Amazon Web Services’ (AWS) Snowball Edge and Microsoft Azure Stack both aim to bring a subset of the public cloud to the private data center. While both have the same high-level objective, they are very different approaches.
On the surface, the two solutions only have faint similarities. However, both solutions are strategic to the two cloud providers’ on-premises public cloud expansion.
Public cloud at the edge
One of the largest stumbling blocks to public cloud remains system locality. There are still several use cases where organizations require micro-data centers at the edge of the network. The Internet of Things (IoT) proves a popular argument for having local compute, as the amount of data generated by IoT devices favors on-premises data processes. Another use case is integration with legacy systems–network latency to public cloud may prove too great an obstacle to place dependent workloads in the public cloud.
Enter on-premises public cloud, which aims to resolve the technical and business challenges of the edge. The basic concept of on-premises public cloud is to provide local cloud service with management existing in the cloud. Today’s on-premises public clouds are only a forethought of what the future may bring. These solutions consist of hardware dedicated to a single customer and managed by a public cloud vendor. With future advancements in software-defined networking (SDN), public cloud providers create resource pools using customer premises equipment. This is not too dissimilar to Comcast’s leveraging residential broadband for public hotspots.
AWS introduced the original Snowball as a data transfer appliance. At the AWS Re:invent 2016 conference, however, Amazon announced the Snowball Edge. The Snowball Edge is a 4-node scale out appliance that acts more as a permanent installation instead of a one-time data transfer appliance.
SEE: How Amazon is moving closer to on-premises compute with Snowball Edge (TechRepublic)
Snowball Edge speaks to a trending public cloud feature known as serverless in that it supports AWS Lambda code. As data objects are written to Snowball Edge, user developed code can process that data. For example, in the IoT use case, an end user may identify temperature thresholds for an IoT sensor. As sensors write logs to Snowball Edge, the readings are compared against the threshold using Lambda code. If the temperature reading is out of range, Lambda could kick off a process by calling a script that may run on an external system.
Microsoft’s Azure Stack is more of a traditional hyperconverged infrastructure (HCI) solution, sold by the company’s OEM partners. I recently had Azure Stack’s chief architect Jeffrey Snover on The CTO Advisor Podcast where he pointed out that Azure Stack is Microsoft Azure inside of the customer’s data center. While AWS Snowball Edge focuses on data and data processing, Microsoft seems a bit more ambitious with Azure Stack.
Azure Stack runs several of Azure’s PaaS and IaaS services. In theory, a customer could replace an entire edge micro-data center with Azure Stack. Azure Stack is expandable to 12-nodes and, according to Snover, an Azure Stack customer can run as many as 400 large cloud virtual machine instances on a fully-expanded Azure Stack deployment.
Both Azure Stack and Snowball represent the initial concept of providing public cloud computing at the edge. Customers gain the advantage of outsourcing infrastructure management to a public cloud provider, but meet the technical and business requirements presented by edge computing. I’d closely watch both Azure Stack and Snowball Edge as both of these cloud providers expand the capability and reach of these systems.