Astaro Performance vs direct connection - Page 2

Billybob · #11 (**permalink**) 07-03-2009, 04:09 AM

Quote:

Originally Posted by khardeveld

The only thing I can't explain is the fact that download speeds with proxy are way faster then with direct connection, even with cache off and cleared. Any ideas ?

That is by design. The speed test is nothing but a jpg file(or any file that can't be compressed by your browser). When you download a file and you have http proxy av scan enabled, the proxy downloads the file and scans it while you are waiting and then dumps it to your client so you get your local network speed reading which is not what you want. That is why I had suggested turning off av scan for speakeasy

Speedtest.net collaborates with different ISPs all over the world so you can't add speedtest.net to your av skip list. If you want to test on any other site, you can simply disable the av scanning on http proxy and you will get similar results to what your line speed is.

Here is a crude diagram

Speed test site-->http proxy downloads the whole test thinking you are downloading a file-->sends it to your browser after its finished scanning--> Your browser gives you braggin rights that your download speed is way more than provided by your ISP

khardeveld · #12 (**permalink**) 07-04-2009, 04:55 PM

Of course, Why didn't I think of that

Thanks BillyBob!

Unfortunately Hardware test was aborted due to a major problem on another system

Will try again asap...

FrancWest · #13 (**permalink**) 07-04-2009, 10:27 PM

Hi,

you also might check the duplex settings on the NICs and the switch. I had a similar issue and it appeared that the switch didn't have the correct duplex setting on one of the ports. Switch was set to auto and router to full. Because of the auto setting on the switch it defaulted to half duplex operation, since it couldn't negotiate with the NIC about the duplex setting (because it was set to full instead of auto).

Franc.

khardeveld · #14 (**permalink**) 07-11-2009, 05:04 PM

Ok, we did some hardware and other tests today.

First we chaged the NIC to exclude hardware failure: No effect (still slow upload, both proxied and non-proxied)

Second: change Internet Connection. We inserted another router between Astaro and the fiber connection. This router did PPoE to the fiber and normal ethernet to devices attached to it. Then we changed our external Interface on the Astaro from PPoE to 'Ethernet Standard' and assigned static address.

We then tested without proxy and got perfect results (10MB up and 10MB down)

Then we switched to proxied http traffic. And again upload became slow (max 4.7Mb). We switched everything which could possibly interfere off (AV off, cache off, cache cleared) but nothing helped. There is no real difference between av scanning on or off of cache on or off with upload speeds (download of course is huge, but that's what we expected after the comments in this topic)

We then upgraded to the latest version (7.404), still no difference.

For now, we have returned to Astaro doing the PPoE connection again because of DNS entries we have to change if we put the other router permanently on the fiber, but somewhere this week we will switch. Can't explain why PPoE is so much slower on Astaro whith proxy turned off. MTU values etc are exactly what they should be.

If anybody has an brainwave about why this is happening I'd appreciate it. Meanwhile I'm trying to get this reported to Astaro as well. Will keep you informed.

BAlfson · #15 (**permalink**) 07-11-2009, 07:27 PM

I hope Astaro gets invovoled in this. I have gotten complaints about this and have just told people that the tests are unreliable when going through a device like the Astaro. I hope I haven't been lying, but I'm not comfortable because I don't understand why things appear to be so much slower.

If they really are slower, then how can it be that the Astaro never registers more than 30% of CPU?

Cheers - Bob

William · #16 (**permalink**) 07-12-2009, 02:06 AM

you say p-4 but what speed p-4? Also how many users and what speed hdd? Any of these can have HUGE impacts on overall performance.

khardeveld · #17 (**permalink**) 07-12-2009, 09:49 AM

Number of users: 0 (Only Astaro itself)
HDD: 7200 rpm SATA

We specifically did testing without any other user or machine attached to the Astaro because we wanted to benchmark as cleanly as possible.

If users or load would be an issue, we'd expect CPU to go up, which it didn't. Also memory load is marginal, so no swap activities.

Under these circumstances we'd expect some perfomance degradation using proxy, but not this much

William · #18 (**permalink**) 07-12-2009, 12:49 PM

again..what SPEED cpu?

William · #19 (**permalink**) 07-12-2009, 12:53 PM

looking at your motherboard it's very possible it's your motherboard. SIS chipsets tend to be flaky.

William · #20 (**permalink**) 07-12-2009, 12:55 PM

Quote:

Originally Posted by BAlfson

I hope Astaro gets invovoled in this. I have gotten complaints about this and have just told people that the tests are unreliable when going through a device like the Astaro. I hope I haven't been lying, but I'm not comfortable because I don't understand why things appear to be so much slower.

If they really are slower, then how can it be that the Astaro never registers more than 30% of CPU?

Cheers - Bob

These tests are also highly affected by distance between yourself and the test site as well as load on the test server as well. I am not seeing this behavior on any astaro(or any other firewall software i have used). When i have seen this issue it's been either a network issue or something with the server hardware itself.