Have you ever noticed how much easier it is to buy new hardware than it is to actually get it working? How a small project often turns into a large one?

I recently bought a new wireless router, Apple's Airport Extreme, for our home network. Since it was expensive, I wanted to make sure that I was really getting my money's worth. In order to do that, I did some measurements of my existing network and then the new one. The process took a lot more time than I expected (always does) but the process was informative.

I will post a review of the Airport Express in another post but I wanted to start with a post on home networking.

Even though I am a techie, I was surprised to learn how much I was personally affected by advertising. My assumption when I started this process was that I needed a faster network. What I learned in the process is that a faster network WITHIN my house wont do me much good whereas a faster network TO my house would be huge.

Which leads us to the first topic. How we measure networks.

bits and bytes

Most of us are familiar with the basic measurements of storage, either memory or hard drives. The capacity of these devices is measured in bytes. Bytes, KiloBytes, MegaBytes, GigaBytes, TeraBytes. In each case, 1024 units gets you to the next level. 1024 bytes is a kilobyte. 1024 kilobytes is a megabyte, etc.

Networks however are measured in bits not bytes. A bit is a single digit (1 or 0) of information and there are 8 bits in a byte. So a million bits is 1/8 as large as a million bytes. Moreover, networks measure a rate so their units involve time, specifically bits per second. Bits per second is what we call bandwidth and the standard unit is megabits per second or mbps.

This bits/bytes thing is where the marketing speak got me. After a million commercials that talk about megabits (and the Slowsky's), I had (wrongly) assumed that Internet speeds were faster than they actually are. 8 megabits per second (mbps) are only 1 megabyte per second. More importantly there arent any Internet connections around my house that go that fast anyway!

Another thing to note about network speeds is that manufacturers use decimal math instead of the binary math used in storage. A kilobit per second means 1,000 bits per second not 1,024 bits per second. This makes a noticeably difference when we talk about mega or gigabits but mostly it is just annoying.

To make things even more confusing, many application that show you your network speed in real time use kilobytes per second or KB/s. *phew*

a typical network

Our household is probably typical of a lot of homes. We have networked computers in two rooms: the office and the living room. In the office we have a Windows PC and two Apple Macs along with two printers, an external hard drive for backups and a cordless phone. In the living room, we have a home theater PC connected to the TV and an Xbox. That makes 5 devices that need to be networked. Some can use an ethernet cable but others needs a wireless solution.

The reason I mention the cordless phone is that I learned all these devices, as well as bluetooth toys like keyboards and mice, share the same 2.4 Mhz frequency. That means they can (and do) interfere with each other, causing slowdowns or failures. And dont forget your neighbors - any devices of theirs that are close will also have an impact.

the many network connections in our home and the 1.5 mbps connection to the Internet

a series of tubes

When most of us think about our "network speed" we are really thinking about the speed of our connection to the Internet. We want to download music or movies or files or web pages into our house.

The important thing to note here is that home computers actually have two networks: the network within your house (called a LAN or local area network) that connects your computers to each other and the connection of your household to the Internet network.

Although I have heard the numbers for a long time, I never really thought much about the difference between the speed of those two networks but the difference is huge. Take a look at these two speed comparison charts to see what I mean.

Compare the blue LAN speeds with DSL (orange) and cable (green) speeds

It makes a big difference whether you are talking KB/s (top) or kbps (bottom)

Internet connections to homes today, so-called high speed broadband, come in two flavors: DSL or Cable. These speeds top out at 6 megabits per second for a cable modem while DSL speeds are generally slower.

The important thing to notice here is that those Internet connection speeds are a FRACTION of the speed within your LAN. Wired ethernet cables transmit at 100 mbps and some now do a gigabit per second. Even wireless speeds or 11, 52, or 90 mbps are massively larger than that Internet connection speed.

In other words, if you buy a faster router for your home thinking it will speed up your access to the Internet, forget about it. Whatever you already have now is much faster than the real bottleneck - your connection to the Internet.

This observation is something I should have known about before but I will admit I hadnt really thought about it. I was expecting a faster "network" and was rather disappointed to realize that what I really want to speed up is my Internet connection. If you are thinking about digital media applications like watching movies over the Internet, you need to think about when someone can put fiber optic connection into your house because current speeds are a big problem.

So how big is my pipe?

If you have a 1.5 mbps DSL line like I do, what does that really mean?

Have you ever downloaded a large file and found your webpages slow down and your Internet radio stops? This is because our networks have a limited bandwidth - they really are kind of like a pipe.

If you listen to Internet radio like I do, different feeds have different speeds listed in kbps. A 128 kbps connection uses up about 9% of your total bandwidth. You could listen to ten stations at once or listen to one station and have less bandwidth left over for webpages. Uploading a file, such as a bittorrent, will also take up some of your bandwidth.

Another measure of your Internet connection is download speeds. If you download a large file with a browser and look at the download speeds, what is a good speed? I typically see speeds between 20KB/s and 140 KB/s. That means my download is using roughly 20% to 80% of my total bandwidth.

But the really interesting discussion comes when we talk about downloading video.

In this table, I took the file sizes of an hour of HDTV recording and a typical DVD. Then I did some simple math to determine how many bits per second one would need in order to watch in real-time. The results show a) why we dont have streaming HDTV over the Internet yet, b) why video services involve a long download first, c) and why popular video sites like YouTube use very small video sizes and low quality. Moving video takes a lot of bandwidth.

How long does it take to do a typical download? Again I used some simple math to look at the time it would take different connection speeds to download 400MB file and a 4GB DVD. Get ready to wait.

What we see from these tests is that it is quite manageable to move DVD's ad HDTV around within a house at LAN speeds but that Internet connection just isn't ready yet. (Over six hour to download a DVD with 1.5 mbps DSL!) And recently there have been several news articles by large network providers complaining that even they dont have the speed to serve video to clients.

reality checks

Look at those speed comparisons again. Note that they are for theoretical maximums, ie marketing-speak. The reality of how fast your network gear operates can (and does) vary substantially from the advertised speed.

You can run a DSL test from a number of websites. Doing so, I found that my DSL operates fairly close to the 1.5 mbps or about 8% less than advertised.

On the other hand, the actual speeds on my LAN were astonishingly slower than the advertised speeds.

To test my network LAN speeds, I moved a 138 MB file between two computers over the network. Some of the test were between computers in the office and others tested the connection between a computer in the living room and one in the office. Overall my network had 5 variations to test:

1. both computers have wired 100 mbps ethernet
2. one computer has 100 mbps ethernet and the other has wireless 52 mbps - both in the office
3. both computers have wireless 52 mbps - both in the office
4. one computer has 100 mbps ethernet and the other has wireless 52 mbps - the office and living room
5. both computers have wireless 52 mbps - the office and living room

Granted, this is not a scientific experiment and there was significant variation when repeating the same test. But after spending a few tedious hours moving that 138MB file around, I feel comfortable that these results are indicative of my reality. And my reality is mighty slow.

Ouch! My network does not perform as advertised (or assumed)

Actual speeds on my network are several hundred percent slower than advertised speeds

Only the wired ethernet connections were even close to the advertised speed. The purely wireless connections were very slow and having one of the computers on a wire almost doubled the performance of file transfers.

The important thing to remember though is that even my slowest speeds were 4 or 5 times faster than my Internet connection. Thus the Internet operations I do most of the time are unaffected by my LAN speeds. (Which probably explains why I never noticed how slow my network actually was.)

Configuration

Before closing I should note that when I started this process, I did not realize how many features one could/should configure in their network. Having configured many networks before, I knew about encryption and security issues and I knew how to get things working. What I did not know about was the many settings that affect network speed or interference with other devices.

With my original router, I could choose the frequency channel to use as well as whether to use a mixed 802.11b and 802.11g network or specify a single type. With the Airport Express had similar decisions to make and it was clear that using a single router to handle both 802.11g and 802.11n networks would cause the 11n performance to suffer. This MacWorld article confirms as much (and shows that someone can get amazing performance).

Lessons

While this little upgrade morphed into a fairly large weekend project, it was informative. I learned a little about network structure and I learned a lot about network performance and Internet access speeds. (I also learned that my USB 1.0 hub was a major bottleneck.) I will be keeping these lessons in mind as I think about digital media applications in the future. (And as I shop for a faster Internet pipe :)