This is the last of a 4-post blog series comparing Amazon Web Services (AWS) vs. Oracle Cloud.
In this post, we discuss the Service-Level Agreement (SLA) provided by each vendor, and share key takeaways from the blog series.
Service Level Agreement
Service Level Agreement (SLA) is an umbrella concept that some people tend to attach meaningless business value to without taking the time to consider the true impact of the agreement terms.
A service level agreement (SLA) is a contract between a service provider (either internal or external) and the end user that defines the level of service expected from the service provider. SLAs are output-based in that their purpose is specifically to define what the customer will receive.
It means nothing to say that your SLA is XX%. What matters is that the SLA document clearly outlines the following:
- At what point will the customer begin to receive a discount because they could not access their service?
- What is the realistic likelihood of downtime?
- Can the customer application tolerate any downtime?
It is also important to remember that the SLA is a shared responsibility. The customer can keep their end of the SLA up by doing some of the same things the cloud provider does to provide redundancy.
Amazon says if their SLA drops below 99.5% they can give credit up to 10%, and if it falls below 99% they will issue a 30% credit for the time the service was down. But that commitment is not ironclad. They add this caveat: “At our discretion, we may issue the service credit to the credit card you used to pay for the billing cycle in which the unavailability occurred.”
Oracle provides a 99.9% SLA for virtual machines. At Amazon it is 99.95%. Amazon says they will make a “commercially reasonable efforts to make Amazon EC2 and Amazon EBS each available with a Monthly Uptime Percentage (defined below) of at least 99.95%.”
Regarding the Oracle Cloud database, Oracle says their SLA = (100% – what they define as “the average of the Error Rates from each one-hour period in any given month of the Services Period).”
Now, let’s put this into perspective. If the vendor provides an SLA of 99.9% uptime for any unit of time, the question to ask is: After how many units of time will the probability of the system going down equal to certainty, i.e. 100%?
At Oracle, with an SLA of 99.9%, the chance of it being down is (1 – 99.9%) = 1/1,000 at any given time. So after 1,000 units of uptime one expect 1,000(1/1,000)=100% chance of at least 1 outage.
The Amazon SLA is given in months. If you look at 1 month as equal to 43,400 minutes, you could expect downtime not to exceed 43.4 minutes per month. So almost an hour, which is not negligible. But neither it is really too important given the larger picture.
With data and geographical replication there is not going to be much chance for the entirety of Amazon or Oracle to go offline at any one time. It’s most likely that a customer will knock an application offline themselves due to errors in deploying code or problems with the ISP. Customer web servers will generate Error 500 on occasion. But that Error 500 is the fault of the customer and not the cloud vendor’s, since they do not run the customer’s applications.
SLA: Customer Responsibility
Lastly, clients cannot be the weak link in maintaining the SLA with their cloud vendor(s). Client organizations should do what Amazon and Oracle do to protect the network by maintaining multiple ISPs and distinct physical facilities, ensuring constant connectivity from the client site. Internet customers can protect themselves from problems in global networks by signing up for a Content Distribution Network (CDN). Customers who choose to go with Oracle will have to sign up with Akamai or another vendor as they do not offer this solution. Amazon does.
What’s the Takeaway?
With so many vendors in this space, one could argue that cloud services have become a commodity product. An Oracle-centric company might pick Oracle if they are using lots of Oracle SaaS products. And since Oracle makes sales by telephone it might be possible to negotiate a discount.
Amazon is going to be an easy sell to the CIO and CEO since Amazon has such a long track record, including keeping Netflix running for so many people streaming Breaking Bad.
As to prices, you’re going to have either do some pilot testing, talk to other companies, or try to wrestle some actual numbers out of your Oracle account manager to get the bottom line on that.