Transcript for this video
Hello there, are you working on your vacation?
Oh hi, glad to see you again. I'm just checking in with our house sitter.
Enjoying the cruise?
Very much, but I'm getting frustrated. The network on the ship is painfully slow.
It's slower than a cucumber doing calculus!
And for this I'm paying 40 cents a minute!
Is something wrong with the network?
The core of the problem is that it's difficult to string wires from shore out to a ship at sea.
So in the absence of wires we are connected to the heart of the internet by a radio link that's being bounced off a geosynchronous satellite.
Really? Geosynch satellites are about 22 thousand miles up!
26 thousand if we measure from the center of the earth.
From here, off the coast of Mexico, our network link goes 22 thousand miles up…then another 22 thousand miles down…and finally lands
you'll never guess where…
In New Jersey.
That's like going around the world twice.
And our data has to come back the same roundabout way!
This causes problems you won't see on your home or office network.
Because the office network uses regular land lines?
Most internet protocols require several rounds of packet exchanges.
When you read email or look at a web page there can be hundreds, even thousands of packet exchanges.
Packets do not move instantly. It takes time for them to be carried.
Generally the faster that packets move the more responsive the internet feels.
From your home or office packets travel on wires that follow the surface of the Earth.
Those packets reach their destinations fairly quickly.
But when a geosynch satellite is involved the distances become much greater.
Even at the speed of light it takes a really long time for packets to move up to the satellite and back down again.
A few minutes ago I measured the time it is taking for my packets to go from the ship to my email service and then back again…
It was more than half a second.
Half a second doesn't sound so bad.
Remember that most internet transactions take lots and lots of packet exchanges.
That means lots and lots of half-second delays.
I guess it adds up.
For instance, each time I fetch a web page it takes perhaps a dozen such exchanges just to figure out the address of the web server and to get the data moving.
That means I wait at least five seconds before I see anything happen.
To make it worse many websites lard-up their content with hidden trackers, images, and scripts.
So that what we see as a single web page may actually require fifty or more inner web-page fetches.
All those delays could easily accumulate into minutes!
Which leaves me sitting here impatiently waiting.
But it does leave me plenty of time to talk…
… and drink.
Can we do anything about these delays?
I wish we could crank packets up to Warp factor nine or something…
Unfortunately Einstein told us that the speed of light isn't just a good idea, it's the law.
So it's hopeless?
A professor of mine used to say “If you can't get ‘em on the merry go round you can get ‘em on the swings.”
In other words, rather than trying to go faster than the speed of light, we can take another tack.
We can work the problem by reducing the number of packet exchanges that we do.
So let's figure out how to get the job done using the minimum number of trips over that satellite.
We can use a technique called “local caching”.
Oh, I've heard of that!
These are fancy words for saying that when somebody on the ship does a transaction the network on the ship tries to preserve the result for a while.
That way, if someone else on the ship does the same transaction we can use the saved result rather than going over the satellite again.
Did I get that right?
Yes. Caching is often very effective.
There is a lot of repetition on the net…
for instance, just think how often different people look at the main Google search page.
There are two things that could be cached to provide a big win for users:
Domain Name look-ups and pieces of web pages.
Let's take a trip to my lab so I can show this to you in action…
This might be a bit rough, so hang on!
Begin cheesy wiggly cross-fade to next scene.
What happened to my shoes?
Oops .. the cross fade still has a few bugs.
Don't worry, they'll eventually show up.
In the meantime let me show you around.
Our company, InterWorking Labs, builds a couple of products.
We call them “Maxwell” and “Mini Maxwell”.
These are used by network developers and providers to create real-world network conditions in their lab.
A cruise-ship company could use Maxwell to try various caching approaches to see how much benefit they might obtain…
and how much happier their customers might be.
Here, let me show you…
Here is the Internet.
Here's the satellite link to the ship.
And here is where we were, on the ship.
We could try to rent a ship and satellite for experimentation.
That would be terribly expensive!
And even if we could afford it, real ships and satellites aren't easy to adjust to create different network conditions.
Are there better ways?
This is a Maxwell box.
As you can see, the Maxwell takes the place of the satellite link.
And this mimics my laptop on-board the ship.
We can adjust the Maxwell so that it makes the emulated network here on this table seem just like the satellite network used by the ship.
Let me show you.
I'll add some simulated satellite delay while pinging our website.
Here's a normal non-satellite situation.
So let's dial in a typical satellite up-down delay of about 250 milliseconds each way.
Notice how the ping times increase?
As I mentioned before, a solution to our satellite delay problem is to reorganize things to eliminate some of the transactions we send over the satellite.
So let me add some equipment.
OK, I've just added a local domain name resolver and a web cache onto the simulated ship and configured my laptop to use them.
Let's see the difference…
Both of these web browsers are going through the same satellite delay of quarter of a second each way.
The left side uses a local, on-ship, web-cache.
The right side is un-cached, like on the ship.
You get the picture…or not.
With Maxwell we were able to reproduce the shipboard experience in our lab.
That let us experiment with solutions to the satellite delay problem.
And as you can see, we found a couple of easy and inexpensive solutions that really do help.
Now that we have some answers let's go back to the ship. Hang on!
Looks like we got back in one piece.
I hope our little trip suggests that we need not suffer in paradise, at least not network-wise.
A land-line will always be faster than a satellite.
But as you have seen there are tools that could be used by equipment vendors and by the cruise ship companies to investigate solutions that would improve things.
Where can I learn more about Maxwell and InterWorking Labs?
Come visit our website at IWL.com
Want to Know More?
|Find more videos||How to buy||Contact an Expert|