Forum Discussion

itakepix_39096's avatar
itakepix_39096
Icon for Nimbostratus rankNimbostratus
Apr 17, 2009

Differences on Client Devices

Hi Guys,

 

I have configured a single F5 Big IP 1600 to accept HTTP requests and forward them directly to an Internet Gateway bypassing the application that the traffic will eventually be passed through for optimization.

 

The clients consist of 3G Mobile Handsets (iPhone, HTC mobiles etc) and Laptops with 3G connectivity.

 

Both devices therefore make HTTP requests and currently we are pitching the devices against each other in the following manner on two different APNs.

 

Live APN Laptop HTTP D/L speed 340kb/s

 

F5 APN Laptop HTTP D/L speed 340kb/s

 

When we test browsing with these two devices the difference is non existent.

 

Live APN iPhone Browsing Pages 10sec

 

F5 APN iPhone Browsing Pages 30sec

 

When comparing page loads on the two iPhones of exactly the same page, with a request made at the same time, the iPhone that traverses the LB system is always significantly slower.... yet as stated, the laptops see no difference.

 

The traffic path for all of these devices is identical in so much that they enter through the same GGSN and are routed on to the same Internet Gateway as follows:

 

Live APN

 

GGSN1 > Internet GW1

 

LB APN

 

GGSN1 > LB1 > Internet GW1

 

I am using the standard TCP profile with the MSS modified to 1380 and all the VLANs that the traffic traverses on the LB are configured with an MTU of 1420.

 

On the HTTP Virtual Server the Client Side uses the tcp-wan-optimized profile, the Server Side uses the tcp-lan-optimized and I have disabled the Naggle, Slow Start and Bandwidth Delay. It also uses the http profile.

 

So, my question is, where would I start looking or what could be the issue for handsets that make HTTP requests?

 

Please let me know if you need any more information or if you have any ideas...?

 

Thanks in advance

 

Phil

2 Replies

  • Just as a quick test, try using the lan-optimized profile on the client side as well. Sounds counterintuitive, but I've had it solve latency problems quite a few times.

     

     

    Denny
  • Thanks dennypayne,

     

     

    unfortunately when we test it with the lan-optimized it forces the speed down when the laptops are using the test APN.

     

     

    so instead of getting 350kb/s we get 250kb/s.

     

     

    It's an odd one because as far as we are concerned the handsets are doing exactly the same kind of HTTP requests as the laptops... I am going to try and use firefox to emulate a client and see if that makes any difference however these same client devices all work ok through the Live APN