Enterprise website auditor Seopler.com has highlighted the challenges startups face around billing and support when building a business in the cloud.
The Dublin-based startup’s website-crawling software is used by enterprises such as Nissan Ireland to sniff out broken links, site redirects, invalid HTML markups and other elements that could adversely affect how highly their pages are ranked by search engines.
Once a site crawl is completed, a PDF is generated and sent to the user advising them of the corrective action they must take to improve the search engine optimisation (SEO) credentials of their website.
The compute resources that the Linux-based software carrying out the crawl requires to do its job varies depending on the size of the website and how many pages it contains, which is one of the reasons why Seopler decided to run the software in the cloud.
Also, the company rarely gets much prior warning about the size of the website it will be called upon to audit before a customer enlists its help, which makes it difficult to know how much compute power will be needed to perform its tasks.
“As Seopler crawls larger websites, the more data it needs to collect, and we used SQLite databases and kept those in-memory, so it’s very fast to collect that data when it crawls the site,” Graham O’Shea, the company’s CEO, told Computer Weekly.
Time is of the essence during the process, because the longer a crawl takes, the more likely it is to cause a level of disruption to the website undergoing the audit.
In the lead-up to the company’s launch last year, O’Shea set about finding a suitable infrastructure-as-a-service (IaaS) platform to run the software on, and initially opted for Amazon Web Services’ EC2 service.
However, the unpredictability of the size of each job undertaken by Seopler’s software created some unforeseen problems when it came to costs and resource provisioning.
“The problem with Amazon EC2 was we had to specify them before we did a crawl and, based on a million- page website, we would have had to commission an EC2 virtual machine with 8GB of RAM and 50-60GB of hard disk space before we ran the job,” he said.
“It started off and it looked quite cheap, but when the bills started coming in, we realised this isn’t going to work for us in the long term because you have to pay extra for support in the long term.
“Also, we don’t know the client’s website. They sign up for it at a switching level, and we don’t know what size their site is. So to do that for each client is really expensive. We tried it with Azure, and it was the same thing.”
While the company was working through these challenges, it had the added pressure of needing to have a workable demo in place to present to the crowd at the Web Summit Conference in Dublin on 3 November 2015.
Seopler was already using Elastichosts and its cloud servers to stand up some of its websites, so O’Shea and his team decided to give its container service a go.
“To be honest, it was there staring us in the face all the time, in that they had this new container service they had launched, and we didn’t really understand it,” he said. “It was only after going down the EC2 and Azure route that we worked out that it did exactly what we were looking for.
“We were able to set up the scripts to build everything in a day, whereas with Azure or EC2 it took a week or more, and it still didn’t do exactly what we needed it to do.”
The Web Summit deadline was met comfortably, and now Seopler, having worked through its hosting challenges, is looking to build out the software’s functionality and expand its operations overseas with the help of Elastichosts’ international datacentres.
“When someone performs a crawl, we want that done as quickly as possible, no matter where they are in the world,” said O’Shea. “So, if we get a subscriber in the US, we will automatically put them on a server being run out of the US. That will just accelerate the crawl and speeds things up because they’re performing it local to them.”