Ping.sg on Amazon Web Services
This is Simone Brunozzi, AWS Technology Evangelist for APAC.
It has been a while since my last blog post here. I’ve been traveling a lot in Asia lately, and now that I’m back in Singapore for a couple of weeks, I finally had the time to check out an interesting one-man company.
Ping.sg started in 2006 as a blog aggregator and a community for bloggers and blog readers in Singapore. They have enjoyed tremendous growth and the site is now aggregating more than 100,000 blogs around the region and it is still growing fast. In fact, my personal blog is now there as well
With the growth, their need for computing power increased as well. When they found out that a single 1U dedicated server could no longer cope with the heavy traffic, it was time to upgrade. They had to choose between getting a new and more powerful server or using this situation as an opportunity to move their infrastructure to the cloud.
They chose the cloud, for better expandability and flexibility in the long run.
After evaluating a few cloud computing providers, Amazon Web Services (AWS) was chosen because they would have the ability to fine tune their server and architecture themselves to better suit their needs and requirements.
The range of services and API that AWS offers allow them to program things like automatic daily incremental backup via Amazon Elastic Block Storage (EBS) snapshots, pushing their static data to Amazon Simple Storage Service (S3), something that they were not able to do before switching to AWS.
Migration from their physical server to AWS took about 15 hours, with most of the time spent learning about tips and tricks of AWS, as that was the first time they were setting it up. They started with two small EC2 instances with Linux, one intended to act as a web server and the main compute server with various daemons and services running while the other acts as a dedicated database server. However, after running it for day, and with some experimentation, they realized that Ping.sg performed much better by running all services on a single high CPU medium instance, which costs the same as two small EC2 instances, but with more CPU power than the pair of small instances.
Amazon CloudFront and Amazon S3 are used to server static files to lighten the load on the compute server and to exploit the content distribution capability of CloudFront.
Log files are stored on ephemeral storage in order to avoid incurring charges for I/O requests to EBS volumes.
The Amazon Machine Image (AMI) is recreated when there are major changes to the system setup.
>With 80% of their visits coming from Singapore, and over 95% from Asia, they were overjoyed when Amazon announced the launch of AWS in Singapore, back in April 2010. They quickly shifted from US-East to Singapore and immediately enjoyed a speed boost due to the significant reduction in network latency: they report ping times dropping from ~250ms (US-East) to less than 30ms (Singapore).
Let’s hear from U-Zyn Chua, the owner and primary developer of Ping.sg:
“Besides the cost savings, by having our infrastructure on the cloud, especially AWS, we are happy to know that we are able to acquire more compute resources as demand grows and we will be able to scale both vertically and horizontally very easily. On top of that, the availability of AWS in different regions all over the world and its CDN capability is also inline with our plans for further expansion outside of Singapore and into other regions.”
Talk to you soon,