(this is page 88 of 121)





Latest

Coming of the cloud

Doc always talks about how much he hates "vendor sports—" well this weekend, we're getting it in the guise of cloud computing all over the place. First, Businessweek has a cover story on some random Google engineer who is fiddling with the CS curriculum at UW which is really just cover for a puff piece on how we're at the beginning of this whole new type of computing which today's CS students need to learn to program for. Then, today's NY Times has a big piece on how Google is headed for a showdown with Microsoft over a new way to deliver applications to computer users. The blogosphere (which unlike Doc does loves vendor sports) is naturally all over it.

I can already see that people are throwing this cloud computing term around so loosely that it's going to cause more confusion than the whole Web 2.0 thing did. For some people, cloud computing means running applications over the web with the interface painted inside the frame while the executable sits far away in an undisclosed data center. For others, the cloud is about a massive amount of CPU cycles available on demand (but metered). And for others still, it's about a hard drive in the sky.

The reality is that the cloud is about all of these things and more. Perhaps the easiest way to see it is as the extension of anything that might be connected to a PC over the public Internet to some magical data center that gets to provide as much you might need. Need more space? Extend it with a Google hard drive over TCP/IP. Ditto for the CPU. And especially ditto for all of the installed applications.

Most importantly, in what is hardly mentioned in all of the articles, the real kicker is the access device doesn't even have to be a PC at all— in fact, most of the real leverage from the cloud is likely to come to all of these really powerful mobile phones that are capable of running a web browser and mimicking that part of the PC. That is where the rubber really hits the road on this stuff!

If you look it as extending the computer on your lap (or in your pocket) on demand, then the next natural question is: what other things are you using today in that computing experience that you'd like to have re-attached in this fashion? Let's take a retro example: what about the 400MM printers we've shipped at HP over the years? What is the cloud version of the print driver? (Our team is actually working on a version of this, but more on that later) We're not doing our PR well here because every cloud article talks about the big four: Google, Amazon, Microsoft, and IBM, and leaves HP— the only big IT company that spans every type of access device and peripheral with services and servers— out of the running altogether.

Finally, it's funny to see Schmidt (Google's CEO) in the role of the guy with the new disruptive platform, mostly because he's spent almost his entire career (Sun, Novell) fighting for the losing platform— each time getting disrupted by none other than... Microsoft. I'd hardly count those guys out of it at this point, but it is interesting to see how much more the platform disruption can count for over the leadership (stewardship) from the guys in the corner office. Schmidt has had a few losing hands, but this round he seems to be holding a full house and that, more than anything else, may make all the difference.

Comments Permalink

Yesterday I wrote about how micropublishing comes out of intercepting a communications stream and making the messages chunks of published content, ending the post with an argument that this will only work for a mainstream audience if it can be totally unobtrusive to the communications channel. Even then though, I have my doubts as to whether communications-parasiting micropublishing could really go mainstream (whereby I mean the "regular" folks piling iPhones under the Christmas tree this year).

What I do think serves as a great model for micropublishing is when applications or better yet, devices, can be properly instrumented to emit a steady excrement of relevant metadata that can then be "massaged" into something that resembles a micropublishing content chunk. Since that statement takes the Abstract Award for the weekend, let me follow it with some concrete examples:

Application: I love Google Reader. While the province of true geeks, RSS readers strike me as the next killer application for managing the vast information flows that we're all crumbling under, and as far as they go, Google Reader is tops. It is fast, reliable, and best of all, keeps sprouting these interesting social features that are likely to make it a very interesting product to watch over the next year. While Reader has always had the notion of "shared items" that friends could subscribe to, the process for doing this was cumbersome in that it required your friend to seek out the syndication URL of your shared items and the subscribe to it. This past week, the Reader team launched the ability for you to see what friends have shared without having to explicitly subscribe, a great improvement that I think will finally raise shared items to the level of first-class feature inside of Reader. This matters because every time a user hits Shift-S on an item, that is in itself a micropublishing event (on Tumblr this is called a "reblog" but Reader makes it seem like more of a non-event). And these events strung together make a great frozen flow of ideas.

Device: A location aware mobile device is the best possible excrement generator for review-based micropublishing that I could possibly imagine (assuming of course the phone is open enough to application developers). While we constantly debate the merits of the "thumbs-up" flyby reviewers at the office, there is no doubt that when I'm evaluating a restaurant I've never eaten at, a barber who hasn't yet nicked my ear, or a hotel I'm about to entrust my sleep to, I'd rather have that information no matter how hastily it may have been generated than not. And these are categories of services that I research today— with an ever-present ability to thumbs-up/down anything at any time, who knows what else we might start micropublishing in binary about? Playgrounds? Busy intersections? Barristas? "Hot" spots?

In both of these cases the work we expect from the user publishing is as close to zero as possible within a certain limit. In fact, taken to the limit, you get models where just by doing stuff, you're generating metadata. The Google Pagerank index for instance is the inadvertent result of billions of micropublishing events (links being created) and its aggregate value is huge. Similarly, the new GPS in my car gets its traffic information from over 1MM radio-enabled GPSes that get a free ride on top of all of the semi-tractor trailers you see on the highway.

However, there does seem to be a limit beyond which it really is just metadata— potentially valuable in the aggregate— but wholly useless to the type of micropublishing that facilitates social connections and fires off the creative endorphins in people. I'm not sure where to draw the line (after all some types of metadata can go either way depending on context— for example, my Nike+ running sensor data), but the key is that the closer we get to the line without going over, the faster we might see micropublishing getting a foothold with a huge audience.

Comments Permalink

When one of my favorite tech blogs titles an article "The Evolution of Personal Publishing," I am most definitely going to bite on it (after all, this is a topic on which I like to noodle). As it turns out, the piece mostly adds color to a post by Fred Wilson who argues that "microblogging" (Twittering, social sharing, etc.) is actually not just blogging under a new guise but a new type of communications/publishing hybrid that not only deserves its own category, but may actually reach into the mainstream Internet userbase in a way that its progenitors (personal publishing and blogging specifically) have not been able to.

Why? Because according to Fred, micropublishing at its core is lightweight (low cognitive load), social, and interactive (which strikes me just social under a different name). In other words, it is easy to do and keeps you in touch with your friends and family as well as the world at large, when appropriate.

And in fact, it seems that this last bit is the critical piece. The relevant part of all of these lightweight tools is that they intercept a communications flow and create a permanent record of it that people who may not have been in the original dialog— people who you may not even know— can have access to it via the archive (the frozen flow).

A little over a year ago, I was eating lunch at Google with Jason Shellen who was the first person to pitch me Twitter. As I was arguing that this seemed just like SMS + groups he said something which has stuck with me ever since: that Twitter allowed every SMS to have its own permalink. I remember my initial reaction to his statement as though it had been yesterday; I thought: Silly Googler, not everything needs to be in your index.

He was absolutely right about the significance being the addressability and persistence of each SMS/tweet, but not because we might be turning up the fact that he had too much to drink for Thanksgiving in the Google index 3 months down the line. Instead what the permalink implies is that what was a disposable message is now a micropublishing event which can in effect become a vector for socializing with people who would otherwise not be in your regular communication stream.

Now the question still remains: how much do regular folks really care about the potential for this type of publishing-sourced social serendipity? Especially when the insertion of micropublishing into some communication channels can have unanticipated adverse effects? An example: I recently tried to give Twitter the good ol' college try by getting everyone at work to use it as well as a few key friends and family members. My older brother (who ironically in this small world was once one of Fred's entrepreneurs) puked on it because he decreed his SMS inbox to be a high-priority near-synchronous channel for communication with a select few. By overflowing his phone with tweets, he was convinced that Twitter was "breaking" the promise of SMS. The serendipity of social experiences on top of micropublishing was just not worth it.

(And before people write to me to tell me that you can turn SMS off on Twitter and just view the tweets on the website, think about this: why would a normal person want one more website to have to go check every day?)

I think that in fact Fred is right that micropublishing done right can go mainstream— but we have to look to models like Amazon reviews and not just the progeny of Blogger to see how we take it there. More on that tomorrow though.

Comments Permalink

First they made it so that you didn't have to worry about disks with their S3 storage service. Then they took away the need to manage hardware and a colo with EC2. And now for the final piece: Amazon has just announced SimpleDB, a database hosted as a web service that application developers can plug into. Despite one curious choice, this is just brilliant strategy on their part.

Everyone who hosts a website that has any level of traffic spends an inordinate amount of time thinking about how to tune the database. And as with almost every infrastructure layer in the webapp cake, I'm fairly certain that 80% of this tuning work is just good RDBMS hygiene and adds preciously little "special sauce" to the core application. In fact this is why there are specialists in tuning MySQL or Postgres who bounce from project to project with the same basic bag of tricks.

If Amazon does it right, making a database scale is one time-consuming headache web application developers will never have to worry about again. But it goes beyond scaling too: backups, replication, failover— these are all common database chores that Amazon just makes go away.

On to the curious choice: no support for full (or even partial) SQL. I'm guessing that this may have had something to do with the complexity involved in supporting SQL on their super-scalable, super-distributed architecture as I'm fairly certain the Amazon folks realize just how much momentum SQL has in the web application space. And if anything holds people back from adopting SimpleDB (outside of potential uptime/performance issues), it will be the need to learn a new way to store and query data (albeit a very simple one).

A potential work around: open source libraries that substitute the back end of the most popular ORMs (Rails, Hibernate, Django) with the SimpleDB service. Now that could be a game changer.

Here's to never thinking about sharding databases again.

[Update: More info on the guts of this beast.]

Comments Permalink

It usually takes me hearing really clever smart (but crazy sounding) ideas three times before they begin to sink in. So hopefully, noticing them at two is an improvement. A couple of months ago, while attending a conference with Alan Kay, I heard him talk about how any new modern computer should be designed with "Internet principles" at its core. That is, the hardware and OS should be designed to support small isolated "components" that behave like nodes on the Internet, sending each other messages through a TCP/IP-based networking fabric even on the bus of the computer itself. Knowing that he is the perennial object & message nut, I filed his statements away as curiosities, thinking only that while his observations about how the unstoppable trends of virtualization and multiple cores as the base substrate for this fundamental change in the architecture of every individual computer seemed clever, the practicality of it seemed elusive.

After all, the stored-program, shared-memory, single instance computer exists for good reasons (simplicity, efficient, cost) right?

Imagine my surprise then while running today to the latest episode of the Google Developer Podcast (hit-and-miss, but pretty good as of late as far as podcasts go), the "All About Android" episode which basically described an implementation of what Alan was talking about in almost every aspect of system design: from the message-passing model for app interop to the way resources are described internally. It gave me pause for thought— after all, if resource-constrained mobile devices are being designed around these architectural principles by smart people, then maybe there is something to this notion of turtles all the way down, from the services we use from the cloud to the devices we carry in our pockets.

I think I've let Fake Steve with all of his Googletard ranting color my perspective too much, as I've paid scarcely any attention to the Android project. And truth be told, I wanted to write this before digging into it— after all, ideas are much prettier as abstractions, and Android seems to be one weird beast of a mix of things on the face of it: Linux + Java + Webkit, all on very resource constrained but heterogeneous hardware platforms. I mean, we're talking about serious potential for a fly in the transporter with this baby.

Still, it's the second time I've heard it, so time to pay attention...

Comments Permalink