Sign in to follow this  
Followers 0
CA3LE

Speed slower than it usually is

1 post in this topic

I actually have two questions.

1.)  Do you currently have any test servers in Tampa?  I ask because up until today I never say the CDN Push => Tampa, FL (it was always TX).  Also, I noted that your site IS more reliable when testing my speed over a distance, FL to TX for example.  However I noticed that the CDN Push (to a local server) is SLOWER than what my ISP offers.  I am testing from two different locations.  The first is my house where I have an Airport Extremem (802.11ac model) and a macbook air with a connection between the two ranging from 500Mbps to usually around 100Mbps (much more than my 50Mbps).

Speedtest.net (Just one of my reference tools) shows that I usually get around 58-60 Mbps (pretty nice for Fios to go over the limit).  However, your tool fluctuates MUCH MUCH more.  Sometimes you will match Speedtest.net almost exactly, other times you vary wildly even down to 5mbps.  Now I KNOW my speed isn't that bad because I immediately test again with both Fios's test (I know it's biased) and with Speedtest.net (sometimes with other Java based ones but I try to avoid those usually).  They usually, barring some weird network thing which happens every 100 days (wiring in my house seems to get funky after a while) they report normal.  Is this happening because your getting a lot of hits or is there something on my end that I'm missing that I need to fix?

2.)  What is the recommended upload/download speed to host a testmy.net server?  I am looking into hosting one, but I want to ensure that if I do I have the most appropriate speed.  I currently have 50/25 on Fios but I don't think this is enough for you.  Can you please verify if there are any such requirements?

Thanks for your time with these matters,

Andre N.

 

Hi Andre,

 
Yes, there is a TMN server in Florida.  It's not my own but is hosted by a powerful host.  I've seen some crazy speed out of sietec's servers.  You can see it at http://testmy.net/mirror/sietec set this sietec in Miami FL to default.  Read sietec's two cents, posted in response to someone questioning TMN results.  Interesting how far he put my method under the microscope.
 
But you actually misunderstood the "CDN Push => Tampa, FL" means that it's testing from CDN Push to Tampa FL... if it said TX before it was due to your location being incorrectly detected.  Hosts move IP addresses around and my databases for connecting your IP to your location are only 83% accurate to the city level.  Especially true for mobile users, who can grab an IP from one area and then drive it with them to a new city.  It's hard to detect that without being invasive.  It's usually correct.  In the future I'll implement GPS and the ability to define location.
 
CDN Push uses CloudFlare's network.  Looking at the network map you should be testing out of Miami using that method.  If you get worse results using that method it may be due to route congestion.  Personally, my home connection is weaker using CDN Push but my verizon 4G LTE usually fairs better.  CDN method is good to try because it's the way that many websites are delivering large volume and static content.  When I test from my home connection the CDN method seems to have a little lag, then huge burst... on my LTE it's just a huge burst.  So, something along that route to my Cox connection must be causing that.  I test using other servers and I never see that.  I offer an array of testing methods because there is an array of possibilities to how you download and how the download is delivered.  Good, clean running ISPs have less variance between the different servers and testing methods.  
 
I always suggest testing with my main server and comparing those results to the other mirrors.  The default is the most powerful and overstocked server I have.  It's also centrally located in the US right at the heart of server valley.  Besides, the location of the server is irrelevant.  Others and myself have proven many times that distance does not have to be a factor.  I have records of people pulling many hundreds of Mbps across 1/2 the planet, and still have super low ping times.  There is a HUGE misconception that you need to be close to the server to achieve your speeds.  This is a misconception that your ISP loves... but the truth is... if the intermediate links were better there's no reason why you couldn't pull your full speeds at any distance, under the right network configuration and circumstance.
 
Regarding speedtest.net... read Why do my results differ from ookla / speedtest.net - there are MANY factors that differ between TestMy.net and ookla speed tests.  
 
The one major one that sticks out in my mind when I hear a FiOS customer with speeds like yours... multithreading.  What I see more often than not with FiOS (and 100+ Mbps consumer level broadband in general) is that you need to multithread to see anywhere near the top speed.  This, I believe is due to either bandwidth shaping or due to the connection being split over channels.  Commercial connections I work with will often perform better or at least the same with my classic test.  If you take a multithread speed test it will be instantly obvious.  My own home connection is effected, Cox Communications supposed 150/25... Using the classic linear speed test I get around 50-60 Mbps but when I multithread it opens up to nearly 90 Mbps.  I know it can't do any higher because sometimes it will cause Pandora to skip and have issues... meaning I've sucked my connection dry.  The max I've seen on Giganews is 92 Mbps, across 50 threads.   ... in my opinion, you shouldn't have to open multiple threads to max out your connection.  I leave my classic method as default because I'm not here to stroke egos, I'm here to help you all get the most out of your Internet.
 
Okay... question 2, lol.
 
The make a speed test is intended for 100 Mbps uplinks and higher.  I personally use a minimum of 1000 Mbps on all my mirrors, my home server has 2 1000 Mbps uplinks on a round robin to balance the load across both interfaces.  ... there is A LOT of head room.  One of the many differences at TestMy.net is that I truly want to deliver the highest level of accuracy.  Rule number 1, a speed test can't be accurate if there isn't enough bandwidth.
 
So, your slow down is not because I'm getting a lot of hits... TMN gets a lot of traffic but I closely monitor what's going on.  I have live monitors on my phone with me at all times.  The mother server I have right now is so powerful I estimate that I can load it with at least 10X the traffic I have right now and users will still be unaffected by other users testing.  It's a 24 core beast, to be honest it's by far the most powerful machine I've ever had the pleasure of working with.  I'll have it replaced again before overloading it becomes an issue and if my traffic exploded insanely overnight, I can have more beasts online in minutes.  >:-) --- it's never an issue but if you wanted to double check it's as easy as going to http://testmy.net/live and look at the results others are getting.  Also, researching into recent results for your city, country or ISP can show if others locally are able to get better speeds than you are.  I see MANY verizon users in your area pulling better speeds than you are.  So you might even have an issue with your TCP or internal network.  Which often don't show up on other speed tests... because most other speed tests multithread by default and actually ignore portions of the result.  Leading to a false sense of pride.
 
You took the time to write a lengthy query so I wrote a lengthy respond... I hope you read it, it helps and you learn from it.
 
- Cheers,
- Damon - TestMy.net

Share this post


Link to post
Share on other sites
Sign in to follow this  
Followers 0