Thursday, July 09, 2009

Google Needs Lafayette

“Give me a place to stand and a lever long enough and I will move the world”

...Archimedes, 220 BC

Google needs Lafayette, and Amsterdam and Vasteras and....any of the fibered-up cities you might care to name. And, of course, Lafayette needs Google. That's been true for some time. But it recently became much clearer. The big news on the internets these last few days has been Google's newly announced Google Chrome OS. Most of the coverage has been predictable and mediocre but more thoughtfully analytical stories have finally begun to appear. (cf. the NYTimes) Even in the better articles the focus is inevitably on Google vs. Microsoft. While that might be understandable given that a battle between the two has become a journalistic stock-in-trade that is used to "explain" every move that either makes it really doesn't seem like the best analytic starting point for understanding what is going on. The fact that Google's OS isn't good for Microsoft is incidental to what Google—and a few other web players—are trying to do aid an ongoing process. Exactly what that process is requires a little explaining:

What's Going On Anyway? The backstory

The world is shifting yet again; this time onto the web from the computer. Not so long ago we moved much of our activity onto the computer —be they mainframes, PDAs, desktops, or laptops. The world shifted from only having physical objects that were unique or functionally identical copies of the unique object (think newspapers) to having perfect digital copies that paradoxically almost infinitely changeable, copyable, and decomposable (think email). The myriad internets focused on finding other computers and on transferring files between them. Mostly you worked on files locally in your own complete environment—even when you were actually a client "your" computer desktop had a separate copy of the document that you worked on. No more: while we struggle to come to grips with the social changes accompanying digitalization we find ourselves undergoing yet another shift off computers and onto the web. This shift widens the scope; it is easy to have a single unique copy that many people alter in addition to single, stable copies and many transforms of the original. That shift promises to make it possible to do our work with less duplication—of files, of storage, and of processing power and promises to pass the savings on to the final user.

Really, it's all about leverage
The world is shifting and Google, with one of the longest levers, is trying to increase its leverage by moving the fulcrum ever closer to the weight it wants to move. The whole point of levers is to move a huge weight with a small force and the closer your fulcrum is to the weight you want to shift the greater you mechanical advantage. [image] The huge weight that Google wants to move is the "dead weight" of the existing paradigm of single, local, users that periodically transfer files. The emerging model is one which shifts toward multiple, distributed users that remain connected to files that are, themselves located in multiple, distributed "places."

The new Google OS is all about building an OS that is optimized for that new environment. Right now we have an operating environment in which we are using a computer/local-user-centric OS to access the web. From the standpoint of web-centric use such OSs are bloated, slathered over with useless "features" and surprisingly anemic when it comes to operating quickly and securely within in the new "always-connected" world.

Note that moving us in this direction is what Google has been from the beginning: making it easy and cheap to move to a web-centric mode of interaction. Google's innovation in web search is all about using web links and web stats to make good guesses about what is sought. That made finding things much easier—and then they made if free...It displaced a hierachical organization (cf. Yahoo's (still extant!) example) arranged by respected experts that more closely resembled the library's Dewey Decimal System or Linneaus' taxonomy than anything that we'd now call search. You can perform pretty much the same analysis for Google Apps, Google Chrome, Android, and, now, the Google OS. Those are all fulcrum points that give Google (and Google's user) additional leverage as we shift the weight of the past. With Google OS that point is very near the center of gravity of the opposing paradigm.... The point here is not that Google does NOT have want to "beat" Microsoft (or Apple or Linux) at any of these tasks. It will be sufficient for the purpose if the new browser or operationg system forces a shift on the rest of the field. It will be quite alright with Google, I suspect, if MS beats them in the browser war as long as the winners all support HTML 5-Ajax-multiple threading and the like. Google will have won if its Apps—and similar web applications that rely solely on nonproprietary foundations—run beautifully on all browsers. It is investing in winning the war; not the battles.

If Microsoft, or Apple, or Linux responds to a Google OS with popular instant-on, secure, web-centric OSs and Google's dies a slow and embarrassing death the larger battle will have been won. And, for my money, that is the most likely outcome. Google to date has done an amazing job of creating the ecology in which it can thrive. Google Search made an impossible-to-navigate complexity suddenly usable—and that encouraged the myriad of small, eccentric, impossible-to-classify sources to find an audience and thrive. That in turn made search ever more dominant and gave Google search the page views it needed to thrive through even the lightest-weight advertising. The old hierarchical web was designed by and for graduate students. The new searchable web is usable by almost anyone who has a vague idea of how a topic is discussed.

Now, back to the topic

Google is leveraging the brutal fact of efficiency, its method is so much more cheaper per person than the oldr way that it can afford to give us significant services for free. We do waste enormous amounts of processor cycles and memory storage. The current system is inefficient by design: We buy memory to store our copy of a file stored (but not easily accessible) in a myriad of other places. How much space do you devote to browser cache alone? We purchase computers with several times the processor power necessary to do what used to be called supercomputing (and was illegal to export only a decade ago). Indeed, much current supercomputer design is consists basically of hooking up many personal computers or even game consoles together through a very fast network. We only very occasionally need the enormous power that is at our fingertips in the current personal computer. Web-based apps and systems do not need to waste anything like that amount of firepower. The difficult, processor-intensive tasks can be done on the web. The big storage can be on the web.

The web is, or can be conceived of as, a big, oddly configured computer. It's got great memory and a great, if wildly distributed, CPU. And it can be radically cheaper to use because of those facts.


The Catch
But, the catch is that the web is great computer that has lousy and expensive I/O by comparison. It is only the beginning of a great computer. You have to be a touch geeky to recognize all three parts of a computer...memory, cpu, and I/O. We are sold computers and parts on the basis of memory and CPU speed; not I/O. I/O is code for input/output. It defines what sort of and, crucially, at what speed, information can flow in and out of the computer. On your personal computer I/O is seldom a bottleneck and its expense trivial. Not so for the web where the I/O is the network itself. On the web I/O IS the bottleneck, always.

Most of Google's initiatives can be conceived of as trying to find ways to minimize the effect of the webs' I/O bottleneck. When we hear talk about running faster or yielding a better user experience that is what is typically where the real bottleneck is. Google Apps, Google Gears, Google Chrome, the Google OS and more are all shaped by getting more out of a slow and expensive connection. They've bee surprisingly successful. (The idea that you can do good word processing over the web is really pretty shocking.) The Google OS is merely the latest and potentially most powerful way to evade that constraint and keep that huge weight moving.

But, really, it's all a sad hack.

Google needs Lafayette, and Amsterdam and Vasteras and....
What Google really needs is for everyone to have better, much better, bandwidth. And damn near no latency too, while you're at it. Google needs Lafayette, and Amsterdam, and Vasteras and every other local fibered-up high-bandwidth network in the world as testbeds to showcase what is really possible. It (and others) need a place with no I/O constraint, with a network that has the quality to take advantage of the infrastructure that it is building and surely wants to extend. It needs to build an on-network cache and server system to explore how it can use a decent I/O network to compliment its current products and develop new ones. It needs real communities to really test those new ideas. (Like Google Wave, which could be launched today in a place with real bandwidth.) Google is creating the conditions for the next big shift. It'd be a pity if like xxx it moved the world only to find that the effort had left in a place where others benefited first and most.

If Google's attempts to move the system can be understood as trying to shift the fulcrum to give them more leverage, promoting big-bandwidth communities might well be likened to making the lever longer...that is what most needs to be changed to really shift the old world to a new place. And Lafayette just might provide that crucial place to stand and use that longer lever.

Lafayette is a special case...
because Lafayette is a campus—it provides 100 mbps of speed, with amazingly low latency, between every household it connects. It's hard to overstate the value of that. What make most great networks less great is, ironically in this context, network effects. In most cases network effects are good [] things...the value of your phone connection only increases when your neighbors also get one. But if your network is great and other networks that contain the people you want to contact are not then the added value of what you get from a great network is seriously diminished. So Google, with its large suite of apps that emphasize interaction finds it difficult to find a population that has a large enough population to use its products who all have the same fortunate circumstance. Even networks, like Verizon's here in the United States, which have some higher bandwidth tiers sell mostly lower bandwidth tiers. And they do NOT give their customers large bandwidth between themselves. These networks do not form a cohesive pool of high-bandwidth users.

Lafayette's will.

And, wait, there's more! What Vasteras teaches us is that a high-bandwidth community can flip from having most of its traffic connect to places outside of the local community to making most of its connections inside its own network. Various reporters say that 70% to 80% of Vasteras' traffic is internal. That really shouldn't surprise us; it has happened before. When the first phone networks were built they were conceived of as substitutes for the long-distance telegraph and few thought their use would extend beyond the business world. In short order, of course, it became apparent that the people we actually want to talk to are right down the street; those are the people we know. Phone traffic is, and has been for a long time, mostly local and the widespread adoption much less expensive long distance calling has not changed that.

There is no reason to think that a more robust network, one that is rich in ways to communicate will not follow a similar pattern. People want to communicate and trade information with each other, not someone far away.

Lafayette et al. needs Google
Google can make the local network truly valuable, it can significantly erase the negative weight of the old network by locating caches and services on the local network. Local networks like Lafayette's need that support to make their own business case. Such networks would be wise to court Google (and many others, Google here stands for the new web aborning) and to suport the company in its efforts. A partnership would be of enormous value to both sides. And would help in shifting that weight.

There's a major shift underway; it's hard not to feel everything straining toward that change. But a single constraint keeps the current edifice from falling: Bandwidth. Kick out that constraint and the new web comes into its own. Quickly. There are a few places where that bandwidth constraint is not in place. Those are the places where, with a little judicious midwifery, the new web could be born. And Lafayette shows how the initial densely interconnected communities that would kick-start the process could be developed.

It is a dream. But it is just barely beyond our grasp.


Speed said...

I've read about this Google OS, and it's basically going to be ANOTHER Linux distribution with whatever packages Google thinks should be installed. There are already tons of lightweight Linux distributions that COULD work for a lot of people, although Ubuntu is the closest thing to user friendly there is.
The main problem I see with this is that every PC out there has to have an operating system. The main problem with any Linux distribution? Drivers and software compatibility.
For software, not everything can be web-based. Even though most people say they only browse the web and check their email, there are still times when they are doing things, like looking at photos on the computer, that you wouldn't want to funnel through the web.
And as far as drivers, when people get All-In-Ones or special equipment they want to use, they are at the mercy of the linux community to make drivers, where just about every device now comes with Windows drivers.
I'm not trying to say that a faster internet won't help us in web apps. It definitely will. I just think this Google OS is a bunch of hype that won't amount to anything soon if at all. Even though Ubuntu has made great strides for Linux distributions, it's still nowhere close to dethroning Windows. As a matter of fact, one motherboard manufacturer was going to make a "quick boot" option to where the PC boots to a mini OS and allows you to run firefox. So booting quickly to a web browser is not something new.
I think one day we will have MOST of our apps web-based. It definitely helps with not having to worry about keeping the software up-to-date and local PC problems (unless it affects your browser or internet connection). The problem I see is that until the internet becomes much faster and reliable for EVERYONE, I don't see that day being soon. And no matter what, there is always going to be the need for some local apps, so you sill need an OS that can do what you need it to.

John said...


No quarrel with what you's all pretty much true. Considered alone the OS announcement isn't huge. It is only in the context of the change I think is in motion that it's interesting. I do think the import is significant from that vantage and the recent OS news is just my point of departure to look at a larger skein which Google is taking the lead in weaving.

They're not alone in that and may not always be the bellwether. But they are right now.

The skein I see is the distributed capacity of the internet in terms of both computational power and static storage is already easily accessible in such quantities that it is the equal to the laptops most of us use...That pretty much is my explanation of netbooks and iPhones and why they can be (of limited) use for some purposes even while being pretty radically underpowered.

What is NOT working well is the I/O and if you really want to bring out the full potential of computing on the internet (rather than being passive consumers of it) you need faster I/O. That's what A faster network is what I think Google needs to realize its larger vision and that's all I'm really trying to say here.

I don't think a revved up OS like OSX or Vista or Ubuntu is going away...but it will not be the only path for casual users. Where cost or frictionless ease of use/familiarity are predominate factors a full OS may become the second option though.

You're right too about the day not coming soon. Except it could come very soon in Lafayette an in a few other places. And companies like Google would be wise to see that and do the obvious things like locate servers and caches here just to try and understand how the new environment will shift things. To really do that right they'd have to locate servers and caches on-network.....

Speed said...

Well I definitely do see this main point, that using the web for apps is where a lot of things are going right now. One thing I forgot to mention in my previous post is backups. Usually, you don't have to worry about Google losing your docs or your emails. But if your PC hard drive dies, I'd say 95% of people don't have a backup. With web apps, Google takes care of the redundancy or backups, so it takes a lot of complexity out of making sure your data doesn't disappear.
And since Google is trying to make everything web-based, fast internet is something they should be very interested in. The more their web apps can feel like a local app, the more it strengthens their position.
Another hurdle with Google though, is privacy. I'm sure most wouldn't have a problem with having their docs on a Google server. But putting sensitive information on their servers is something I wouldn't do. Take for instance the security issue a while back where some users could see other user's docs.
However, with new technology always comes new security issues. I just think companies like Google need to be more clear about how they are securing your data. When they say they can't guarantee that even deleted emails will really be gone off their servers when you delete them, that increases privacy concerns. Google can sure use a faster network but they could sure use giving people a better sense of security about using their apps.