Archive for January, 2007|Monthly archive page

Load-balanced EC2 setup on Podango.com

Disclaimer: This is not live yet.

At Podango, we have been working on migrating to Amazon EC2 from our existing hosting solution. We now have a load-balancer in front of multiple web servers and are set up for multiple processing servers and a DB server behind them. We are still a few weeks from going live with it, but when we do, we will have a very scalable system with at least a half dozen servers running all for less than the price of one Rackspace server.

Jeff Barr (Amazon Web Services Evangelist) is visiting Utah for a few days

Jeff Barr is coming to Utah for a few days in February. He is going to be speaking at an event lined up by Paul Allen at Provo Labs, and will also be available to meet with a few other individuals/groups.

As I’ve mentioned to anyone who will listen, Amazon has created the most significant technology available to small and medium-sized companies with their web services offerings. Here is a chance for those of us in Utah to come hear Jeff first hand.

The power of Amazon EC2

So, one thing that is nice about Amazon EC2 is the price. The cost of a server is $0.10/hour, or about $75/month.

Another nice thing is that you can have an instance up and running in a few minutes once you have an AMI ready that you are happy with. No need to order and install hardware, or even wait 24-48 hours for a new server from somewhere like Rackspace.

Yet another really cool thing about EC2 is that you can use it to give you bursts of computing power. Let’s suppose that I have a cron job that does some sort of regular processing (off on a processing server or other non-web server). Eventually, as I get more and more traffic to my site, the processing ends up taking all day. So, eventually, I add a second server. However, suppose that from an end user perspective, it would be beneficial if I could have all that processing done in 1 hour rather than 24-48 hours.

In a traditional environment, that would be really expensive. I would have to purchase 24-48 “processing” servers in order to get all the processing done in an hour. Then, all those servers would sit idle for 23 hous/day.

With EC2, on the other hand, I could instantiate 24-48 servers for one hour, knock out all the processing, and then shut them back down. The end cost is about the same since EC2 charges by the hour. So, I can appear to have 24-48 servers for the price of 1-2 servers (and I would in fact have 24-48 servers, but for only 1 hour/day).

One of today’s Amazon Web Services Blog entries has an entry about a Perl interface to Amazon EC2 written by Jeff Kim that allows you to script the starting and stopping of EC2 instances. This would allow you to programatically determine the number of servers you need, automate the process of starting up new instances, and automate the shutting down of instances that are no longer needed.

Another simple use of this would be to automatically monitor the load on your web servers and instantiate new instances or deactive unneeded instances depending on the amount of traffic coming to your website.

Bottom line: EC2 is amazing.

Google PPC is a waste

What has Google done that makes it so special? They have a search engine that is arguably no better than Yahoo’s. They have maps (big deal). They have gmail–again bested by Yahoo. They have a smattering of other offerings that aren’t doing a whole lot. For some reason, Google has become the media darling of the technology industry and has a market cap many times larger than it should be as a result.

To top it all off, Google’s biggest cash cow is there PPC revenue–a cash stream that is seriously bloated because of marketers that don’t know what they are doing who have been caught up in the Google craze. A few days ago, I sat in a meeting where we reviewed our marketing strategy for the coming year. Of the many ideas we are implementing and/or considering, PPC was mentioned. It was almost immediately dismissed. Why? It is rarely cost effective.

Over the last 2 years, I have repeatedly seen companies try and drive large amounts of traffic to their web sites using Google PPC. At times, they have been successful. However, in the end, they have all lost money. Google is making a ton of money off of marketing people who don’t know either a) the value of a customer for their business, b) the cost per acquired customer using PPC, or c) both.

On more than one occassion, I have seen companies who are happy having an overall marketing budget that comes in lower than the overall revenue being generated for those marketing efforts. What they failed to realize is that some efforts (like affiliate or SEO) made up for losses in PPC. The end result was that they happily spent their PPC money not knowing that it was actually causing them to lose money.

For example, suppose that for every $100 spent on marketing, they were making $120 in revenue. The CEO would look at that and be happy. However, the breakdown of those numbers would look something like this:

PPC
$50 – spent
$30 – revenue

SEO
$30 – spent
$40 – revenue

Affiliate
$20 – spent
$50 – revenue

Net
$20 – in the black

However, if they remove PPC from the equation, then they would be $40 in the black despite a decrease in gross revenue.

Now, I fully understand that different keywords perform better or worse, and that it is imortant to create landing pages to optimize the experience for each keyword, etc, etc, etc. But, the bottom line is that spending on Google PPC is out of control in MANY places. In some cases, I absolutely know that the top 10 bidders for a particular key word are losing money. There is no way conversion rates can possibly be high enough to support the money being spent per click.

Eventually, people will figure it out. Eventually, bids will drop to something much more reasonable. However, I’ve been waiting to see that happen for 2 years, and I have yet to see any real signs of dropoff in the PPC madness.

S3 Infinidisk

The latest Amazon Web Services Blog entry highlights a new product called S3Infinidisk for EC2. This is a really cool idea implemented on top of Amazon S3 and EC2.

At Podango, we have been using S3 for awhile, we are beginning to use EC2, and I will be looking in to S3Infinidisk next week. When dealing with things like mp3 files (like we do at Podango), S3, EC2, and now S3Infinidisk are allowing us to compete with anyone in the industry. SQS has also proved to be very valuable. I’ll talk more about that some other day.

For now–kudos again to Amazon and also to Openfount.

Biggest technology news nobody is talking about from 2006

I am amazed at how little buzz two Amazon web services have generated since launghing in 2006. Amazon S3 and Amazon EC2 are groundbreaking services in my opinion.

A few years ago, Amazon did something very smart. As they discovered that they needed highly scalable and reliable platform components that they couldn’t buy, they decided to build them in a modular way such that they could then profit from these platform pieces by selling them to others. I have read that Amazon’s CEO describes Amazon as a software services company. I would guess that most people equate Amazon to an online Barnes & Noble done right. However, I think Amazon will outgrow that image over time with what they are now doing. Right now, Amazon is my favorite internet company with Yahoo and Google coming in a very distant second.

Amazon S3 allows you to store gigabytes or even terrabytes of data for pennies a month ($0.15/gigabyte to be exact) in a highly scalable and reliable environment. If you need to serve that content up to someone over the web, it only costs $0.20/gigabyte of data transfer. The product is sooo easy to use, and those prices are sooo low, it becomes a no-brainer to use S3 to store anything that takes up any space. People are dumping audio, video, images, mail, backups, and many other things on S3, and having great success.

Amazon EC2 makes the typical small business IT infrastructure almost obsolete. It allows you to “instantiate” servers in 2-3 minutes time. Suppose I have a website with traffic that has outgrown the single-server setup. I can go to EC2 and set up a load-balancer, multiple web servers, a database server, and multiple utility servers for $450/month. The time it takes to do this is very small, and the ongoing costs are very small. Once again, I only pay $0.20/gigagbyte of bandwidth consumed. The ease of use and low price just blow everything else out there away.

The bottom line is that Amazon with S3 and EC2 are making it so small and medium sized companies can have big-company style infrastructure for a VERY affordable price. Yes, people are talking about these services more and more. However, they are not getting publicized nearly enough for what they really are. If Google had released these services, it would be all anyone could talk about. But, since it wasn’t Google, they still continue to fly under the radar (relatively speaking), and that’s too bad IMHO.

A few good links:
Amazon Web Services Blog
Some sites using Amazon’s web services