Optimizing Small Network Speeds

[Batspiderman17]

Proud to announce my first post! I'm relatively new to building networks and I was hoping to get a little guidance from the experienced users on this forum. Thank you in advance for your help. My setup is below:

Router:

• Belkin N600 DB (f9k1102 v1) (dual stream, dual band, with 100 Mbps ports)
• Security: WPA2 (AES)
• Wireless channel set to auto on both bands
• Extension channel set to 4 on 2.4 GHz band and set to 40 on 5 GHz band
• 20/40 MHz: set to Auto on both 2.4 and 5 GHz bands
• Stock firmware 1.00.10

Desktop:

• HP Media Center m7167c
• Windows XP Media Edition 2002 SP3
• 10/100 Ethernet wired to router
• 2 external USB 2.0 (High-speed) HDDs (permanently shared to network via mountvol)
• Norton Internet Security 2011

Laptop:

• Sony VAIO FW-490
• Windows Vista Home Premium SP2
• Intel Wifi Link 5100 AGN (dual band, 1x2 stream)
• With my laptop ~5 feet away from the router, Windows states that I have 5 bars and 150 Mbps connection.
• Norton Internet Security 2011

I turned off all other devices using the router (android phones, other laptops).

When I run LanSpeedTest v3.4 on my laptop and connect to a shared folder on my external HDD connected to my desktop, my upload/download is ~35 Mbps for both on the 2.4 GHz band and ~45 Mbps upload and ~22 Mbps download on the 5 GHz band.

My initial reaction is that this seems low. After looking into bottlenecks, I believe my desktop's ethernet card/Router's is the weak link. While I'm not naive to expect 100 Mbps, I was expecting more than ~33% of the theoretical bandwidth.

My questions:

1. Is this expected behavior?
2. How can I optimize any additional settings on my network?
   • Router settings: Perhaps selecting a more appropriate channel and extension channel? (I've attached jpegs of inSSIDer's wifi analysis. I would appreciate any advice. My SSIDs are the Belkin_TL_N, Belkin_TL_G, and Belkin_TL_N.)
   • Desktop/Laptop Settings: Perhaps configuring Norton better? (I assume it was since I was able to access the shared folders, but are there ports I should be opening?)
3. If this is the maximum realistic thoroughput I can expect with a bottleneck of 100 Mbps from the desktop ethernet card, is it safe to assume that I can't improve on this bottleneck without getting a Gigabit PCI card AND also getting a new router that has Gigabit ports?
4. That said, as my Wifi card is 1x2 stream, could I expect some improved thoroughput if I upgrade to an Intel Wifi Link 5300 which would enable me to fully utilize the dual streams of my router?
5. Finally, am I supposed to see an appreciable improvement in thoroughput by disabling the Guest Wifi, thereby reducing interference???

I know this is a lot but I really appreciate any help/advice. Thank you in advance!

 

[thiggins]

Run your test with all Ethernet connections to see if Ethernet is a bottleneck.

Use the maximum # of packets to be sure you are getting as accurate a measurement as possible.

You should also try timing a large file drag and drop to get another measurement point.

The Guest Wi-Fi network doesn't use additional bandwidth unless it is in use. It won't reduce interference because it is on the same channel as your main SSID.

 

[Batspiderman17]

Thank you very much your advice! And you were absolutely right, the ethernet was not my bottleneck...wifi throughput is much worse than I expected.

On Lan Speed Test, keeping the default 1 MB packet size and maxing out at 1000 packets, I noticed that the numbers changed a tad:

• 2.4 GHz: ~33 Mbps upload / ~32 Mbps download
• 5 GHz: ~47 Mbps upload / ~25 Mbps download
• LAN (wired): ~80 Mbps upload / ~60 Mbps download
• When I transferred a 1 GB file back and then forth, I noticed that there was an increase of about ~5 Mbps in either direction (possibly due to my timing the transfer).

So after all this, my understanding is that it is not the 100 Mbps ethernet ports on my desktop/router that is the bottleneck but rather the wireless connection despite it being rated for 300 Mbps per radio. I stumbled upon your article on this website (How To Buy a Wireless Router: 2012 Edition), and I was astounded to see find out how poorly real-life performance correlates with theoretical throughput specifically on Page 2, Table 3, your table with expected throughput based on the router's max throughput.

So for my Belkin N600, I should be expecting 60-70 Mbps on 20 MHz and 70-80 Mbps on 40 MHz. But the catch in my case is that my network adapter is a Wifi Link 5100, which is not a full 2x2 stream but rather 1x2 stream (1 transmit and 2 receive as per Intel's site). So, is it right to assume that my typical throughput should be higher than for 1x1 but lower than 2x2?

If so, then the following statements should be correct based on Table 3 of the article:

• 1x1 wireless n card has a theoretical max 150 Mbps, with a typical rate of 30-40 Mbps at 20 MHz and a typical rate of 50-60 Mbps at 40 MHz.
• 2x2 wireless n card has a theoretical max 300 Mbps, with a typical rate of 60-70 Mbps at 20 MHz and a typical rate of 70-80 Mbps at 40 MHz.

My setup compares as such:

• 2.4 GHz band: both measured upload and download rates are within 30-40 Mbps
  • This is consistent with a 1x1 wireless n card at 20 MHz. Because I live in a low-rise condo/apartment setup with a number of interfering APs, I can assume that my router is defaulting to 20 MHz. So, no problems here.
• 5 GHz band: there was a marked disparity between measured upload and download with measured upload being ~47 Mbps and measured download being ~25 Mbps. This makes little sense to me.
  • First off, there are no other 5 GHz bands in my building and I don't have a cordless phone utilizing this spectrum. Therefore, my router should be operating at 40 MHz.
  • Since the Intel 5100 specs say that I only have 1 transmit antenna, then I should only be able to upload at a theoretical max of 150 Mbps (which translates to a real-world 50-60 Mbps at 40 MHz). Clocking in at ~47 Mbps, this is pretty much on par.
  • But the real issue is that my download is only ~25 Mbps. Since the network adapter has 2 transmit streams, then I should be able to upload at a theoretical max of 300 Mbps (which would translate to a real-world 70-80 Mbps at 40 MHz). Therefore, my download throughput on the 5 GHz band is 2-3x less than it should be.

So, is there any explanation (software or hardware) for why diminished download throughput on the 5 GHz might be occurring?

Again, thank you in advance for your help!

 

[stevech]

on both ethernet and WiFi, if you run 2+ large file transfers simultaneously, you usually see that the sum of the two's throughput is much better than one (flow). In WiFi, it's because the 802.11 medium is half-duplex, i.e., an acknowledgement (ACK) for prior data frames cannot flow concurrently with new frames. On Ethernet, assuming yours is able to do full duplex (you have dig in details to find out), 2+ flows can be faster due to the overhead in TCP and file system I/O.

The raw 802.11 speed is best measured with a tool that does not use the Windows file system. There are some free ones. But it's kind of moot what the raw rate is... it applies only to file-less transfers like video streams. A realistic expectation for the net yield at IP layer WiFi one-stream is about 60-70% of the indicated bit rate on the WiFi connection. About 15% of the ideal is lost just in TCP overhead; more is due to the file system and file sharing protocols.

 

[Batspiderman17]

Thank you for your reply! I've been learning so much about networking these past few days and there is quite a wealth of information on this website alone! As for the duplex mode, I have the desktop's card set to 100-Full Duplex and Belkin support stated that the router will be on 100-Full Duplex for wired connections if the wired ethernet card is also set to it. So I should be good there.

However, I'm still scratching my head over why my download speeds (~25 Mbps) are much slower than my upload speeds (~50 Mbps) on my 5 GHz dual-stream band when I also have a 1x2 stream wireless card (Intel Wifi Link 5100 AGN)???

 

[Batspiderman17]

Sorry stevech, it took a while to get it but it suddenly hit me how I could more objectively test what you were stating, how two simultaneous flows could more accurately benchmark my network.

On LAN Speed Test configuration options, instead of running the streams successively, I ran them simultaneously. A message popped up saying that the program supports up to 2 simultaneous streams when testing a folder. So, I'm capped to 2 simultaneous streams. Regardless, when I ran the test with 100 MB packets. My results:

• 2.4 GHz band: ~55 Mbps upload and ~55 Mbps download
  • Well, I'm not sure how accurate this can be because after reading more, I have set the router to operate at only 20 MHz on the 2.4 GHz band because I live in a condo with many interfering 2.4 signals. So, if this were true, I should only be expecting 30-40 Mbps - but not complaining
• 5 GHz band: ~90 Mbps upload and ~53 Mbps download
  • Again, this band never ceases to confuse me. This one I still do have set to operate at 40 MHz because there are no other 5 GHz bands in the area visible to inSSIDer.
  • Upload: my wireless card (Intel Wifi Link 5100 AGN) supposedly only has one transmit stream, so my theoretical max on 5 GHz band at 40 MHz should be 150 Mbps with a typical rate of ~50-60 Mbps (as per the aforementioned article). Yet somehow I'm achieving upload rates of near 90 Mbps?!?!
  • Download: my wireless card supposedly has 2 receive streams, so my theoretical max on 5 GHz at 40 MHz should be 300 Mbps with a typical rate of 70-80 Mbps. Yet, I am underperforming at only about 50 Mbps...

Now, I'm not sure if I'm totally off base here, but it almost seems like my card has 2 transmit streams and only 1 receive streams (though this would be the exact opposite of what Intel documentation states). Can this at all be possible?

I make a big deal about all this because I'm trying to optimize download speeds so I can stream movies from my networked desktop onto my laptop - dedicated NAS not an option yet