regarding mcshield.exe , damn AV tool ;-)
|
|
|
02-27-2012, 12:29 PM
|
#15
|
FlashFXP Developer
FlashFXP Administrator ioFTPD Beta Tester
Join Date: Oct 2001
Posts: 8,012
|
I discovered an issue that can result in higher than normal CPU usage when connected to two servers, specifically when connecting to both servers at the same time.
During the connection and login process CPU usage was max for the core/cpu, once logged in the CPU usage returned to normal.
I have not seen this issue during any site to site transfers, however that doesn't mean it didn't occur.
I am preparing a beta update with some improvements to prevent this type of situation.
|
|
|
Thread Tools |
|
Display Modes |
Rate This Thread |
Linear Mode
|
|
Posting Rules
|
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
HTML code is Off
|
|
|
All times are GMT -5. The time now is 11:47 PM.
|