Cloud means many things to many people. Some think cloud computing was invented when HotMail was released in 1996. Is HotMail or any online email system a cloud? If you use Gmail for your email, does that mean you are in the cloud? Maybe.

Defining the Cloud

People who implement clouds define the cloud as a pool of resources available on demand that run in an infrastructure environment. These can be broken down into three layers. Consider these layers of the cloud:

Software as Service

The highest level of applications in the cloud is Software as a Service (SAAS). Gmail, Salesforce.com, Google Apps are all examples of SAAS applications. If your business runs on these applications, you are in the cloud. Advantages of running your applications in the cloud are aplenty: first, your end users will have access to their data and applications from anywhere. No more worrying about VPNs and in most cases (as in Google Apps), no more worry about the platforms your end-users are using. Second, assuming you choose the right service, your data is robust, reliable, and secure.

Platform as a Service

The next level down is Platform as a Service (PAAS). PAAS is a full development stack, running as a service, in the cloud. Examples of PAAS are EngineYard or Heroku for Ruby on Rails development. Services are acquired on demand, resources are metered and billed by usage, and an entire Ruby on Rails stack is available through the service.

Advantages of using PAAS are:
• Pay per use:
• Quick, easy provisioning of resources
• Standardization: all development is done on standard platforms
• Agility: New projects are easy to start; run, expand, and tear down
• Scale up and Scale down: easy to scale up applications as they grow; easy to de-provision applications as they need to be replaced.

Infrastructure as a Service

The lowest level of the cloud is Infrastructure as a Service (IAAS). Infrastructure can be defined as an OS with resources, including compute, memory, storage, and networking. Amazing’s EC2 is an example of a Public Cloud IAAS. Customers can provision, for example, a High CPU Medium Instance. It’s easy to bring up additional instances and add them to the same system. Or de-provision these instances when they are no longer needed. Advantages to an IAAS are the same as PAAS minus the standardization of development platforms.

Public Private Hybrid – What’s this?

Many small to medium size businesses start their operations in a public cloud, such as Amazon EC2. At some point, however, running an entire operation in a public cloud might become cost prohibitive. Or, some companies simply won’t run their operations in a public cloud due to security reasons. It is still possible, however, to leverage the benefits of PAAS and IAAS described above; this is simply implemented as a Private Cloud that runs in a company’s data center (or co-location facility). The Hybrid Cloud is useful when a company wants to leverage multiple resource pools, such as the internal/Private Cloud and possibly a Public Cloud (e.g Amazon). Reasons for doing this are to handle peak demands or for failover / redundancy.

So far so good – look for another post about comparisons between Heroku and EngineYard.

One thing that was difficult (but then easy) was figuring out a way to rewrite from the canonical domain (whackanumber.com) to www.whackanumber.com. There are various opinions about whether this is necessary for search engines as it used to be believed that search engines (Google) would treat content indexed at whackanumber.com and www.whackanumber.com as duplicate content. Regardless, I like it better when I only see one of the above in my site indexes, and so I want to rewrite to www.*

In EngineYard, this was easy using Nginx config, so this:

1
2
3
4
5

server {
  listen 80;
  server_name example.com;
  rewrite ^/(.*) http://www.example.com/$1 permanent;
}

And then using the EngineYard “keep” configs, this was easy:

http://docs.engineyard.com/appcloud/howtos/customizations/make-changes-to-nginx-or-monit-configuration.

However, Heroku’s environment is much more locked down. So no dice on modifying web server config files. Heroku’s DNS docs recommend adding a before_filter to ApplicationController that will ensure the correct domain:

http://devcenter.heroku.com/articles/custom-domains

But I don’t like this approach because this applies to all environments (test, dev, prod, etc.) and I only want this rewrite to be in prod (otherwise it breaks all of my tests).

Instead, I found an article that discusses how to do this rewrite using rack. I have seen postings that have said that the most efficient places to do rewrites in Rails applications are in this order:

1. Web server (can’t do this in heroku, can do it in EngineYard)

2. Rack

3. Rails (via the app)

So Rack is better than using the code suggested by Heroku for two reasons a) it performs better and b) you can specify the configuration by environment (more on this below).

So, steps to use rack are as follows:

1. Install rack-rewrite gem:

a. add rack-rewrite into your gemfile

b. bundle install

2. Add the following lines into your environments/production.rb file:

config.gem 'rack-rewrite'

require 'rack-rewrite'
config.middleware.insert_before(Rack::Lock, Rack::Rewrite) do
r301 /.*/, Proc.new {|path, rack_env| "http://www.#{rack_env['SERVER_NAME']}#{path}" },
:if => Proc.new {|rack_env| ! (rack_env['SERVER_NAME'] =~ /www\./i)}
end

That’s it. Redeploy your app to heroku and every time you type domain.com/whatever you’ll be rewritten to www.domain.com/whatever.

I have a web site that uses MySQL, and the database has a table that contains a 2 column primary key where each column is varchar(3), 2 text columns, a created_at and an updated_at column. The functionality of the site requires both insert and select on this table, and since I don’t need transactions I chose MyISAM per the MySQL documentation. The table acts like a cache of text data, where each row represents 130,000 characters of text divided into the two text columns. So, the table started empty, and each time someone on the website requests something that isn’t in the table, the data is created and inserted into the table (to avoid creating the data the next time it is created).

At first, it looked like everything was super fast. But then after a day or so it was REALLY slow. The table is up to >13,000 rows and the response time was > 12 seconds for each query.

Immediately, I noticed that if I do this query (col1 and col2 are the primary key of the table):

select * from mytable where col1=810 and col2=394

that my result took FOREVER.

However, if I did this:

select col1, col2 from mytable where col1=810 and col2=394

my result came back in .01 seconds. So I’m thinking hmm, it must be the text fields that’s making it slow. But then I did an explain:

mysql> explain select * from mytable where col1=810 and col2=394;
+----+-------------+-----------+------+---------------+------+---------+------+-------+-------------+| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+-----------+------+---------------+------+---------+------+-------+-------------+
| 1 | SIMPLE | mytable | ALL | PRIMARY | NULL | NULL | NULL | 13911 | Using where |
+----+-------------+-----------+------+---------------+------+---------+------+-------+-------------+
1 row in set (0.00 sec)

This doesn’t look good, since it’s not using my index and it’s scanning every row in the table. Then it hit me – my primary key columns are VARCHAR, not INTEGER, but yet I was writing my query as if they were INTEGERs. Not being a MySQL expert, it didn’t occur to me that this would cause MySQL to do a scan of the entire table rather than using my PK. DUH!. So I changed it to this:


mysql> explain select * from mytable where col1='810' and col2='394';
+----+-------------+-----------+-------+---------------+---------+---------+-------------+------+-------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+-----------+-------+---------------+---------+---------+-------------+------+-------+
| 1 | SIMPLE | mytable | const | PRIMARY | PRIMARY | 18 | const,const | 1 | |
+----+-------------+-----------+-------+---------------+---------+---------+-------------+------+-------+

What a difference! With this simple change, my select query went from taking >12 seconds on a 13k row table to taking .01 seconds. Voila, you learn something new every day.

A few days ago I went about trying to detach Twitter and FB. It turns out this isn’t that straight forward because I couldn’t remember how I linked them in the first place. When I researched how to link them, I was directed to the Twitter app on Facebook (http://apps.facebook.com/twitter/). So I figured great, simply press the “Deny” button and this should work great… NO! It didn’t work. Maybe I needed to logout first for the settings to take affect.. NO! I kept posting “Testing” messages on Twitter and they kept showing up in Facebook. Annoying!

I finally found this, and it worked:

1. Login to Facebook and go to Application Settings

2. Look for Twitter in the list. If it’s not there, use the “Show” drop-down box (upper right) to filter on “Authorized”.

3. Edit settings for Twitter

4. Under “Additional Settings”, uncheck “Publish recent activity to my wall”

That did the trick. Woot!

Since I will still want to post some things to Twitter, I am now using Selective Twitter (http://www.facebook.com/selectivetwitter) where you can append #fb to the end of any tweet and then Twitter will post this as your status in FB.

These days, location has become a lot more interesting because of the inclusion of GPS technology on mobile devices. All of a sudden, we are using apps like FourSquare on our phones to checkin and receive badges and coupons for businesses that are close to us. The other day, I even received a text on my phone from Papa Johns close to where I was. If I happened to be hungry for pizza at the time, this would have been a great thing (as it was, I wasn’t hungry and it was highly annoying).

So where are we headed with location? It seems like this is wide-open territory, especially as people start to replace traditional PCs with devices like an iPad 3G where you will always be connected and have access to location information. I wonder if this will create new types of collaborative opportunities. Or maybe advertising (this is obvious as it’s already happening). Maybe there will be apps that do things like collective rating of local businesses. I like this because it is a nice way to filter through the massive amount of information on the Internet and deliver to you (the user) data that is most pertinent to you, based on your location. That’s good stuff.

But maybe that’s because I’m old-school and I like to find value where there’s real money changing hands. So the concept of a corporate business model for Web 2.0 is intriguing. Many of the companies listed are ones that we are all familiar with, such as Avenue A / Razorfish and salesforce.com. But the one I really like is Jive Software. These guys get the concept of collaborations and the value of communities in this space. What’s great about Jive is that they get the value of the group in collaboration, which is something that (surprisingly) not all enterprise software vendors understand.   

For me, Facebook was an obvious place to start as I have received the most invitations to join Facebook or view other people’s information in Facebook. So I joined and uploaded some information about myself. Then I started receiving super pokes and all kinds of other gimmicks from Facebook friends. I can see how this is a fun tool and a great way to keep in touch with friends and family as I can upload pictures of vacations and post status about what I am up to.

However, I am struggling to figure out how this is valuable in my professional life, especially inside a company. For developing my career and networking, it seems like Linked In is a lot more valuable. Either way, there are certain things I would like to publish on Facebook that I would not like my work colleagues to see. For example, pictures of my family hanging out in bathing suits in our last family vacation. No problem on the family seeing this, but probably not the guy down the hall at work.

Inside the company, I have heard stories about companies as large as GE who are heavy users of Facebook and rely on it to get their work done. I don’t get it. Are they using it to share documents? Or maybe this is a simple way to publish project information and status? Aren’t wikis and blogs a better way to accomplish this?

Am I missing something?

It is great to see such validation for the product that so many people have contributed to over the years, including the super-smart team of developers that originally took Parlano out of UBS Warburg and the amazing team of people that we have on staff today. Needless to say, Parlano would never have reached this point without the help of so many talented people. Now it’s exciting to think about how much bigger the product will become as we integrate it under the Microsoft Unified Communications umbrella.

At Parlano we have always thought that Persistent Group Chat, and MindAlign in particular, is a perfect way to combine groups of people around topic-based conversations, and then to use these conversations to launch into other modes of communication. While a partner of Microsoft’s, we have been working toward this goal for years. Now with this announcement we will be able to put this integration into overdrive.

Thanks to everyone at Parlano past and present. Also, thanks to the great teams of people that we have worked with at Microsoft over the years.