Under The Hood of HP Cloud Services SLAs

Under The Hood of HP Cloud Services SLAs

View photo

Under The Hood of HP Cloud Services SLAs

With the launch of HP Cloud Compute into General Availability (GA), there has been a lot of talk about the relevance or real value of IaaS Service Level Agreements (SLAs). At HP Cloud Services, we believe that SLAs have to be a critical part of any credible cloud solution. Before we take a service into GA we make sure that we have an SLA that we are ready to stand behind so enterprises and customers of all sizes know that they can count on the services they are using. In this post I will discuss how we approach SLAs and how we design and build our services to achieve the commitments we make in our SLAs.

The Logic Behind SLAs

An SLA is really a sign of trust between a service provider and a customer and as such it is important that the SLA is a true representation of what a customer can count on from a service. Therefore, as part of our decision making process for the functionality of a new service, we  consider what the SLA for that service will look like. We talk to existing and prospective customers, we look at what others are doing in the market and we think about the overall vision for the service. Once we know the direction that we want to take with an SLA, we create some guiding principles which we provide to our engineering teams for them keep in mind as they develop the service.

The HP Cloud Compute SLA Defined

Since our launch, we have received several questions about the specifics of the HP Cloud Compute SLA, so I want to briefly further explain the details of this SLA.  To summarize the HP Cloud Compute SLA in a single statement: we commit  to customers that if they have an instance that fails they will be able to launch a replacement instance in the same Region or they will receive a credit in accordance with the provisions of the SLA.   Here are a few points that are critical to understanding the details of the full SLA (which can be found here):

The SLA covers failures of even a single instance

Downtime starts when you issue the first call to launch a replacement instance
You do not have to be running instances in multiple AZs prior to the downtime
You do have to launch or try to launch a replacement instance in a different AZ
Downtime continues until your replacement instance becomes accessible
Delivering on SLAs

As I described above, delivering on an SLA is a lot more than just putting some words on a piece of paper (or a webpage). It starts when you first begin to design and architect your service.  The SLA you can offer depends on the technology stack that you select, the hardware you order and the configuration you run. It also depends on your organizational discipline and operational excellence - two areas that are a big part of HP’s DNA.

For HP Cloud Services we selected OpenStack® as the base technology platform, but that was only a beginning. We then built our cloud using HP hardware and made architectural decisions to provide maximum availability and isolate points of failure. As an example, we built each of our Regions with three distinct Availability Zones (AZs).  To harden security we incorporated the HP Security Intelligence Platform which includes a suite of HP security products such as ArcSight™, Fortify™ and TippingPoint™.  Additionally, we extended the capabilities of the OpenStack Keystone Identity Service in several areas including encrypted tokens, stronger role enforcement and optional multi-factor authentication.

For HP Cloud Object Storage, we leverage self-healing algorithms to continuously monitor the system, detect issues or corruption and make repairs automatically. In addition, the service is spread across three AZs per Region with replicas of your data stored in each AZ. For the HP Cloud CDN we partnered with Akamai, leveraging their massively distributed and redundant worldwide network to eliminate single points of failure. Finally, for HP Cloud Compute we put extensive time into making every component in the control plane highly available so we are able to seamlessly upgrade components and successfully survive hardware failures.

The Road Ahead

Over the next few months/quarters you will see more announcements from HP Cloud Services around SLAs.  Although we are proud of where we are today, we know there are still opportunities to further enhance and differentiate our offerings.  Stay tuned for more announcements as we bring additional services into General Availability and start moving from service-oriented SLAs to solution-oriented SLAs (more on that in a future post).
View Comments (0)