For some reason the fact that intranets can significantly outperform Internets and in particular THE Internet is a shocking concept. It shouldn't be.
An intranet is a contained network managed to optimize the performance to meet the needs of the participants on the network. Your corporate network in your local office is an intranet. You probably have IT people working hard to make sure its up and as fast as they can make it.
Your home network is an intranet. Wired or wireless, you are probably able to get throughput that far exceeds the speed you get from your Internet connection.
The INTERNET on the other hand is owned by no one, optimized for no one. Whoever you buy your Internet connectivity from, particularly if you are a broadband customer, has a network that is physically, locally accessible through a persistent connection (as opposed to dialing in to a remote network). They do everything they possibly can to make it as fast as possible. Your speed and throughput is dependent on a variety of variables, the most important of which is the type of wire/fiber that connects your house to the network.
Their ability to control the quality of service and throughput you receive to your home disappears the minute your traffic leaves their network and is passed of ff on to the route that will take or request your bits to or from their destination.
Think of it as crossing the border between the USA and Mexico in a car. . Neither side cares about how fast the traffic of the other side gets through. If its busy, its busy. If its slow, its slow. Its not their primary concern.
This is exactly why your video buffers , websites timeout and downloads take forever. No one is responsible for making sure its fast.
The times you get the best performance are when the website or video providers put servers on your providers network. Its for this reason that content providers pay Content Distribution Networks (CDNs) like Akamai a premium to distribute content. Akamai has thousands of servers all across the net.
CDNs however are limited in the applications they can run. They are optimized to deliver websites and graphics and downloads. The same things they have been doing for 10 years . They are not a development platform
Which is exactly why the best years of the Internet are behind it. Before you get all mad at me, ask yourself this. What is the increase in your broadband speed THROUGHPUT over the past 1,2,5 or 10 years ? I went from about 200k in 1997 to about 800k (on a 3mbs advertised number) today. An increase of only 600k. True, its cheaper now than then, but 600k is only 600k. Most people think the throughput TO THE INTERNET of our home broadband connections will increase significantly over the next few years. It wont. Other than dropping fiber in the last mile, not much has changed in the last few years and unless you have fiber to your home, not much will change in the next 5 or more years. Face it folks, the Internet as a platform has stagnated. Its dead as a growth platform. Its like Microsoft windows. From about 1985 to 1995 it was a great platform and there was new software coming out continuously. When was the last time you got excited about a new piece of consumer software written for Windows ? Its a stagnant consumer platform. We switched to browsers for most of our PC activity. We are getting to the point where the browser on the net as a platform is becoming stagnant.
Now ask yourself what the maximum possible throughput of your Internet connection ? You probably are connected to a 100mbs or faster port on the other side of your "last mile".
The typical provider throttles you down from the maximum not because they cant support more speed on their network, but because they cant deliver more speed on to the INTERNET. If you get your Internet access from the same provider that you get digital TV from , that provider is already providing you more than 1gbs of throughput of service. Yep, there are 10s of millions of people who get more than 1gbs of throughput of traffic to their home. We just call it digital TV. All those channels that you can flip to take up a huge amount of bandwidth in aggregate. The limit on the amount of bandwidth they give you for Internet is not a physical limit , its a limit based on software, technology and business decisions.
So I asked myself, "Self, could that software change so that a new platform for applications that are built on 25mbs, 50mbs, 100mbs or even 1gbs are possible ? " The answer is yes.
Software is emerging that allows applications to be written that are optimized for very high speed. But those applications can only leverage high speeds on the broadband providers INTRANET. There is no network provider on the planet that can guarantee 100mbs throughput to some random website somewhere in the world. ' Put that website on the same physical network that you buy your Internet service from and in the next couple years your provider will be able to guarantee quality of service of 100mbs
Ive sat with several of these network providers and what I'm telling them, and I think they are listening and following through, is to offer a platform or intranet applications. A platform for applications that confirm that the user and application host are on the same network, or possibly even on the same network segment. Make 100mbs or higher throughput a guaranteed service level to that application.
When that happens, people a lot smarter than me will come up with applications that blow away anything we are seeing now. I dint care if you call it Web 10.0 or whatever, but the reality is that the applications we will see then will be amazing.
When it happens, we will look back at Internet applications and laugh. Kids will call you out with things like
"did websites really time out ? we video really buffer and die even though it was limited in size and quality"
IntraNets vs InterNets
Recent Posts
- Beating Google ? (5/14/2008)
Reader Comments
(Page 1)2. Too much Youtube nonsense taking up too much bandwidth. Growth of file sharing -- bandwidth eating sources -- clearly growing at at a better clip than the technology supporting it
Posted at 4:30PM on Jul 27th 2007 by David
3. Isn't this saying that CDNs just need to get more local? If I buy bandwidth from Akamai, and they've got a "local" account with my cable provider, that data comes in on my last mile connection. That makes it fast without needing a faster connection to the Internet at large. It means that large sites like Google and Yahoo will be very fast (using CDNs) but my little blog will be slow as hell.
Also, the CDNs need to get smarter. Instead of deploying data they need to be application deployment platforms. If they all ran some sort of application server, and you distributed the actual application across the network, you'd effectively get the same result.
I wouldn't say the Internet is dead, I'd say that the monolithic approach to website development is. Your application needs to be as distributed as your users are.
PS - The bottom of your site has the copyright as 2006. You should probably update that.
4. Mark,
What makes the Internet great is who's connected to it, not the actual speed of the network. Sure, the speed makes it more enjoyable, but most of us choose to leave the walled garden (AOL) years ago.
As for local services that will be enjoyable with guaranteed QOS, they will continue to offer services (yes I'm talking about VOD) but there will always be the Internet for entrepreneurs who are smarter -- or just more agile -- than big cable.
As for the last mile, most coax networks have way more than 1Gbps. Most modern cable networks are 900Mhz so; 900Mhz/6=150 QAM256 channels, each capable of 38Mbps or 150*38=5700/1024= 5.56Gbps (minus all the analog channels of course).
Posted at 4:57PM on Jul 27th 2007 by Ben Drawbaugh
5. There is no network provider on the planet that can guarantee 100mbs throughput to some random website somewhere in the world.
Could I email this quote to all of our customers? Because they don't believe me when I say it, maybe they'll listen to you. ;)
Posted at 4:59PM on Jul 27th 2007 by Jonathan Brashear
6. i'm not a tech. someone please offer clarification. are you proposing the mirroring of the most popular aspects of the internet within walled gardens to guarantee high throughput and ideal network capacity for currently difficult-to-implement network applications? Supernodes, as it were?
thanks.
Posted at 5:11PM on Jul 27th 2007 by blyx
7. With digital storage costs falling rapidly, how feasible is it that each ISP or, dare I say, each home, cache the Internet each day (each hour) and make the entire Internet an Intranet? I think this is what you're getting at Mark. ? When are we going to see this happen?
Because I've gotta tell you - I'd much rather see comcast put their resources into this (grabbing the Internet) than another cable channel that i could care less about. They probably think they've got this long tail thing figured out with 500 channels. How about a couple billion websites? I'd pay more for what you're talking about than I would for this cable package I've got. And that's a lot.
Maybe comcast should buy akamai - as a start.
Thanks for the thinking. -Byron
Posted at 5:58PM on Jul 27th 2007 by Byron Warnken
8. I like the way you are thinking. But what do you think about WiMax? See Sprint press release from yesterday - http://www2.sprint.com/mr/news_dtl.do?id=17560. If they get long range wireless up to speed someday as technology advances, that may be an alternative solution for all of us. That way we are not dependent on the type of wire/fiber that we use.
Posted at 6:46PM on Jul 27th 2007 by Brian
9. Well, first of all, the whole point of Internet2 -- now a research project, just like the Internet was at the beginning -- is to be a lot faster. So that may come.
But that may also be years. That's why so many content providers are talking about network neutrality -- they want to have the ability to slow down certain activities in order to speed up others. The objection is: how do you know which is the unimportant packet to slow down, and which is the important one to speed up?
You suggest an interesting twist: instead of speeding up some packets and slowing down other packets, ISPs should locally provide a variety of services at a very high rate of speed. using their own network. "Comcast: now with accelerated YouTube!" "Verizon: now with accelerated Google Docs & Spreadsheets!"
Not a bad idea, but if they're going to host YouTube locally, why not just have the YouTube channel on cable or the telco's triple play service? Isn't the step beyond locally hosting a service specialized transmission, lacking the HTTP overhead, like, say... cable tv? What's the difference?
Posted at 8:03PM on Jul 27th 2007 by Wade Armstrong
10. What if someone has built a broadband accelerator... that allows any subscriber to surf at "line speed" 95% of maximum capacity all the time, regardless of other network slow downs?
Posted at 10:37PM on Jul 27th 2007 by Morgan Warstler
11. This article demonstrates that the primary limitation of increasing home bandwidth is technical whereas it's pure business strategy.
They are NOT trying as hard as they can to provide sppeedy connections in the US. Case in point the 12 to 15 Mbits most folks can get here in Europe (I have direct experience in Germany, France and the Netherlands) with DSL and much higher in Japan.
Posted at 11:43PM on Jul 27th 2007 by Alles Klar
12. Alles Klar:
You are right. Europe, Japan and Korea all have much faster speeds than we do. Much cheaper too. Of course it helps that Japan and Korea have much higher population density.
Posted at 3:06AM on Jul 28th 2007 by Ghost of Tom Joad
13. Where do you get your 800k number? In 1997 I had something like 1mbps cable. I now have 20mbps (5 up) fiber and I can saturate it with many services to get very close to my 20480k max. Sure a single web page doesn't saturate it, but I can download an hour of high def video from usenet (provider not in my ISPs network) in 4 or 5 minutes.
14. No Sir, that's not correct. Generally, your assumptions are incorrect. The throughput obsticles are not caused by the issues that you mentioned, such as bandwidth. Also, the "Internet" is potentially on the verge of a huge innovation, and it's just around the corner.
If your consultants understood the issues of realtime, multiprocessing systems then they would understand that the main problems with end-to-end throughput to a user workstation, over the Internet in the current tech. world, is with the user workstation's operating system. This has been shown many times over, in realtime systems involved with spaceflight systems environment. The throughput bottleneck is in the data handling at the user's end. With most people using Windows O/S, or anything similar, it's the non-realtime nature over TCP/IP "ack windows" or other flow control mechanisms that cause an inherent slowdown, and not the routers or the "Internet Cloud" in between.
With the increase in multiprocessor available, as well as price affordability, processing at the user workstation will be more efficient. The old WIndows O/S with no-thread coding gave rise to the multithreading Windows O/S, which will give rise to a multiprocessing O/S. Remember, WIndows is saddled with "backward compatibility" issues and is basically a big cludge. But, there will arise many new O/S that handles these new multiprocessor CPU (like 8, 16, 32, etc. processors on a chip) well.
The O/S software must now catch up with the multiprocessor hardware. Currently, there is no good "big name" O/S around that is meant for the Internet Surfer. Issues like deadlock and multiprocess allocation are hard to code, towards a general purpose "surfer". It might take a "innovator" company like APPLE to eventually design a good Internet Surfer O/S with a good multiprocessor "design" to place inside it's iTV (ie. the set-top box) or maybe they will come out with a Killer-App tablet with super-functionality. This stuff will only be possible with revamped software (ie. O/S) and multiprocessor hardware. But, you never know when the breakthrough will occur. Mark, when you handed YAHOO the golden egg, they dropped it.
Your "HD" solution is to design, produce, and distribute a highly efficient O/S which is put in a set-top box or tablet, which is meant to effectively handle the higher data rates, as well as utilizing an innovative user interface, to revolutionize the Internet as the new KILLER APP.
Posted at 12:25PM on Jul 28th 2007 by www.freeway2000.com
15. Served apps by ISPs to their clients have already been attempted and failed. MPowered being one. Sure it was a great idea and ahead of it's time, but it will never fly. You need a critical mass of developers to build those apps and you won't get it. You will be relegated to offering the same apps as what is available on the plain jane Internet which again doesn't cut it.
I submit that it is the cable companies that are dead. In 5 years everyone will be downloading their multi-media and apps-on-demand; there will be no need for cable feed movies, TV or Phone. You will need this subscription revenue as most ad revenue will be diluted by digital recorders and the coming auto-elimination of advertisements. In five years we WILL be able to download HD content, so relax.
Five years is a heck of a long time Mark and you know it. Advancements will be made that are going to knock your socks off and YOU know it. HDNET has its place though - you are fine; no need to FUD the market just yet bud.
17. Great post Mark! It would be great to see the ideas you have as well as from some of the people who have commented work soon. If providers shared resources, would this help?
Posted at 10:57PM on Jul 28th 2007 by Chris Dowell
18. I actually agree with some components of net neutrality (such as crippling competitor traffic and such), but in the end it's really a business decision. Net neutrality will be an afterthought once broadband becomes ubiquitous and transparent. At that point, consumers will not care about which provider has the best content peering arrangements since bandwidth will be an afterthought.
Posted at 6:52AM on Jul 29th 2007 by Brian Wohlgemuth
19. Japan/Sweden have what you're talking about. Both intraconnect at vastly superior speeds but slow to a relative crawl when they get to the great wide internet.
Why haven't we seen any of these amazing developments you project from Japan?
For the most part Japan's & Sweden's bandwidth is used for media on demand. Not exactly an amazing technological breakthrough.
Intranets are wonderful for sharing large amounts of scientific data between labs but for the masses they're not worth the effort.
Give the public a 4Mbps connection and they'll stream 1 DVD channel, give them a 40Mbps connection and they'll stream 1 HD channel. The increased investment at the last mile just isn't worth it, especially with carriers beginning to have viable wireless options at the last mile.
Posted at 6:54AM on Jul 29th 2007 by Adam
20. I think both intranets and the internet serve their purposes, and their purposes are different. Intranets are more specific, and as you say - catered to their users.
The internet, the "information superhighway" (haven't heard that term in years), is about the free distribution of information. The open-endedness of it, is what makes it different. Because it's designed for everyone, it is slower as you say, but isn't restricted to just the specifics of the intranet.
Posted at 1:52PM on Jul 29th 2007 by Adam S.
1. This is what you get with net neutrality -- equally slow for everyone.
Posted at 4:28PM on Jul 27th 2007 by Josh Bernoff