Page 3 of 4 FirstFirst 1234 LastLast
Results 51 to 75 of 97
  1. #51
    Join Date
    Mar 2008
    Location
    Los Angeles, CA
    Posts
    555
    Quote Originally Posted by jo2jo View Post
    Arent they the best!! i created an pretty large pay per use wifi network with them. run like champs in all kinds of weather. everyone should take a look at mikrotik. changed my life (not to get too mushy)
    Testing to your server from my home connection is kind of acting like some sort of rate limiting (QoS or speed limiting by http daemon) is going on. When I first started the transfer for the first second I got 17 megabytes/sec (130 mbit) but it quickly died down and trying to download it again I was getting like 300 kilobytes/sec. Oddly enough I then tried pinging your server while downloading just to see what its latency was and I got nearly the full speed of my connection on the whole file:

    While downloading:

    Code:
    64 bytes from 216.45.49.251: icmp_seq=3 ttl=252 time=5.45 ms
    64 bytes from 216.45.49.251: icmp_seq=4 ttl=252 time=37.9 ms
    64 bytes from 216.45.49.251: icmp_seq=5 ttl=252 time=36.5 ms
    64 bytes from 216.45.49.251: icmp_seq=6 ttl=252 time=29.9 ms
    64 bytes from 216.45.49.251: icmp_seq=7 ttl=252 time=31.0 ms
    64 bytes from 216.45.49.251: icmp_seq=8 ttl=252 time=26.9 ms
    64 bytes from 216.45.49.251: icmp_seq=9 ttl=252 time=5.39 ms
    64 bytes from 216.45.49.251: icmp_seq=10 ttl=252 time=21.6 ms
    64 bytes from 216.45.49.251: icmp_seq=11 ttl=252 time=10.1 ms
    64 bytes from 216.45.49.251: icmp_seq=12 ttl=252 time=5.58 ms
    64 bytes from 216.45.49.251: icmp_seq=13 ttl=252 time=5.40 ms
    64 bytes from 216.45.49.251: icmp_seq=14 ttl=252 time=11.3 ms
    64 bytes from 216.45.49.251: icmp_seq=15 ttl=252 time=5.42 ms
    64 bytes from 216.45.49.251: icmp_seq=16 ttl=252 time=8.83 ms
    64 bytes from 216.45.49.251: icmp_seq=17 ttl=252 time=7.87 ms
    64 bytes from 216.45.49.251: icmp_seq=18 ttl=252 time=8.00 ms
    64 bytes from 216.45.49.251: icmp_seq=19 ttl=252 time=5.45 ms
    64 bytes from 216.45.49.251: icmp_seq=20 ttl=252 time=6.79 ms
    64 bytes from 216.45.49.251: icmp_seq=21 ttl=252 time=5.38 ms
    64 bytes from 216.45.49.251: icmp_seq=22 ttl=252 time=5.40 ms
    64 bytes from 216.45.49.251: icmp_seq=23 ttl=252 time=5.41 ms
    64 bytes from 216.45.49.251: icmp_seq=24 ttl=252 time=8.82 ms
    64 bytes from 216.45.49.251: icmp_seq=25 ttl=252 time=11.0 ms
    64 bytes from 216.45.49.251: icmp_seq=26 ttl=252 time=12.9 ms
    64 bytes from 216.45.49.251: icmp_seq=27 ttl=252 time=5.45 ms
    64 bytes from 216.45.49.251: icmp_seq=28 ttl=252 time=10.0 ms
    64 bytes from 216.45.49.251: icmp_seq=29 ttl=252 time=5.67 ms
    64 bytes from 216.45.49.251: icmp_seq=30 ttl=252 time=12.0 ms
    64 bytes from 216.45.49.251: icmp_seq=31 ttl=252 time=13.7 ms
    64 bytes from 216.45.49.251: icmp_seq=32 ttl=252 time=13.1 ms
    64 bytes from 216.45.49.251: icmp_seq=33 ttl=252 time=5.40 ms
    64 bytes from 216.45.49.251: icmp_seq=34 ttl=252 time=5.42 ms
    64 bytes from 216.45.49.251: icmp_seq=35 ttl=252 time=8.45 ms
    64 bytes from 216.45.49.251: icmp_seq=36 ttl=252 time=16.1 ms
    64 bytes from 216.45.49.251: icmp_seq=37 ttl=252 time=7.20 ms
    64 bytes from 216.45.49.251: icmp_seq=38 ttl=252 time=10.4 ms
    64 bytes from 216.45.49.251: icmp_seq=39 ttl=252 time=13.9 ms
    64 bytes from 216.45.49.251: icmp_seq=40 ttl=252 time=5.47 ms
    64 bytes from 216.45.49.251: icmp_seq=41 ttl=252 time=12.9 ms
    64 bytes from 216.45.49.251: icmp_seq=42 ttl=252 time=8.41 ms
    64 bytes from 216.45.49.251: icmp_seq=43 ttl=252 time=8.49 ms
    64 bytes from 216.45.49.251: icmp_seq=44 ttl=252 time=6.72 ms
    64 bytes from 216.45.49.251: icmp_seq=45 ttl=252 time=5.95 ms
    64 bytes from 216.45.49.251: icmp_seq=46 ttl=252 time=8.32 ms
    64 bytes from 216.45.49.251: icmp_seq=47 ttl=252 time=9.08 ms
    64 bytes from 216.45.49.251: icmp_seq=48 ttl=252 time=10.6 ms
    64 bytes from 216.45.49.251: icmp_seq=49 ttl=252 time=10.5 ms
    64 bytes from 216.45.49.251: icmp_seq=50 ttl=252 time=5.40 ms
    64 bytes from 216.45.49.251: icmp_seq=51 ttl=252 time=8.11 ms
    64 bytes from 216.45.49.251: icmp_seq=52 ttl=252 time=7.37 ms
    64 bytes from 216.45.49.251: icmp_seq=53 ttl=252 time=6.61 ms
    64 bytes from 216.45.49.251: icmp_seq=54 ttl=252 time=7.36 ms
    64 bytes from 216.45.49.251: icmp_seq=55 ttl=252 time=5.53 ms
    64 bytes from 216.45.49.251: icmp_seq=56 ttl=252 time=6.66 ms
    64 bytes from 216.45.49.251: icmp_seq=57 ttl=252 time=10.6 ms
    64 bytes from 216.45.49.251: icmp_seq=58 ttl=252 time=10.5 ms
    64 bytes from 216.45.49.251: icmp_seq=59 ttl=252 time=12.4 ms
    64 bytes from 216.45.49.251: icmp_seq=60 ttl=252 time=11.1 ms
    64 bytes from 216.45.49.251: icmp_seq=61 ttl=252 time=11.2 ms
    64 bytes from 216.45.49.251: icmp_seq=62 ttl=252 time=9.82 ms
    64 bytes from 216.45.49.251: icmp_seq=63 ttl=252 time=9.87 ms
    64 bytes from 216.45.49.251: icmp_seq=64 ttl=252 time=12.3 ms
    64 bytes from 216.45.49.251: icmp_seq=65 ttl=252 time=16.3 ms
    64 bytes from 216.45.49.251: icmp_seq=66 ttl=252 time=12.0 ms
    64 bytes from 216.45.49.251: icmp_seq=67 ttl=252 time=8.65 ms
    64 bytes from 216.45.49.251: icmp_seq=68 ttl=252 time=5.42 ms
    64 bytes from 216.45.49.251: icmp_seq=69 ttl=252 time=5.38 ms
    Those latency spikes were caused by when the connection was getting close or being saturated as this was from my home connection.

    Code:
    dekabutsu ~ # wget -O /dev/null http://media7.ipadporn.com:1049/videos/1gb.bin
    --23:58:06--  http://media7.ipadporn.com:1049/videos/1gb.bin
               => `/dev/null'
    Resolving media7.ipadporn.com... 216.45.49.251
    Connecting to media7.ipadporn.com|216.45.49.251|:1049... connected.
    HTTP request sent, awaiting response... 200 OK
    Length: 1,073,741,824 (1.0G) [application/octet-stream]
    
    100%[==============================================>] 1,073,741,824   15.45M/s    ETA 00:00
    
    23:59:11 (15.81 MB/s) - `/dev/null' saved [1073741824/1073741824]
    It seems extremely odd to me that something going over transit would be getting lower speeds than something on peering.

    Traceroute:

    Code:
    traceroute to 216.45.49.251 (216.45.49.251), 30 hops max, 46 byte packets
     1  router.houkouonchi.jp (1.1.1.1)  0.192 ms  0.125 ms  0.077 ms
     2  L103.LSANCA-DSL-29.verizon-gni.net (71.110.63.1)  1.037 ms  0.850 ms  0.807 ms
     3  G0-3-1-7.LSANCA-LCR-22.verizon-gni.net (130.81.146.60)  8.735 ms  1.748 ms  1.175 ms
     4  so-4-1-0-0.LAX01-BB-RTR2.verizon-gni.net (130.81.151.248)  21.260 ms  18.383 ms  51.442 ms
     5  0.ae2.BR3.LAX15.ALTER.NET (152.63.2.133)  2.701 ms  2.199 ms  2.212 ms
     6  ge8-3-10G.ar4.LAX2.gblx.net (64.212.107.121)  5.289 ms  5.196 ms  5.483 ms
     7  xe-3-0-3.ar1.lax1.us.nlayer.net (69.31.127.213)  10.185 ms  6.441 ms  9.757 ms
     8  ae0-60g.cr1.lax1.us.nlayer.net (69.31.127.141)  2.454 ms  2.452 ms  2.343 ms
     9  ae1-50g.ar1.lax2.us.nlayer.net (69.31.127.130)  6.421 ms  4.256 ms  3.786 ms
    10  as29761.xe-1-0-2.ar1.lax2.us.nlayer.net (69.31.121.254)  2.481 ms  2.541 ms  2.437 ms
    11  lax13.distrib.core01.quadranet.com (67.215.251.230)  5.476 ms  5.434 ms  5.673 ms
    12  216.45.49.251 (216.45.49.251)  5.633 ms  5.343 ms  5.392 ms

    Compared to the gig server that is seeing lower speeds:

    Code:
    # traceroute -I 216.45.49.251
    traceroute to 216.45.49.251 (216.45.49.251), 30 hops max, 46 byte packets
     1  ip-66-33-193-1 (66.33.193.1)  0.664 ms  1.018 ms  0.546 ms
     2  ip-66-33-201-113 (66.33.201.113)  0.454 ms  0.755 ms  0.312 ms
     3  peering.1wh.la.quadranet.com.any2ix.coresite.com (206.223.143.159)  0.595 ms  0.699 ms  0.478 ms
     4  96.44.180.98 (96.44.180.98)  0.639 ms  0.744 ms  0.500 ms
     5  216.45.49.251 (216.45.49.251)  0.507 ms  0.676 ms  0.465 ms
    Usually when network issues do occur the lower the latency the better the performance is (this is generally true as well).

    Opening 20 connections from my colo'd 1gbit server resulted in around 70 megabytes/sec (560 megabits) download. A single connection still won't go past 6-9 megabytes/sec though.

  2. #52
    Join Date
    Jan 2010
    Posts
    165
    oh im doing alot of lighttpd restarting and tweaking of the server, im sorry. I should have let you all know. im basing it upon your suggestion of using wget locally to pull different sized files and tweaking the lighttpd.conf. i still think once the max you can push UP (not down, down is fast) this quadranet line is 200-250mbit. but we will see and i will keep trying. im waitng for my main admin to get up and see what he says, but he did alot of the tuning along with me on this server based upon other servers that preform the same role for other sites with HUGE demand.

  3. #53
    Join Date
    Mar 2008
    Location
    Los Angeles, CA
    Posts
    555
    jo2jo I would try using the web100 tool to my server to see what it says:

    http://software.internet2.edu/source...t-3.6.4.tar.gz

    Sometimes the whole package doesn't compile but usually the web100clt will and will be located in:

    ./src/web100clt

    This can give some interesting details about what the actual TCP receive/send buffers are as well as retransmits, out of order packets and other stuff as well.

    Here is some output between my home machine and my server (on 1000/1000):

    Code:
    dekabutsu ndt-3.6.4 # src/web100clt -l -n  208.97.141.21
    Testing network path for configuration and performance problems  --  Using IPv4 address
    Checking for Middleboxes . . . . . . . . . . . . . . . . . .  Done
    checking for firewalls . . . . . . . . . . . . . . . . . . .  Done
    running 10s outbound test (client to server) . . . . .  72.21 Mb/s
    running 10s inbound test (server to client) . . . . . . 148.80 Mb/s
    The slowest link in the end-to-end path is a 100 Mbps Full duplex Fast Ethernet subnet
    Information: Other network traffic is congesting the link
    Server '208.97.141.21' is not behind a firewall. [Connection to the ephemeral port was successful]
    Client is probably behind a firewall. [Connection to the ephemeral port failed]
    
            ------  Web100 Detailed Analysis  ------
    
    Web100 reports the Round trip time = 9.20 msec;the Packet size = 1448 Bytes; and
    No packet loss - but packets arrived out-of-order 0.32% of the time.
    This connection is sender limited 92.23% of the time.
    This connection is network limited 7.77% of the time.
    
        Web100 reports TCP negotiated the optional Performance Settings to:
    RFC 2018 Selective Acknowledgment: ON
    RFC 896 Nagle Algorithm: ON
    RFC 3168 Explicit Congestion Notification: OFF
    RFC 1323 Time Stamping: ON
    RFC 1323 Window Scaling: ON; Scaling Factors - Server=7, Client=7
    The theoretical network limit is 1200.37 Mbps
    The NDT server has a 256 KByte buffer which limits the throughput to 217.32 Mbps
    Your PC/Workstation has a 673 KByte buffer which limits the throughput to 571.42 Mbps
    The network based flow control limits the throughput to 387.73 Mbps
    
    Client Data reports link is '  5', Client Acks report link is '  5'
    Server Data reports link is '  6', Server Acks report link is '  6'
    Packet size is preserved End-to-End
    Information: Network Address Translation (NAT) box is modifying the Server's IP address
            Server says [208.97.141.21] but Client says [ ndt.dhspeedtest.com]
    Information: Network Address Translation (NAT) box is modifying the Client's IP address
            Server says [71.110.63.2] but Client says [ 1.1.1.3]

  4. #54
    Join Date
    Apr 2009
    Location
    Spain
    Posts
    130
    I can only get 60-70mbps on your test file from the server in quadranet.
    but I can get 100+mbps from the server in other DC.
    really weird.

    BTW: I have 50+ servers at quadranet, no problem at all.

  5. #55
    Join Date
    Mar 2003
    Location
    /root
    Posts
    23,991
    We have 1Gbps line with Quadranet and we can use it fine.

    You must have a problem with your server network configuration.

    You better check it my friend.

    Specially 4 U
    Reseller Hosting: Boost Your Websites | Fully Managed KVM VPS: 3.20 - 5.00 Ghz, Pure Dedicated Power
    JoneSolutions.Com is on the net 24/7 providing stable and reliable web hosting solutions, server management and services since 2001
    Debian|Ubuntu|cPanel|DirectAdmin|Enhance|Webuzo|Acronis|Estela|BitNinja|Nginx

  6. #56
    Join Date
    Jan 2010
    Posts
    165
    >>I can only get 60-70mbps on your test file from the server in quadranet.
    but I can get 100+mbps from the server in other DC.
    really weird.


    net what kind of upstream are you seeing on your server? i can download just fine on my quadranet, its the upstream. And your finding match up exactly with my 100mbit -175mbit max as the server was pushing 50-75mbit when you ran your test.


    again as of this morning, pointing my entire site at each of my gig servers for a little bit showed (so a single server gets teh full load). IM seeing ~110mbit on quadranet. When i do this to my server at wholesale i see right at 400mbit. Maybe its bc i have more users on the east coast? but still difference should not be that large. streaming lighttpd is a great way to check the bandwidth of a provider as my users are literaly all over the USA (remember world traffic gets routed to it appropriate server).

    Cgozman sent me a image of another user on my swtich showing him using 200mbit. Again my issue is beyond 200mbit. Many may dismiss this as petty, but that is what i paid for, and quadranet has yet to show me beyond 200mbit on upstream. Im not saying it doesnt exist, but i really need to see numbers. Net what does we can use 1gbit just fine mean. is that 50mbit upstream? 500mbit? all downstream?

    There is nothing wrong with this server, here is a local wget (i know this does not prove that much).

    [root@media7 ~]# wget -O /dev/null http://media7.ipadporn.com:1049/videos/10gb.bin
    --2011-10-09 10:23:08-- http://media7.ipadporn.com:1049/videos/10gb.bin
    Resolving media7.ipadporn.com... 127.0.0.1
    Connecting to media7.ipadporn.com|127.0.0.1|:1049... connected.
    HTTP request sent, awaiting response... 200 OK
    Length: 8589934592 (8.0G) [application/octet-stream]
    Saving to: `/dev/null'

    30% [======================> ] 2,641,498,112 298M/s eta 26s


    -=-=-= let me tell everyone how serious i am, im willing to build a new server, and pay to send to quadranet - this last server cost around $ 4k.

    btw im assuming that Cgotzman, VP of quadranet, is now ignoring me (since last night) as he doest reply to my messages but comes on here to complain about my questions. This entire situation took a more aggressive tone when HE got involved, you can look back at his first message to see this. again a big part of this thread is so that others can see this and know how quadranet and their network work. I know if i had saw this 4 months ago when shopping one of my colos i would have RAN. but instead i was re assured by sales that quadranet is all about speed and obsessed with speed and i will never have issues bc gig is nothing to us. well Wholesale has done ALOT more than you guys for getting me the speed i pay for, and with no attitude or rude VP. Guess what, they said it was my machine at first TOOOOOOO!
    Last edited by jo2jo; 10-09-2011 at 11:35 AM.

  7. #57
    Join Date
    Feb 2011
    Posts
    680
    Where are you testing from? Could there be a narrow pipe in the route to Quadranet but not HE? What do the traceroutes look like to Quadranet and HE from your test site?

    What if you run the test from Quadranet to HE and HE to Quadranet. Remember that again this is not a definitive test as the routes can (and most probably will) differ for outgoing and incoming traffic. You can do a traceroute from Quadranet to HE then HE to Quadranet.

    Dave.

  8. #58
    Join Date
    Jan 2010
    Posts
    165
    Dave tks for contributing, but this is not about one specific route.

    tks

  9. #59
    Join Date
    Jun 2002
    Location
    Waco, TX
    Posts
    5,623
    I can download about 2x faster from the quadranet la test file download link than from this users server. Don't know exactly how to analyze that since neither were that fast at 3.5MB/s and 6.8MB/s

    Downloaded from akamai at over 18MB/s for an 4GB file as a comparison.

  10. #60
    Join Date
    Feb 2011
    Posts
    680
    Quote Originally Posted by jo2jo View Post
    Dave tks for contributing, but this is not about one specific route.

    tks
    How do you know? Even if you test from a couple of places perhaps they are sharing a node or a net some place.


    Dave.

  11. #61
    Join Date
    Mar 2008
    Location
    Los Angeles, CA
    Posts
    555
    jo2jo: Your server is not capped at low speed. I said earlier with 20 connetions I was able to get over 500 megabits/sec (70 megabytes/sec). Also 300 megabytes/sec to that sparse file (if its a sparse file) is actually kind of weak because my machine gets 1.5GB/sec:

    Code:
    ~# wget -O /dev/null http://box.houkouonchi.jp/5tb.bin
    --08:59:11--  http://box.houkouonchi.jp/5tb.bin
               => `/dev/null'
    Resolving box.houkouonchi.jp... 2607:f298:1:100:feed:face:beef:d00d, 208.97.140.21
    Connecting to box.houkouonchi.jp|2607:f298:1:100:feed:face:beef:d00d|:80... connected.
    HTTP request sent, awaiting response... 200 OK
    Length: 5,497,558,138,880 (5.0T) [application/octet-stream]
    
     0% [                                                            ] 17,751,240,704    1.56G/s    ETA 55:31

    that is with apache pretty much getting maxed out CPU: I would honestly expect better from from a machine that likely isn't as loaded as mine as mine was under heavy I/O and already pushing 200mbits outbound:

    Code:
      PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
    10864 apache    20   0  144m  16m 4396 R   94  0.1   0:07.64 apache2
    17969 root      20   0 92960 1804 1428 R   77  0.0   0:05.70 wget
    I assume you aren't seeing high %irq, %soft or even possibly %sys on your machine right?

    You can see values with:

    mpstat 5

    Here is my server but its under pretty dencetly heavy load right now:

    Code:
    09:05:00 AM  CPU   %user   %nice    %sys %iowait    %irq   %soft  %steal   %idle    intr/s
    09:05:05 AM  all    5.54    4.32    5.81   19.26    0.02    5.36    0.00   59.68  11049.50
    09:05:10 AM  all    4.72    0.20    3.89   10.43    0.02    4.49    0.00   76.25  11335.73
    09:05:15 AM  all    5.08    0.07    2.56    8.04    0.10    5.13    0.00   79.01  12434.26
    09:05:20 AM  all    6.00    0.07    2.12    8.74    0.05    4.78    0.00   78.24  12011.75
    09:05:25 AM  all    3.81    0.57    1.62    8.54    0.05    4.31    0.00   81.10  11564.34
    09:05:30 AM  all    5.08    0.00    2.07    8.07    0.07    5.58    0.00   79.13  12371.51
    09:05:35 AM  all    5.55    2.14    5.93    8.49    0.02    4.18    0.00   73.68  12939.24
    09:05:40 AM  all    6.42    1.62   10.58    9.76    0.00    4.38    0.00   67.23  15041.24
    09:05:45 AM  all    4.36    0.00    1.22    9.14    0.05    4.98    0.00   80.25  13157.09
    What kind of nics does the server have? If supermicro I assume its intel as my box is also supermicro:

    Code:
    09:07 AM :~# dmidecode | grep -i product\ name
            Product Name: X8DTU
            Product Name: X8DTU
    09:07 AM :~# lspci | grep -i ether
    01:00.0 Ethernet controller: Intel Corporation Unknown device 10c9 (rev 01)
    01:00.1 Ethernet controller: Intel Corporation Unknown device 10c9 (rev 01)
    0b:00.0 Ethernet controller: Intel Corporation 82571EB Gigabit Ethernet Controller (rev 06)
    0b:00.1 Ethernet controller: Intel Corporation 82571EB Gigabit Ethernet Controller (rev 06)
    0c:00.0 Ethernet controller: Intel Corporation 82571EB Gigabit Ethernet Controller (rev 06)
    0c:00.1 Ethernet controller: Intel Corporation 82571EB Gigabit Ethernet Controller (rev 06)
    09:07 AM :~#

  12. #62
    Join Date
    Jun 2011
    Location
    Miami, FL
    Posts
    825
    Quote Originally Posted by jo2jo View Post
    >>
    There is nothing wrong with this server, here is a local wget (i know this does not prove that much).

    -=-=-= let me tell everyone how serious i am, im willing to build a new server, and pay to send to quadranet - this last server cost around $ 4k.
    You're willing to spend around $4K to put a new server in, but aren't willing to do some basic troubleshooting that has been recommended by the posters here...(Have you removed/powered down the router as I suggested yet?)

    That doesn't make sense to me and I'm beginning to understand why QuadraNet is getting frustrated....

  13. #63
    Join Date
    Jan 2010
    Posts
    165
    Now that i think back, i remember being disappointed in the list of transit provdiers that quadranet Sales provided to me. Not one major name that i recognized except HE (and HE was their slowest link). Im not saying i know much about transit or backbones or even how a DCs upstrems work, BUT i was reassured by sales and posts here that quadranet was ALL ABOUT SPEED and all about their network. Well just dont ask for more than 200mbit upstream. Im not asking for a free lunch while paying budget prices for a server, i have offered to pay for anything related to fixing this issue, remote hands, hardware, whatever. Again not right for a company that says they are all about speed and their network is king. Just look at the lousy speeds some are getting so far on tests NOT to my server. (not trying to be anecdotal, just going off the facts i have.)


    Again not one person at quadranet nor any quadranet customer has shown any tests or info that shows their ports are capable of over 200mbit. Im not saying they are NOT, but i know ive given a hell of a go at it in terms of HARDWARE and time/admin work. I realize this may sound petty and minute to some people, but if you need 1gbit and ask for it, (again i realize you only get a FULL gbit on a 1000$/mo + dedicated gig line)

  14. #64
    Join Date
    Jun 2011
    Location
    Miami, FL
    Posts
    825
    Quote Originally Posted by jo2jo View Post
    >>well Wholesale has done ALOT more than you guys for getting me the speed i pay for, and with no attitude or rude VP. Guess what, they said it was my machine at first TOOOOOOO!
    So you've had this issue before as well? What did Wholesale do to correct the issue (Hint: Anything on your machine?)

  15. #65
    Join Date
    Mar 2008
    Location
    Los Angeles, CA
    Posts
    555
    The issue is definitely not transit anyway. Like I said when testing from my server on 1000m its going over any2 peering link so its completely taking transit out of the equation. Through that same route I get 50 megabytes/sec to pacific racks test file I get 5-6 MB/sec to you. With 20 connections I can get 70 megabytes/sec.

    Your wget results still seemed kind of low. I even tested on a VPS I have setup with lighttpd and I still saw double your speeds and this is a loaded VPS:

    Code:
    [ps8992]$ wget -O /dev/null http://houkouonchi.net:8080/10gb.bin
    --2011-10-09 09:10:38--  http://houkouonchi.net:8080/10gb.bin
    Resolving houkouonchi.net... 69.163.156.178, 2607:f298:1:103::1e8:b3e7
    Connecting to houkouonchi.net|69.163.156.178|:8080... connected.
    HTTP request sent, awaiting response... 200 OK
    Length: 10737418240 (10G) [application/octet-stream]
    Saving to: `/dev/null'
    
    100%[===============================================================================================>] 10,737,418,240  653M/s   in 16s
    
    2011-10-09 09:10:54 (639 MB/s) - `/dev/null' saved [10737418240/10737418240]
    Code:
      PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
    28750 houkouon  20   0 43580 7580  680 R   82  0.6 130:50.82 lighttpd
     2965 houkouon  20   0  106m 1852 1468 S   52  0.1   0:04.75 wget

    Your NIC is properly offloading everything to the NIC right?

    Code:
    # ethtool -k eth1
    Offload parameters for eth1:
    rx-checksumming: on
    tx-checksumming: on
    scatter-gather: on
    tcp segmentation offload: on
    udp fragmentation offload: off
    generic segmentation offload: on
    This is important when we are talking about gigabit level speeds.

  16. #66
    Join Date
    Feb 2011
    Posts
    680
    I think the problem (at least some of it) is just internet weather.

    If I download the quadranet file normally I get 2.15 MB/s If I remove my HE peering session and traffic goes via Verizon and I now get 11.4Meg

    Looking at the trace routes I would suspect nlayer as the possible bottle neck.

    [root@ocean ~]# traceroute quadranet.com traceroute to quadranet.com (98.143.159.202), 30 hops max, 38 byte packets
    1 gige-0-1.core1.silvaco.com (192.73.228.1) 0.177 ms 0.108 ms 0.103 ms
    2 gige-v31.qts.silvaco.com (63.80.200.73) 0.461 ms 0.449 ms 0.477 ms
    3 GigabitEthernet0-1-1-5.GW3.SJC7.ALTER.NET (152.179.48.65) 1.725 ms 1.244 ms 1.156 ms
    4 tinet-gw.customer.alter.net (152.179.48.18) 0.965 ms 0.948 ms 0.974 ms
    5 xe-10-0-0.lax20.ip4.tinet.net (89.149.182.70) 9.263 ms xe-3-3-0.lax20.ip4.tinet.net (89.149.184.149) 8.934 ms xe-9-0-0.lax20.ip4.tinet.net (89.149.186.186) 8.607 ms
    6 quadranet-gw.ip4.tinet.net (77.67.68.238) 9.302 ms 9.149 ms 8.888 ms
    7 96.44.180.245 (96.44.180.245) 11.635 ms 11.172 ms 11.564 ms

    8 *
    [root@ocean ~]# wget -O /dev/null http://quadranet.com/speedtests/100mb.test --09:17:54-- http://quadranet.com/speedtests/100mb.test
    => `/dev/null'
    Resolving quadranet.com... 98.143.159.202
    Connecting to quadranet.com|98.143.159.202|:80... connected.
    HTTP request sent, awaiting response... 200 OK
    Length: 105,472,000 (101M) [text/plain]

    100%[====================================>] 105,472,000 11.42M/s ETA 00:00

    09:18:03 (11.33 MB/s) - `/dev/null' saved [105472000/105472000]

    [root@ocean ~]# traceroute quadranet.com
    traceroute to quadranet.com (98.143.159.202), 30 hops max, 38 byte packets
    1 gige-0-1.core1.silvaco.com (192.73.228.1) 0.138 ms 0.105 ms 0.104 ms
    2 gige-0-1.core2.silvaco.com (192.73.228.2) 0.146 ms 0.107 ms 0.104 ms
    3 gige-v21.he.silvaco.com (65.19.157.249) 4.821 ms 4.793 ms 4.792 ms
    4 gige-2-0-1.he.silvaco.com (64.71.131.153) 5.093 ms 4.878 ms 14.843 ms
    5 10.123.123.6 (10.123.123.6) 7.050 ms 15.378 ms 5.214 ms
    6 paix.99.xe-0-1-0.cr1.pao1.us.nlayer.net (198.32.176.13) 5.212 ms 5.302 ms 5.207 ms
    7 ae1-50g.cr1.sjc1.us.nlayer.net (69.22.143.165) 5.950 ms 6.095 ms 5.844 ms
    MPLS Label=682232 CoS=6 TTL=1 S=0
    8 xe-4-1-0.cr1.lax1.us.nlayer.net (69.22.142.53) 13.831 ms xe-5-1-0.cr1.lax1.us.nlayer.net (69.22.142.126) 54.621 ms xe-4-1-0.cr1.lax1.us.nlayer.net (69.22.142.53) 13.838 ms
    9 ae1-50g.ar1.lax2.us.nlayer.net (69.31.127.130) 17.553 ms 18.485 ms 17.076 ms
    10 as29761.xe-1-0-2.ar1.lax2.us.nlayer.net (69.31.121.254) 13.890 ms 13.840 ms 14.047 ms

    [root@ocean ~]# wget -O /dev/null http://quadranet.com/speedtests/100mb.test
    --09:18:58-- http://quadranet.com/speedtests/100mb.test
    => `/dev/null'
    Resolving quadranet.com... 98.143.159.202
    Connecting to quadranet.com|98.143.159.202|:80... connected.
    HTTP request sent, awaiting response... 200 OK
    Length: 105,472,000 (101M) [text/plain]

    100%[====================================>] 105,472,000 1.84M/s ETA 00:00

    09:19:45 (2.15 MB/s) - `/dev/null' saved [105472000/105472000]

  17. #67
    Join Date
    Mar 2008
    Location
    Los Angeles, CA
    Posts
    555
    jo2jo if you have MSN or AIM or something you can PM me with your info. I am willing to talk to you on chat to help you rule out some stuff (if you are interested). I understand your frustration but if pacific rack already moved you to a completely different floor and your still having the same problem I understand why they think they have done their part and concluded the problem is with your system.

    There is *definitely* a problem here, whether it is indeed on pacific racks end or your server has definitely not been proven but I will definitely say there is an issue and things aren't going the speed they should be.

    I manage a friends server which is on pacific rack's network. I usually always see very good latency/speeds on their network. I can tell you right now I get faster off his server on 100m port than I do on yours which is on 1000m. I get a very stable 11.2 megabytes/sec to his which is right what I would expect to max out on with only a 100mbit port.

  18. #68
    Join Date
    Jan 2010
    Posts
    165
    MiIT guy, please read the entire thread, ive said 2 or 3 times, the router was removed nearly a month ago during trouble shooting. Quadranet will agree ive been troubleshooting this for about 2 months off and on 90% of that has been on my side trying to tweak server, and the issue i spoke of with wholesale was them adding a Gigabit connection to my server bc they dont have gig ports. It was resolved when they tried to hook up a server to the line and could not.

    I really apprecaite dave houkouonchi's help and will report back from my server, but are you saying you got 500mbit from my server with 20 connections? Im watching with nload and 15 sec mrtg graphs and there has been no usage over 70mbit.

    tks

  19. #69
    Join Date
    Mar 2008
    Location
    Los Angeles, CA
    Posts
    555
    Quote Originally Posted by jo2jo View Post
    I really apprecaite dave houkouonchi's help and will report back from my server, but are you saying you got 500mbit from my server with 20 connections? Im watching with nload and 15 sec mrtg graphs and there has been no usage over 70mbit.

    tks
    Yes that is correct. It is likely because I only ran it for about 30 seconds. Here I can let it run for about 5-8 minutes so it shows up on your graphs. Again I should not have to run 20 connections like that. Something is definitely affecting speed on an individual connection basis. Normally I would say saturation except there is no way I would be seeing 0.5ms latency to your server if there was saturation.

  20. #70
    Join Date
    Jan 2010
    Posts
    165
    wow please do run this for a few min, im watching.
    This could be resovled here if you are seeing 500mbit!!

  21. #71
    Join Date
    Feb 2011
    Posts
    680
    If I force things out of Level3 then it goes via nlayer again but the wget hangs. I get similar results going to your server.

    [root@ocean ~]# traceroute quadranet.com
    traceroute to quadranet.com (98.143.159.202), 30 hops max, 38 byte packets
    1 gige-0-1.core1.silvaco.com (192.73.228.1) 0.226 ms 0.108 ms 0.103 ms
    2 gige-v11.qts.silvaco.com (4.28.13.25) 0.467 ms 0.459 ms 2.182 ms
    3 vlan370.car1.SanJose1.Level3.net (4.28.12.225) 0.458 ms 0.483 ms 0.451 ms
    4 ae-4-90.edge2.SanJose3.Level3.net (4.69.152.209) 0.881 ms 0.905 ms 0.868 ms
    5 gblx-level3-20g.SanJose3.Level3.net (4.68.62.178) 0.997 ms 0.853 ms 0.863 ms
    6 ae9-40G.scr4.SNV2.gblx.net (67.17.72.13) 1.358 ms 1.514 ms ae9-20G.scr3.SNV2.gblx.net (67.16.145.117) 1.407 ms
    7 po5-40G.ar3.LAX2.gblx.net (67.16.141.170) 8.949 ms po4-40G.ar3.LAX2.gblx.net (67.16.141.162) 8.764 ms po5-40G.ar3.LAX2.gblx.net (67.16.141.170) 8.790 ms
    8 xe-3-0-3.ar1.lax1.us.nlayer.net (69.31.127.213) 14.064 ms 11.870 ms 11.964 ms
    9 ae0-60g.cr1.lax1.us.nlayer.net (69.31.127.137) 8.679 ms ae0-60g.cr1.lax1.us.nlayer.net (69.31.127.141) 8.592 ms ae0-60g.cr1.lax1.us.nlayer.net (69.31.127.137) 8.828 ms
    10 ae1-50g.ar1.lax2.us.nlayer.net (69.31.127.130) 15.317 ms 14.714 ms 11.971 ms
    11 as29761.xe-1-0-2.ar1.lax2.us.nlayer.net (69.31.121.254) 8.709 ms 8.747 ms 8.949 ms
    12 * *
    [root@ocean ~]# wget -O /dev/null http://quadranet.com/speedtests/100mb.test
    --09:31:14-- http://quadranet.com/speedtests/100mb.test
    => `/dev/null'
    Resolving quadranet.com... 98.143.159.202
    Connecting to quadranet.com|98.143.159.202|:80...

  22. #72
    Join Date
    Jan 2010
    Posts
    165
    WOW he is doing 500mbit. much appreciation i will edit the original topic to update.

    so i guess its just poor transit in the USA or all of my customers are on east coast. but should wholesale still be 3x faster when im pointing all users to either single server.

    To me this also proves it was not the server. Am i wrong to now say that quadranet is much slower than Wholesale for high bandwidth applications delivered to USA users?

  23. #73
    Join Date
    Mar 2008
    Location
    Los Angeles, CA
    Posts
    555
    Quote Originally Posted by jo2jo View Post
    WOW he is doing 500mbit. much appreciation i will edit the original topic to update.

    so i guess its just poor transit in the USA or all of my customers are on east coast. but should wholesale still be 3x faster when im pointing all users to either single server.

    To me this also proves it was not the server. Am i wrong to now say that quadranet is much slower than Wholesale for high bandwidth applications delivered to USA users?
    Ok, I stopped the transfers since you verified. The thing is though I am not going over transit to your machine like I mentioned earlier its peering:

    traceroute to 216.45.49.251 (216.45.49.251), 30 hops max, 46 byte packets
    1 ip-66-33-193-1 (66.33.193.1) 119.262 ms 18.164 ms *
    2 ip-66-33-201-113 (66.33.201.113) 0.441 ms 0.508 ms 0.606 ms
    3 peering.1wh.la.quadranet.com.any2ix.coresite.com (206.223.143.159) 0.733 ms 0.459 ms 0.424 ms
    4 96.44.180.98 (96.44.180.98) 0.970 ms 0.613 ms 0.561 ms
    5 216.45.49.251 (216.45.49.251) 1.778 ms 1.290 ms 1.751 ms


    Even then a single connection is only resulting in 5-7 megabytes/sec which is weird for such low latency and direct routing. The latency did go up a little bit when doing that 500mbit trasnfer though, possibly one of the links in a port channel got saturated along the way since it was from one single IP -> another single IP:

    Without the connection going latency was lower:

    Code:
    # traceroute -I 216.45.49.251
    traceroute to 216.45.49.251 (216.45.49.251), 30 hops max, 46 byte packets
     1  ip-66-33-193-1 (66.33.193.1)  31.977 ms  1.519 ms  0.452 ms
     2  ip-66-33-201-113 (66.33.201.113)  0.482 ms  0.302 ms  0.289 ms
     3  peering.1wh.la.quadranet.com.any2ix.coresite.com (206.223.143.159)  0.439 ms  0.462 ms  0.389 ms
     4  96.44.180.98 (96.44.180.98)  0.547 ms  2.063 ms  0.723 ms
     5  216.45.49.251 (216.45.49.251)  0.647 ms  0.781 ms  0.556 ms
    Here was 50 pings/sec to your server while doing 500mbits. It did lose a few packets. This is normal if there was a link being saturated along the way:

    Code:
    --- media7.ipadporn.com ping statistics ---
    300 packets transmitted, 296 received, 1% packet loss, time 8423ms
    rtt min/avg/max/mdev = 0.325/7.737/106.499/19.065 ms, pipe 4
    Here is with no transfer going:


    Code:
    --- media7.ipadporn.com ping statistics ---
    300 packets transmitted, 300 received, 0% packet loss, time 6434ms
    rtt min/avg/max/mdev = 0.301/4.705/104.495/15.494 ms, pipe 4

  24. #74
    Join Date
    Mar 2008
    Location
    Los Angeles, CA
    Posts
    555
    jo2jo can you traceroute my server (208.97.140.21). I assumed its using the peering both ways but I suppose its possible it is not (and only is me to you). It is using peering both ways to my friends server also at pacific rack so this is why I assumed it but its possible they are going over transit for some reason from your subnet.

  25. #75
    Join Date
    Sep 2006
    Location
    The Not So Deep South
    Posts
    931
    Quote Originally Posted by jo2jo View Post
    WOW he is doing 500mbit. much appreciation i will edit the original topic to update.

    so i guess its just poor transit in the USA or all of my customers are on east coast. but should wholesale still be 3x faster when im pointing all users to either single server.

    To me this also proves it was not the server. Am i wrong to now say that quadranet is much slower than Wholesale for high bandwidth applications delivered to USA users?
    Well from inside Quadranet I don't see it

    Code:
    [root@e3la03 vz]# wget -O /dev/null http://media7.ipadporn.com:1049/videos/1gb.bin
    --2011-10-09 12:31:07--  http://media7.ipadporn.com:1049/videos/1gb.bin
    Resolving media7.ipadporn.com... 216.45.49.251
    Connecting to media7.ipadporn.com|216.45.49.251|:1049... connected.
    HTTP request sent, awaiting response... 200 OK
    Length: 1073741824 (1.0G) [application/octet-stream]
    Saving to: `/dev/null'
    
    16% [=====>                                 ] 177,502,360 3.30M/s  eta 3m 20s
    [root@e3la03 vz]# wget -O /dev/null quadranet.com/speedtests/100mb.test
    --2011-10-09 12:31:50--  http://quadranet.com/speedtests/100mb.test
    Resolving quadranet.com... 98.143.159.202
    Connecting to quadranet.com|98.143.159.202|:80... connected.
    HTTP request sent, awaiting response... 200 OK
    Length: 105472000 (101M) [text/plain]
    Saving to: `/dev/null'
    
    100%[======================================>] 105,472,000 75.3M/s   in 1.3s
    
    2011-10-09 12:31:51 (75.3 MB/s) - `/dev/null' saved [105472000/105472000]
    
    [root@e3la03 vz]# traceroute media7.ipadporn.com
    traceroute to media7.ipadporn.com (216.45.49.251), 30 hops max, 40 byte packets
     1  escapefromla.hostigation.com (96.44.154.33)  0.094 ms  0.127 ms  0.122 ms
     2  hostg-quadra-gw.hostigation.com (72.11.150.105)  0.543 ms  0.601 ms  0.684 ms
     3  216.45.49.251 (216.45.49.251)  0.288 ms  0.292 ms  0.288 ms
    [root@e3la03 vz]#
    So I can pull from QuadraNet's test file at 600mbit but yours at 26mbit?
    Hostigation.com - High Resource Hosting
    WHM/cPanel Servers for Hosting and Dedicated Needs
    SolusVM VPS Hosting - Big Features, Small Prices
    Like us on Facebook or follow @hostigation on Twitter

Page 3 of 4 FirstFirst 1234 LastLast

Similar Threads

  1. Replies: 5
    Last Post: 05-11-2011, 01:02 AM
  2. Replies: 0
    Last Post: 05-01-2011, 05:19 PM
  3. Replies: 0
    Last Post: 04-18-2011, 06:55 PM
  4. QuadraNET launches Secure SSL VPN Passthru to Private Network!
    By ItsChrisG in forum Web Hosting Industry Announcements
    Replies: 0
    Last Post: 03-03-2011, 04:38 AM
  5. QuadraNet announces Asia Optimized Network Segment!
    By ItsChrisG in forum Web Hosting Industry Announcements
    Replies: 0
    Last Post: 08-25-2010, 03:08 AM

Posting Permissions

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