Making sure people get their morning coffee is a cornerstone of modern civilisation. Lavazza has IOT-enabled thousands of coffee kiosks around India to allow real time monitoring of supplies and machine health. They use Sunlight-based edge microdatacenters to collect data in each region, aggregating up to a core datacenter running Splunk on Sunlight to analyse the IOT data without the performance problems virtualisation normally creates.
reduction in infrastructure costs vs VMware
Collection of IOT data at the edge brings several advantages. Where network connectivity is patchy or unreliable, it enables a more resilient store-and-forward model for the data. Where significant volumes of data are being collected, it can reduce the volume of data sent to the core datacenter or cloud by 140x. In Lavazza’s case - collecting data at the edge improved resilience to ensure that the data analytics were 100% accurate. Lavazza deployed 5 microdatacenters based on Sunlight NexVisor nationally, with aggregated data combined in a core datacenter Splunk deployment. Due to Sunlight’s minimal footprint, edge servers can be based on low-power processors, keeping infrastructure costs low.
"Network availability was patchy, meaning that collected data from the IOT sensors could be incomplete and inaccurate. Sunlight made it easy to deploy small microcenters at low cost which has fixed the problem."
Splunk is a data analytics platform which processes huge volumes of data on a continuous basis. Lavazza estimate that 7-8 day data retention of coffee kiosk data generates around 100Gb of data. To process that, a typical Splunk cluster on VMware requires 7-8 Splunk nodes. With Sunlight - due to improvement in IOPS and increased memory bandwidth - that server requirement is now reduced by 50%. This represents not just a basic infrastructure saving, but more data can be processed more quickly - improving the user experience when generating reports.
"Fundamentally, Splunk needs good IOPS. Splunk made lots of improvements in multithreading in the product - which helps with performance a lot – but at some point it has to hit the disk. 80,000 IOPS vs 1M IOPS per VM makes a significant difference in performance."
Managing the complete analytics and monitoring platform through a single pane of glass reduces operational complexity and costs. The Sunlight Infrastructure Manager provides a view across the core datacenter cluster as well as the 5 edge microdatacenters, enabling easy hardware commissioning, deployment and migration of workloads, unified backup and template management. Coupled with the ease of use of the Sunlight NexVisor stack itself, this means a far smaller NOC team is required to keep the infrastructure running.
"It takes my team members about a month to become proficient with VMware vCenter. With Sunlight - they’re fully up-to-speed in about a week."