For some applications, the public cloud is only as good as the slowest connection to it.
Latency of cloud providers — the amount of time it takes for a cloud-based service to respond to a user’s request — is one of many critical factors that customers consider when choosing a cloud provider and monitoring their workloads. So which cloud provider has the best latency?
The real answer is it depends. Myriad factors influence latency: where the user connects to the cloud from; which cloud data centre the user connects to; which network provider is used; the route of the network traffic, among others.
During a 48-hour period in May network-monitoring company Cedexis compiled data across five leading IaaS public cloud providers to take a snapshot of latency in four regions of the country. This is what Cedexis found.
IBM’s SoftLayer cloud performed the best of five cloud providers. SoftLayer had the lowest latency in the US northeast and in the southwest. In the US northwest, Amazon Web Services had the lowest latency and in the Southwest Microsoft Azure had the quickest response times.
To compile the data Cedexis deploys a simple piece of JavaScript code in virtual machines of the five providers: IBM, AWS, Azure, along with Google Cloud Platform and Rackspace. Cedexis then pings those VMs using 10 network providers from multiple sites across the country to create four regional latency reports.
None of the five vendors in the report chose to comment on the data.
Mileage will vary
Individual latencies will vary based on different use cases. The data provided for this report represents an average of 10 different network connections to these public cloud providers. So, for example while IBM SoftLayer may have the lowest average latency in the Northeast, another customer may have a lower latency thanks to a more direct connection or by using a different network provider.
“There’s no one provider that’s the best,” said Pete Mastin, an evangelist at Cedexis, who compiled the data.
Other factors that could skew the results too: in most cases Cedexis uses the closest cloud region or availability zone to where the requests originate from. In the Northeast test, for example, Cedexis measured network connections to IBM SoftLayer’s Washington DC data centres. In the northwest region testing, Cedexis measured the connections to AWS’s US West region in Oregon. Cedexis used only one cloud region for Google however: US Central. Google has a US east region available but Cedexis did not include that region in this study. Google also has plans to launch a US west region this year.
This is also a point-in-time view of latency. These measurements could change minute-by-minute. Jim Davis from 451 Research says 48 hours – the amount of time Cedexis used for this report – is “a good starting point” for measuring latency. He cautions though: “It requires further digging based on your individual criteria.”
One factor
Latency is just one factor that can influence performance, he adds. The speed of the IaaS vendor’s infrastructure (virtual machine, storage and database response times) can impact workload performance. Application-layer issues can have profound impacts on workload response times too, and those are the end user customer’s responsibility, not the cloud or network provider.
No matter which cloud provider has the lowest latency, the closer you can get your end users to the cloud you’re using the better, says Mastin from Cedexis. If you have a lot of customers on the West Coast, then host the data there. “If you’re a provider, a gaming company, or any enterprise that wants to launch a service in the cloud, then know where you customer base is, that will help direct you to which cloud to use,” he says.
What else can be done about high latency connections to a public cloud? Network vendors like Level 3, AT&T and Verizon offer direct connections into public cloud providers. Vendors like Equinix provide collocation facilities with direct connections into the public cloud too. There are a variety of network mapping software tools available that allow organisations to optimise traffic routes from origin to end point.
Latency is just one factor users should take into account when choosing a public IaaS cloud provider, Davis says. Services offered, cost and future road map should also be considered when selecting a vendor. Davis says in his interactions with customers, he hasn’t found latency to be a determining factor in vendor selection, but it is one of many factors to consider.
IDG News Service
Ltd
