Wednesday, September 30, 2009

The Folly of the Amazon Cloud Economy

I have heard of dozens of startups that are offering services that Amazon does not. The big problem with these businesses is that are not taking into account what Amazon is.

Amazon is a Fortune 200 company. They have all the technology that you need to run a huge enterprise. They just have not released it or figured out how to bill for it.

They are just waiting for some small startup to prove that there is a market for the product. They go back to the labs and pull the code out.

If you think that there is no way that Amazon will ever go after your market, it must be too small to be Venture funded.

Thursday, August 13, 2009

Feeling Depressed

I believe that Platform as a Service (PAAS) will be the end result of the Cloud Revolution. One or more of the PAASs will be the new cloud operating system. You can build systems with Amazon or another Infrastructure as a Service (IAAS) but that is like assembling your computer from chips and writing your own operating system.

The reason I am so depressed is that I think that the most likely winner will be Hadoop. Hadoop is written mainly in Java. Java is the result of design by committee like COBOL and ADA. The answer to any API problem is to add more API's.

The only hope for any elegance in programming is if a better scripting language is overlaid on it like Jruby or Jython. I also think that Google AppEngine will remain a viable competitor but I don't think that Google will push it enough for it to win.

The new Windows will not be Linux. It will be Hadoop over Linux over the EC2 API. Hadoop will probably be more reliable than Windows. Thinking about Java surviving is just sad.

Monday, April 20, 2009

Oracle says that it buys Sun for Solaris and HW

I don't buy it for a second. IBM's exclusive to look at Sun's books expired only two weeks ago. Oracle can not have done a proper due diligence. Oracle says that it is buying Sun for the HW business and Solaris. They barely mention MySQL.

The vast majority of Oracle's software revenues come from it's database. $3B out of $4.4. Of the remaining $1B in revenues, $758M come from consulting services. This probably has to do with it's database.

Oracle paid $5.6B for Sun. It made $11.64B last year in database revenue. It is projected to make $13.35B this year. If it could increase it's database revenue by 50%, it could pay for Sun in 1 yr.

MySQL is the most popular database for new startups. Facebook has 1800 MySQL servers and zero Oracle servers. I remember asking an Oracle employee if Oracle was concerned about having approximately zero market share among the hot new companies.

This is very smart. They have to say that they want to sell HW. That battle is lost. Oracle will sell Sparc to Fujitsu or shut it down as soon as it is not embarrassing.

Oracle will not kill MySQL. It will just neglect it. If it gets any further revenue from Solaris, Java, etc. That is just a bonus

Monday, April 13, 2009

Platforms make headway

Amazon announced hadoop support on EC2. Open source is slowly moving to creating a platform to compete with Google AppEngine. There is no doubt in my mind that cloud computing will evolve to new platform.

The two top contenders for dominent cloud operating system are Hadoop and AppEngine. Google AppEngine scales easily and is slowly building out more features and becoming more open source. Hadoop needs more admin features to make the scaling automatic but it has access to the machine level.

Simon Phipps has complained that AppEngine does not support all of the Java API. That is like complaining that planes do not support saddles. It is not possible for AppEngine to support all of the API and still remain fast.

Many Java programmers wanted Java support so Google did the best job it could. The Java folk should just consider it J2Cloud or J2Fast.

Sunday, March 1, 2009

Is Cloud the new DotCom?

I saw the title of this article on Elastic Vapor. It is interesting because I have presented at 1 cloud meeting and attended 2 others in the last 7 days. I was at the meeting where Reuven is mocking George Zachary.

Any industry that has 3 meetings within 7 days has far too high a level of froth. Is Cloud Computing the next DotCom? Of course it is. There are too many conferences and companies being formed for most of them not to tank.

There were many big companies that came out of the DotCom era like Ebay, Yahoo, Amazon, Expedia, etc. There were many that died like Webvan. The first Platform as a Service play has already died. Coghead has already gone bankrupt.

There are far too many Software as a Service plays being started up. How many online CRM systems does the world need?

Cloud Computing is real. I believe that it will be bigger than the first Internet wave. Cloud Computing is the next huge opportunity and the next crash all rolled into one.

Cloud Computing is an enabling technology. It lets startups build companies faster. It also let's them die faster. No data center to break down and sell on Ebay. The company stops paying it's compute bill and the data vanishes. Like a cloud. :)

Tuesday, February 10, 2009

GoGrid command Line tool

GoGrid has a command line tool called cloudcontrol. I replicated most of the features in a script called gogrid in cloudwizard/bin. I did not bother with some of the load balancer features. I did add a nice ssh feature.

gogrid -c ssh --id IdOfServer
will find the password and give you a commandline direct to the server.

Monday, February 9, 2009

Adding GoGrid

I added the GoGrid API over the weekend. It was very simple. It only covers the compute portion. I could not get the storage function to work.

There were a few confusing things like the IP interface. If I start a server, why doesn't it just use the first unassigned IP? I am guessing that the API did not want to guess whether you wanted a public or private IP. I added that functionality with a public default.

There is a goofy thing with the API. Each server can be identified with a name or an Id. The name is not required to be unique but they do not issue an Id until the server is completely up.

Sunday, February 1, 2009

Why Cloud Wizard?

The biggest annoyance with Open Source software is the time it takes to install. Much of it is poorly documented and tested. I might download a package and find that it conflicts with something I need. It might take 3 hours to uninstall the application.

One of the big reasons I wrote Cloud Wizard as a command line language was that I wanted to be able to exchange scripts. I thought that an Open Source author could write one script that could be copied off the web page. I would run that script against any compute cloud like Amazon. I would test a fully configured system for an hour and pay amazon 10 cents.

Just not worrying about the problem of uninstalling would be worth 10 cents. If the system could be started fully configured, that would just be a bonus.

Saturday, January 31, 2009


As I was plugging away at CW, it has evolved. The first version was going to be online. Then it was a scripting engine with some command line tools. Now, I think that there is no need for the additional command line tools. Each script can be run by itself with default authentication and a special command line argument parser that is provided.

Each cloud is implemented as a Python module. You can go

from import Aws

You can also go directly to the service you want

from import Ec2

Or you can go directly to your machine

from import Instance

All clouds will follow the same organizational structure.
Cloud/service/machine, directory, queue, or whatever.