massimhokuto - stock.adobe.com
How Sporting Group tackles the Rugby World Cup
Sporting Group has readied its system monitoring to ensure the technology is able to cope with the huge demand for bets during the Rugby World Cup final
Sports betting technology firm Sporting Group experienced a 43% increase in activity around the England vs New Zealand Rugby World Cup semi-final on Saturday 26 October, compared with the quarter-final week before - and the company is anticipating a further boost for the final, which takes place on 2 November.
“We expect to have a similar level of increase around the final,” says chief technology officer Peter Wallis.
Sporting Group provides betting services both to consumers and bookmakers. Wallis, who joined the company in 2017, says that from a technology perspective, the nature of the business means IT is focused on the velocity of data, low latency and predictable spikes in usage.
Wallis says the company implemented a log management and analytics software as a service (SaaS) tool from Sumo Logic nine months ago to support the peaky demands systems experience when people place bets during sports events.
Since Sumo Logic runs and scales elastically in the cloud, Wallis says: “We expect to cope with the increase in activity automatically without any impact on our ability to proactively spot issues that affect customers.”
In its B2B business, Sporting Group provides services for 50 bookmakers, taking prices across 23 different sports. “Football is half of our business,” he says. “At 3pm on Saturday, when the English Premier League kicks off, we see data going up by 200 to 300 percent over the 90 minutes of a game.”
From his past experience of working at ETX Capital and Camelot, Wallis describes the data requirements very similar to that of a trading platform in banking.
“We rely on a third party feed to give us real time data,” says Wallis. “If you are slow, people will pick up on that.”
Impact on infrastructure
From a systems perspective, each sport will have a different impact on the IT infrastructure at Sporting Group.
“All sports behave slightly differently,” he says. “You’ll see variations in patterns depending on how that sport operates. For instance, in basketball, which is fast-moving, lots of points are scored, which means there are lots of frequent, small updates. In football, there is a lot more contextual data around where the ball is on the pitch, whether it is in an attack, a corner or a throw-in.”
The company’s move to SaaS-based IT log analytics began seven years ago, with Splunk’s cloud-based platform, to help it manage and monitor systems more effectively and reduce downtime.
But following workshops with Splunk in October 2018, Wallis says he realised that Sporting Group needed another SaaS provider due to the restrictions of its existing Splunk contract.
“Our previous commercial deal with Splunk was very good because they [Splunk] were incentivising people to move to Splunk Cloud,” he says. “It was a low margin deal. Everything they spoke to us about was an additional cost. In some cases, the cost of these additional tools would have meant a trebling of our annual fees.”
Handling peaks without cost hikes
Sporting Group’s head of IT infrastructure suggested the IT team have a meeting with Sumo Logic. This took place in November 2018, two months before the Splunk contract was due to end.
“After a two hour meeting we decided Sumo Logic gave us enough out of the box functionality compared to Splunk. The primary reason was that its commercial model was more attractive to a peaky business.”
With the old Splunk contract, Wallis says Sporting Group needed to provision and pay for peak usage, even though it only experienced this peak four times a month.
“In both Splunk and Sumo Logic, users are billed on the amount of data ingested into the monitoring platform. With Splunk, we had to buy the maximum to cover our peak usage.”
However, with Sumo Logic, Sporting Group only pays for the average amount of data ingested into the platform over the whole month, which suits a peaky business. “We kind of felt quite annoyed paying for something that we only hit four times a month,” says Wallis.
Another reason for switching from Splunk is that a lot of features that are standard in Sumo Logic are an additional cost in Splunk.
For Wallis, one of the most useful features included in Sumo Logic is support for dynamic thresholds. In his experience, one of the challenges in trying to monitor peaky systems is determining when it’s genuinely running slower than it should.
Read more about log management and analytics
- Log analytics tools are boosting real-time capabilities to monitor application performance, and users are in the market for tools that best slice, dice and store log data.
- The path to more resilient, bug-free and available IT deployments is hiding in plain sight, in log data. Parse short- and long-term data these three ways to improve operations.
If during peak usage a particular calculation should take five seconds, the threshold may be set to 10 seconds. However, this threshold needs to be much lower during non-peak days to identify that the system is running slower than it normally does.
“If the threshold is set too high, you will never detect a problem that happens on a Monday afternoon,” says Wallis. Dynamic thresholds are based on using time series analysis to understand different “normal” usage patterns, to identify anomalous behaviour, such as when a system runs abnormally slow on a given day.
Going forward, Sporting Group’s strategy is all about being cloud first. Wallis says monitoring elastic on-demand IT infrastructure will be a key business requirement.
The company already owns two systems on Google and is working with Sumo Logic to ingest Kubernetes log data in a meaningful way.
“Using Sumo Logic, we can consume that data and see what challenges exist or where we may need to take action,” he says. “This ability to scale up and manage data as it is created, alongside approaches like dynamic threshold alerting, means we can get an accurate view into our applications and how we are performing, and we can do the same for the third-party applications we consume data from.
“We can be ahead of any potential problems in our stack – or in that of our partners – before they impact customer experience,” says Wallis.