Sholto Macpherson
Smarter Writer

Sholto Macpherson builds business models and tech stacks for cloud accounting practices on DigitalFirst.com

Cloud scaling spares you red faces and angry customers

Sholto Macpherson
Smarter Writer

Sholto Macpherson builds business models and tech stacks for cloud accounting practices on DigitalFirst.com

Highlights
  • Website outages due to traffic spikes can mean lost revenue and reputational damage.
  • Cloud scaling can help businesses deal with traffic fluctuations, avoiding outages and costly idling.

Any business with a digital presence needs to get serious about cloud scaling – to avoid the embarrassment and irate customer response prompted by an underperforming or unavailable website.

One of the ironies of having a website is that overwhelming success can have fatal consequences.

two people working on a laptop together Website outages can cause frustration for customers and loss of revenue.

Think of a bricks-and-mortar clothing store like Country Road, running a highly effective promotion. Success in a physical sense is a store packed with buyers fighting over clothing racks while other customers line up on the street to get inside. The energy and excitement of the event would generate curiosity, envy and FOMO in passers-by, the success of the sale converting into even more successful marketing.

But what would an equivalent scenario online look like?

When buyers overwhelm an online store, success can turn into failure. Not just lost sales, but reputational damage that could result in permanent loss of customers.

Stories abound of e-commerce websites grinding to a halt after a deluge of visitors. In our example, customers rushing to a Country Road website expecting to pick up a bargain would instead find they couldn’t access the site – their browser showing a server error instead or, worse still, a site that only half-works. Pages load so slowly that a customer gives up in frustration before they click the pay button on their shopping cart.

“The biggest loss is brand damage,” says Tim Otton, General Manager Cloud Strategy and Platforms at Telstra. “You can recover lost sales, but the brand damage for digital natives is very significant. They’re not tied to bricks and mortar and traditional brands, they’re tied to experiences online.”

Digital natives (loosely defined as consumers born from 1980 onwards) have high expectations of online interactions. Any type of business must not only have a website – well designed and easy to browse – but that website must also be available 24/7. The digital natives’ point of comparison is Facebook and Google, massive websites that deal with billions of users. If they can stay up despite unimaginable traffic loads, why can’t a local store?

“Anyone running a digital presence needs to be up there with the best. Having [a poor user] experience or an outage is almost unacceptable,” Otton says. “They don’t have an experience of Facebook or Google having an outage.”

Product launches are particularly sensitive times for e-commerce websites, particularly if a marketing campaign creates a lot of buzz and expectation. The launch of Nintendo’s latest games console, Classic Mini NES, crashed the website of electronic games retailer EB Games two days in a row when a flood of eager games fans all logged on at the advertised time.

Frustrated shoppers slammed the retailer on Twitter when they were unable to pre-order the console. Rubbing salt in their wounds, by the time EB Games’ website had recovered, the Mini NES had sold out and scalpers were selling them on Gumtree for three to four times the retail price.

EB Games hadn’t adequately scaled up their infrastructure in readiness for the massive spike in customers hitting the website at the same moment – approximately fifteen times the usual daily number of page views. As a result, the retailer may have lost out on a huge number of sales.

Spikes in traffic to websites are becoming more common and more unpredictable as businesses move to sophisticated marketing tactics. Advertising campaigns across social media and other platforms are increasingly personalised. The ad will refer to an individual’s interests or preferences, or contain a specific offer for smaller and smaller segments.

“You’re actually targeting more refined value propositions,” Otton says. “They tend to hit very well, or if they miss, you might not get a big response or [get] a very small section of market.”

Marketers want the flexibility to tailor messages and offers, and split test variations to see which will get the higher conversion rate.

One factor driving unpredictability in traffic is the contextual impact of a marketing campaign. If a news story breaks on Facebook and a marketer inserts a highly relevant offer, then that can drive a high number of conversions. This may be intentional or coincidental, and may occur so quickly there is little time to prepare for the response. Offline factors that can boost traffic include the time of day, time of year or more random variables such as weather.

Companies running websites on their own servers are not always able to prepare for high spikes in traffic. So how do Google and Facebook do it? Instead of working on a ratio of servers to traffic and ensuring it’s within the capacity of their data centre, they build a pool of computing and storage power. 

The concept of cloud scaling gives a business the ability to instantly ramp up or down the number of servers required to meet the traffic. This eliminates the scenario where a company either wastes budget on extra physical capacity and servers sitting idle “just in case”. It also eliminates the reverse – the e-commerce website grinding to a halt during peak sales season.

The beauty of cloud scaling is that it can be automated as soon as specific parameters (latency, processor usage) are met. When traffic spikes, cloud providers can spin up the necessary firepower with no human intervention required.

“When utilisation or latency has reached a certain point, the cloud service fires up another server so that the e-commerce app doesn’t degrade in performance,” Otton says. “They don’t require planning or operational support to do that, it happens automatically.”

A cloud platform can also help businesses scale in another way – geographically. It is easy to copy an application and spin it up on a new server to support a new office.

Civica, an IT consultancy that works with government agencies, councils and libraries, helped an education customer set up a management system for schools in the Northern Territory. It took eight weeks from the time of signing the contract to set up the system, despite the remote locations of the schools in question.

“If you’re dealing with remote locations like the Northern Territory, it’s expensive and time-consuming,” says Daryn Vanstone, Civica’s marketing director. “If things go wrong, you have to wait for them to be fixed. With cloud you can [respond] quickly and avoid the cost of sunk IT. The customer can be up and running very quickly.”

Cloud scaling requires a different mix of skills to conventional data centre operators. Software architects examine how the application and supporting infrastructure will function under load and set parameters for adding or reducing capacity as required.

“Not every application has the ability to scale in this way,” Otton says. “You need to have the ability to manage cloud platforms yourself or find someone to do it for you.”

Learn about why you should design for failure in the cloud

Find Out More