So, who is next?

Recently Pinterest has gained a huge momentum. As per recent Mashable/Experian report, It surpassed Linkedin and Google+ as being the third largest social network after Facebook and Twitter.

• 1. Facebook: 7 billion
• 2. Twitter: 182 million
• 3. Pinterest: 104 million
• 4. LinkedIn: 86 million
• 5. Tagged: 72 million
• 6. Google+: 61 million

My first guess especially after this move by Facebook is Google. Google is struggling to make Google+ relevant in social media. Experian says that their traffic has increased by 27% recently and all the interviews with Larry Page on his emphasis on social web. They have a great advantage because of youtube and I am sure they will use it well. Instagram would have been great addition for them but looks like they missed the boat, so Pinterest be ready for a call from Google or who knows, it may have already happened.

Other suitors may include Facebook, who still have a ability to do this after Instagram. Microsoft, who is still sitting on plenty of cash. Lets see how it unfolds.

The numbers sound impressive. Growth seems great. The requests per second is actually a more impressive number - 650,000 request per second is something to think about. Lets compare that to Facebook, which does not say the actual numbers but here is what they say in this post from 2010

A few of the big numbers we deal with:
* 500 million active users
* 100 billion hits per day
* 50 billion photos
* 2 trillion objects cached, with hundreds of millions of requests per second
* 130TB of logs every day

Now, going to Twitter, I could out find a lot about exact latest numbers but a quick search led me this Techcruch article from last year.

I know this comparison is apple to oranges but gives us some idea when arbitrary numbers are thrown around.

Now, for the big question, what does Object mean in Amazon S3, lets say in actual storage terms that we can understand, like gigabytes or terabytes. Again, it was not difficult to look it up, on Amazon S3 FAQs

The total volume of data and number of objects you can store are unlimited. Individual Amazon S3 objects can range in size from 1 byte to 5 terabytes. The largest object that can be uploaded in a single PUT is 5 gigabytes. For objects larger than 100 megabytes, customers should consider using the Multipart Upload capability.

So, that means that these billions of objects could be of any size and we will not know how much data we are really talking about.

BTW, that original blog post written by Jeff Barr was written as a recruitment post adding a number of positions that people could apply too. As was the data presented by Twitter at UC Berkley last year at a recruitment drive. So, while the blog buzz is one, I would wait till Amazon announces their quarterly numbers before jumping to conclusion on their growth numbers. They did pretty well last time on AWS

“Other,” the revenue category in Amazon’s reports that encompasses Amazon Web Services, is growing like mad — 70 percent over last year, in fact. This matters because it likely means AWS is outpacing its projected growth and is rapidly approaching a $1 billion run rate.

Lets wait till end of this month for their earnings to come out.

After all, Amazon is a retailer. An Internet pioneer but it is still in retail business. Retail business in general are very low margins business. So why then would Amazon make heavy investment into these Amazon Web Services?

The general explanation given by some gurus, Amazon computing power is built to handle the Christmas season load and more. Rest of the year, the systems utilization is really really low. And thus AWS is lets them monetize that infrastructure. That makes sense to an extent. Does it mean, during the heavy infrastructure usage Christmas season, AWS users will see slower response time?

So, again I started with some of their financials to see if they do a better job at detailing this than Google does. I was pleasantly surprised. To be fair to Google, Amazon has been doing this much longer, both the utility computing and publishing financial reports. Here is what I found in their Q42007 report

The other one where they actually talk about developers as customer

Upon further examination, we can see that their technology and content expenses are nearly 25% of thier total operating expenses. That is very high investment. So, monetizing that should be pretty

Shifting gears here, I put down the financial reports and their marketing blurbs and then pick up a ‘how to’ book from O’Reilly, called Programming Amazon Web Services. Browsing through the book, we can see that the techies at Amazon have done a great job at the service oriented architecture for AWS. As the book tries to explain, they built it for AWS as their own platform to develop applications. I am impressed by the loose coupling (which all SOA should be), support for multiple access systems (REST, Query and SOAP) which makes it open for access through different systems. What I do not like, at least so far from my browsing, is that are SlAs are very weak. They talk mostly about giving a user service credits if the service goes down. This would make it tough decision to deploy a mission critical application in an environment where uptime is not guaranteed with some indemnities. This is one thing Amazon will need to reconsider, though I am sure there is a battle between lawyers and techies going on about it right now.

Going through the resources on AWS site, I also discovered that Amazon has another program in beta called Amazon Devpay. This is where the developers can showcase their applications that are developed on top of AWS and they have billing and an account management application for their customers. This a great motivation for the developers to build on these services. It takes a community to promote these type of innovations.

So, to conclude, I think Amazon is selling a real service. Based on feedback from people who have used it directly or indirectly such as customers of Basecamp from 37 Signals, it seems like this model is here to stay.

I also think that Amazon has laid the model for other larger companies to deploy and leverage their internal systems to generate revenue [anyone from IT reading this?]. I am pretty sure that other large sites will open up their services for external usage. I am sure there are companies that would like to integrate with services from travel sites such as Travelocity or Expedia. The is true implementation of the web services vision and I am glad it is finally coming together.

Can the ISPs in US do the same for content on Youtube or video google where you have watched the laughing baby thousands of times, who guarantees the QoS, the content provider or the transporter?

If so, where does this stop? If you start using services such as Amazon S3 and pulling large files over the ISP provided bandwidth, who pays for the extra bandwidth usage? Will enterprises use cloud based services if they have to pay for extra bandwidth? Will it be metered or unlimited services? Is this the reason, Google wants to invest in spectrum in the air and fiber in the ground?

Again, interesting times!

So, this cloud is really massive data centers spread across the globe with lots and lots of servers, both real and virtual serving hardware and software services for customers. (Of course, these data centers better be green and energy efficient.). So the winners in this game will be the companies that can run a data center most efficiently. Google, Amazon, Yahoo and other large portals base operations have had head start.

But what about the internal IT teams at large corporations? In the last few years, many of these have outsourced their data center operations to companies such as Verizon or Sungard, so I am not sure that would more of difference to them.

In any case, this is interesting times for all.

As a side note, if you are in IT professional in data center operations, system administration, virtualization savvy, good times ahead for you.Be sure to take advantage of that

Let’s start with SaaS, Software as a Service. SaaS as the name suggests is a delivery model to provide software as paid service. The software is not installed on your own machines, but the vendor provides the complete package as hosted solution including the infrastructure. The software vendor may in turn rely on hosting providers who manage their server farms, provide disaster recovery, load balancing, mirroring and other infrastructure needs, but the end user does not have to be concerned with this. The end user is mostly interested in functionality, costs per user, uptime, associated SLAs, ability to integrate data with other systems etc. The customer pays for the software on a subscription basis. The poster child for this is Marc Benioff’s salesforce.com.

In general under SaaS, the software delivered is multi-tenant model, as in multiple customers can use it with the clear understanding that one customer will not be able to access another customer’s data. Multi-tenant is not the same as virtualization, a misconception spread by even the gurus in this world. Virtualization is the ability to slice hardware to represent multiple virtual machines (VM). So for example, a high end Sun server can be sliced into 10 different virtual servers. Each VM then represents a server. One VM may host the SQL database on one virtual machine and a web server on another etc.

Providing computing resources including software, storage or hardware as subscription based service is utility computing. The key difference here is utility computing is not limited to software. It could be any form of computing resources. The mainframes providing time-sharing was earliest form of utility computing. Any ISP renting disk space to store backups is another simplest example.

On-demand software is synonymous with SaaS and On-demand computing is synonymous with utility computing. It is opposite to on-premise where you purchase the software/hardware and deploy it on your own servers/networks.

Now coming to cloud computing. There has long been the dissociation of physical infrastructure and software application design. Cloud computing takes this a few steps further where designers and users need not be concerned with how the different computing resources and services that are combined in an application or service are coming from. For example, Amazon S3 is providing storage space to many applications including Basecamp.

Even though we generally talk of external services such as Amazon, clouds could also be internal. Many IT departments have been functioning in this way for a long time. A simple example is having a centralized group to manage all SQL DB resources. The DB group within IT provides an SLA based service to all other groups where developers from can request DB resources for development, testing, staging and production environments. They work with the DB group getting sizing information based on their logical models, number of users and growth expectations.

There are a number of other concepts that come into picture with cloud computing including business services management [BSM] where your business and IT services may reside in or outside of your enterprise boundaries, business process management [BPM] where the services orchestrated may be provided by multiple vendors or internal resources. Enterprise IT architecture will now need to go way beyond the enterprise.

Does this give a sense of impact of cloud computing? The promises are many but fulfillment of this on the way. Reminds me of an old Arabic saying, cloud is a promise, rain is the fulfillment.

]]> http://cloudinsight.com/2008/04/hello-world/feed/