Comment Hybrid cloud is so yesterday – multi-cloud is where it’s at. Spreading your cloud apps between different providers is now becoming a trend.
Four in five firms were using multi-cloud services last year, according to cloud firm RightScale.
Six in 10 UK companies used more than one provider, said Adapt a year earlier.
There are good reasons for this. Some firms want to run instances in different jurisdictions for data sovereignty reasons. Others might fear being locked into a single provider, or use multiple providers to combat latency issues. Some might crave redundancy enough to straddle multiple IaaS providers.
Another driver for multi-cloud environments is that different providers can often provide optimal services for specific workloads, explained Gordon Davey, UK cloud strategy director at Dell.
“Platforms chosen for a specific purpose will often have less over-provisioning, and will usually out-perform a generic multi-purpose solution,” he said. “The aggregated cost and performance benefits of using the right platform for the right workload can often make a very compelling business case.”
But whatever their reason for using multi-cloud, companies often aren’t doing it particularly intelligently. They don’t manage it in a uniform way, and will find themselves maintaining different tools and juggling workloads across different cloud-based services. Surely this digital plate-spinning was something that cloud computing was supposed to avoid?
The shortcomings are partly down to a lack of sophistication on the users’ part, but also to disparities in the way that cloud service providers themselves handle their own metrics. It’s built in from the very start, according to the chief technology officer of capacity planning-as-a-service company Sumerian Peter Duffy.
“All the cloud providers sell you compute instances in different sizes. So there's complexity right there from the get go,” he said. “If I move a workload from Dell to Amazon, then how many of these should I be buying?”
Another problem is that cloud SLAs aren’t tailored to keep providers accountable for how fast your instances are running. Rob Greenwood, technical director at Steamhaus, a consultancy that builds and runs cloud-based applications, told The Register: “The types of SLAs that public platform provide aren’t about performance, they're just about availability.”
One way around this problem is to use a cloud broker to wrangle your multi-cloud environment for you. These services sit in between the customer and a variety of cloud service providers, aggregating the latter into a single set of services.
NIST’s Cloud Computing Security Reference Architecture (500-299) (PDF) defines two types of cloud broker: the business, and the technical. The former handles tasks on the contractual side, managing paperwork, billing, and arbitration between the two sides. The latter is the one that can pull together a coherent performance story and hopefully remove some of the headaches for your ops team.
Kalyan Kumar, senior vice president and chief technologist at IT services firm HCL Technologies, calls brokers “simplifiers.”
“Brokers help in unifying the different services, standardising the implementations and taking care of governance, risk and compliance,” he said. “They can also provide support and expertise that may not necessarily exist within the organisation, such as managing the use, performance and delivery of cloud services.”
Cloud broker firms are a big opportunity, but it seems the earliest opportunities are for those who are already systems integrators or telcos and who want to offer packaged services based on the cloud.
The blurred line between the cloud service broker and integrator could be an asset. The broker that has an integration background could also potentially provide advice on how to engineer applications and network infrastructure for the cloud, improving the performance of applications for the cloud.
Such brokers probably are unlikely to work in a very detailed way, doing things such as balancing workloads dynamically between different cloud environments. It’s a utopia to imagine that a cloud customer will be shifting containers around on a minute-by-minute basis between AWS and Rackspace, say.
Rather, they will employ different services according to needs and employ a uniform way to find, access, monitor, manage, provision, and govern.
A lot of these needs take the form of particular managed services such as application hosting, where one provider may have a better offering than another. Customers may fire up a deployment for a particular project with a specific cloud partner and then scale it back down again. Doing that kind of thing smoothly, and providing a single, balanced view of what’s happening across the infrastructure, is where the technical broker comes in.
Cloud brokers are in their infancy, however, so be careful: It still isn’t clear in many cases who owns the SLA and or how SALs are enforced. Such things must be ironed out before a contract is signed – it’s classic managing an outsourcer or supplier relationship.
Then there are the internal wrangles that a cloud broker relationship can create, too. “There's a struggle between company culture and IT strategy, because using a broker means that lots of different groups get access to what they want rather than on an allocated basis,” according to 451 Research vice president William Fellows. Managing internal procurement can be a tricky political problem when working with a broker-based service.
There are alternatives. The IT department could decide to be its own broker, balancing the performance in a multi-cloud environment using a cloud portfolio management service like RightScale, for example. Services like this could help aggregate your view across different cloud providers, and give you a platform to manage them from a single point. You’ll still have to manage all of the relationships and paperwork, though.
No matter what, however, the growing complexity and sophisitication of cloud deployments means intermediaries are becoming real – needed to tackle both business and technical challenges. It’s just looking like a question of whether you choose different companies to provide each kind of service, or throw in their lot with a single provider. At least then you can move all those spinning plates to a single stick. ®