|

No one shoe fits all… Defining the Enterprise Cloud Definition

By Steve Conner

“No one shoe fits all” is a good motto for businesses to adopt as they begin to look at the cloud. Smaller businesses have very different needs than large enterprises. The same is true for clouds. While we can all agree on the basic frame work of what a cloud is, the feature sets and functionality can differ greatly from solution to solution. But what exactly is an enterprise cloud? The wants, needs and requirements of a cloud solution for a large enterprise will be very different than those of a smaller company. What differentiates it from other cloud solutions? We wanted to explain our thinking and help build out the definition of what is an enterprise cloud.

Let’s begin by defining what is the “Enterprise Cloud”. Breaking this phrase apart, the first word Enterprise is easy to define. BusinessDictionary.com simply defines enterprise as a business or company, so for enterprise we will assume that our cloud will fall in the boundaries of the business.

Cloud, although the smaller of the two words, is a much more complex topic, but fortunately the National Institute of Science and Technology (NIST) developed a formal definition of cloud (NIST SP 800-145). According to the publication, there are 5 key characteristics that define a cloud:

  • On-Demand Self-Service – consumer can provision computing capabilities, such as server time and storage as needed without requiring human interaction with their provider.
  • Broad Network Service – Capabilities are available over the network and accessed through standard mechanisms.
  • Resource Pooling – Resources are pooled to serve multiple consumers using a multi-tenanted model.
  • Rapid Elasticity – Capabilities can be elastically provisioned and released, and in some cases, automatically, to scale rapidly outward and inward commensurate with demand.
  • Measured Service – Control and optimize resource use by leveraging a metering capability at some level of abstraction appropriate to the type of resource (e.g. storage, processing, bandwidth, and active user) (NIST SP 800-145).

The large, public cloud providers, such as Amazon, Google, and Softlayer are definitely capable of checking all of the cloud check boxes. When we step into the enterprise, things get a little fuzzier on how vendors can address the definition of being a true cloud inside the four walls of a business. Let’s break this down in terms of today’s hyper-converged players (e.g. Nutanx, VxRail, Simplivity) verses composable platforms (e.g. Cloudisitics):

  1. Broad Network Services are easy from a network connectivity standpoint.  These solutions attach to standard networks and are accessible via standard network access protocols.
    1. Hyper-Converged – YES
    2. Composable – YES
  2. On-Demand Self Service is a bit more complex.  When looking at the hyper-converged players, the only player in that footprint that addresses this requirement out of the box is Nutanix.  VxRail to achieve this, the consumer must invest in expensive vRealize software expansion.  Simplivity/HPE simply do not have this capability natively. 

The Cloudistics solution was built from the ground up with multi-tenancy in mind, so the attributes for RBAC are built in.  Using both physical and logical partitioning, each tenant can actively administer their allocated resources.

  1. Hyper-Converged – Nutanix Only
  2. Composable – YES – Cloudistics
  1. Rapid Elasticity is in the eye of the beholder here.  In the case of the hyper-converged players, the architecture is capable of spinning resources up and down, but due inability to address multi-tenancy, resources need to be dedicated to individual customers.  This not only causes the potential for resource islands, but it also means that there are tenant boundaries that resources cannot easily cross.  So, if you are looking it from the lens of a single use case, organization, or customer, then these architectures address the requirement.  If you are looking ate elasticity across multiple use cases, organizations or customers, then hyper-converged architectures fall short.

Composable architectures on the other hand, specifically Cloudistics, have multiple capabilities that address this requirement.  First and foremost, these platforms support end-to-end multi-tenancy – this will be discussed in greater detail below.  Second, because the resources in the Cloudistics platform are federated pools of resources – allowing storage, network, and compute to scale independent of one another – the platforms have greater levels of elasticity, allowing enterprise granular scaling of independent resources.  The resource federations give the enterprise operated complete control over the ebb and flow of individual resources

  1. Hyper-Converged – Partial
  2. Composable – YES – Cloudistics
  1. Resource Pooling is where things start to breakdown in the hyper-converged world.  The foundation of hyper-convergence is the merger of storage and compute resources into a single component – a node – that eliminates the complexities of operating complex storage area networks.  Hyper-converged architectures do not address or control the north-bound switch components that are part of the cloud architecture.  Because, HC platforms cannot control the network, they are incapable of managing tenant boundaries from the switch down.  

Another issue with HCI solutions revolves around their distributed file systems.  To achieve scale, the HCI platforms leverage distributed filesystems that span the breadth of their clusters.  Because the storage footprints reside on every node in the cluster, there is no way to actively separate tenants in the storage pools.

Composable architectures like Cloudistics incorporate all aspects of the cloud stack – network storage, and compute. 

By having full control of the network, composable architectures can segment tenant traffic from the switch port all the way down to the storage block. 

In addition to network segmentation, because the storage components are presented to VMs at the block level, specific storage segments can be either physically or logically separated, providing complete separation at the storage level.  Finally, because network micro-segmentation is built in at both network and hypervisor, using VNet/distributed firewall capabilities across both layers of the platform, the composable platform provides full segmentation of VMs within specific tenant boundaries.

  1. Hyper-Converged – NO
  2. Composable – YES – Cloudistics
  1. Measured Services need to be addressed both technically and financially.  HCI solutions have the ability to measure services in general, based on resources such as Virtual Machines, memory, and/or storage.  VxRail, by virtue of VMware’s VSPP program, does provide Manages Service Providers with the ability to consume software resources based on cost units associated with RAM.  At the time of the writing of this article, it is unclear if VMware will extend this offering to the enterprise.

Other HCI solutions, including Nutanix and Simplivity, do not have a commercial offering that allows consumers to consume resources using standard metrics or metering.  In August of 2015, Nutanix launched a program designed to provide Service Providers with a node based consumption capability, but as of 2016, it is unclear as to the status of this program and currently, there is no public mention of consumer based consumption models.

In January 2017, Cloudistics launched its utility consumption model.  The model allows customers to acquire standard, commodity hardware (Dell) and pay for resource consumption/management based using two axes,  similar to Amazon. 

The first axis, just like AWS Reserved instances, is based on time, the longer the commitment to use resources, the better the individual unit price.  The second axis measures resources.  Because the pools are separate, the consumer can commit to consumption for CPU/hours for compute and GB/month for storage.  Just like AWS, the consumer can choose to commit to nothing (On Demand) or fully reserve the resources of the platform.  As with all cloud providers, this offering is available to all consumers.

  1. Hyper-Converged – Partial – VxRail/Nutanix – Service Providers Only
  2. Composable – YES – Cloudistics

Using NIST 800-145 as the foundation for defining enterprise cloud, it is easy to make a clear determination on which technologies get the closest to being a true enterprise cloud. HCI solutions check the box on networks services and partially check the box on on-demand self-service, rapid elasticity, and measured services. By the nature of span of control and how HCI addresses scale, the standard HCI solutions fail to address resource pooling.

In comparison, composable architectures, can check all the boxes. Solutions like Cloudistics have complete control of all tiers of the cloud – network, storage, and compute – providing the ability to the resource pooling the entire box. With complete resource separation and federation, these solutions also can effectively provide the elusive metering capability that AWS has made the standard of clouding computing.

HCI vendors such as Nutanix claim to be a complete enterprise cloud solution. Now that we have a common nomenclature and a basis for definition for both enterprise and cloud it is truly up to you, the cloud consumer, to determine which solution truly meets the definition of enterprise cloud.