AWS's policy of requiring users to run services across multiple AZs costs users more money than if applications are running in a single AZ. "Every AZ that a customer chooses to run in effectively imposes a cost," Leong writes. HP's SLA, which requires all of the AZs to be down before the SLA applies leaves customers vulnerable, she says. "Most people are reasonably familiar with the architectural patterns for two data centers; once you add a third and more, you're further departing from people's comfort zones, and all HP has to do is to decide they want to add another AZ in order to essentially force you to do another bit of storage replication if you want to have an SLA."
The SLA requirements basically render the agreements useless. "Customers should expect that the likelihood of a meaningful giveback is basically nil," she says. If users are truly interested in protecting their systems and received financial compensation for downtime events, she recommends investigating cyber risk insurance, which will protect cloud-based assets. AWS has recently allowed insurance inspectors into its facilities to inspect its data centers for such insurance claims, she notes.
A strict requirement of service architecture isn't the only aspect of the SLAs Leong takes issue with. They're unnecessarily complex, calling them "word salads," and limited in scope. For example, both AWS and HP SLAs cover virtual machine instances, not block storage services, which are popular features used by enterprise customers. AWS's most recent outage impacted its Elastic Block Storage (EBS) service specifically, which is not covered by the SLA. "If the storage isn't available, it doesn't matter if the virtual machine is happily up and running it can't do anything useful," Leong writes.
Leong does note that AWS has voluntarily refunded customers impacted by major downtime events even when the SLA did not require it.
Not all IaaS cloud SLAs are as bad as AWS and HP, she concludes. "The norm in the IaaS competition is actually strong SLAs with decent givebacks, that don't require you to run in multiple data centers," she writes. Data Dimension, for example, has a per-VM SLA with 100% uptime. That compares to a 99.95% uptime guarantee from AWS and HP, which only kicks in after at least five minutes of an outage. AWS and HP also have caps of how much of a percent of a customer's bill can be refunded during a downtime, whereas Dimension Data will refund up to 100%.
HP did not immediately provide a comment in response to Leong's claims.