By: Caroline Donnelly
The big four cloud firms are set for price war over serverless computing, suggests 451 Research
According to 451 Research’s first foray into analysing the pricing and cost models of the serverless computing market, IBM’s offerings in this area generally work out cheapest, until certain Microsoft Azure configurations are taken into account.
“When users’ memory requirements match pre-defined size allocations, we find IBM is cheapest for 0.1-second duration scripts and Azure is cheapest for 10-second duration scripts,” states 451 Research’s Economics of Serverless Cloud Computing report, which is authored by Owen Rogers, research director of the firm’s Digital Economics Unit.
“IBM also has the smallest granularity, which gives it a major cost advantage across many scenarios by allowing users to choose exact memory requirements without rounding up, thereby reducing waste capacity.”
That said, the cost of using serverless computing resources will vary from scenario to scenario, making it difficult to make a definitive declaration of whose offerings work out cheapest. “Enterprises need to do their homework and work out the costs of their own likely deployments; risk-aware enterprises should also assess how expenditure will change depending on growth or shrinkage,” said Rogers.
All of the big four, including Amazon Web Services (AWS) and Google, offer free serverless computing tiers, which Rogers said should fuel adoption of the technology by essntially allowing developers to try before they buy.
“Freemium serverless offerings from the four big cloud providers are already fuelling the growth of serverless services by stimulating experimentation and helping enterprises gain skills. This could lead serverless to be the next cloud price war battleground,” he said.
For most new applications, it works out cheaper to run them in a serverless environment, rather than rely on virtual machines (VMs) or containers.
“The total cost of ownership of serverless tends to be lower than VMs, even when the VM is hosting containers, because there is no need for developers to provision, configure and manage the infrastructure,” the company said in a statement.
“As an example scenario: when a serverless function is active for just three-quarters of the month, it only takes a 10-minute saving in operational overhead for serverless to beat virtual machines on TCO.”