Iguazio builds Google-like Outpost for IoT retail edge

Unlike Amazon’s Outpost and Azure Stack,  Google has no public cloud on-premises presence. Iguazio says it can provide one for Google customers.

The Israeli start-up claims superstores with multi-branch retail outlets want to use public cloud facilities but won’t use Amazon because it is a direct competitor through the Whole Foods acquisition. 

Iguazio is working with Trax, a retail edge app vendor, and integrating Google Cloud Platform to provide an Amazon Outpost-like service. It says the Iguazio/Trax/GCP combo delivers public cloud benefits that enable retailers to optimise in-store operations across the estate.

Yaron Haviv, Iguazio founder and CTO, said: “Retailers who left AWS due to the Whole Food acquisition are a target market for Google and this solution enables them to build intelligent stores and compete with Amazon Go.”

Making Trax

Trax ranks in the top 25 Fastest Growing Companies on Deloitte’s Technology Fast 500 list. The company “monitors, predicts and optimizes store-and-field performance in real-time to improve on-shelf availability, optimize click-and-collect processes and modernize the shopping experience”.

In other words it uses video and still cameras and software to say where customers have gone in a store and what they have bought. To do this in real-time is compute+store intensive.

Trax CTO Yair Adato said: “We recognised that we needed an edge-to-cloud solution that…allowed us to focus on our application, versus the management of our infrastructure.”

This is where Iguazio comes in. In essence it provides the glue between Trax and Google Compute Platform. Iguazio supplies the storage platform and infrastructure, and its software integrates with store-level workflows and apps for local analytics. Also, Iguazio enables data movement to the cloud where it integrates with federated store-level workflows.

Trax-generated data points include videos, images of products on shelves, from cameras and other devices and are stored in Iguazio and processed by Trax software to turn analogue imagery such as known products on shelves into digital information,

This data is then used to track real-time store performance at product level and to analyse performance with different pricing and shelf placement options, and in turn drive shelf re-stocking.

The Trax software analyses tens of thousands to hundreds of thousands of data points, depending on store size, which are stored in Iguazio databases on Iguazio array hardware. This is linked to Google Cloud which receives selected data for further analysis looking across a set of retail branches, and also using Trax software and other applications.

Iguazio storage and analytics

Israel-based Iguazio is not your usual storage array or software-defined storage company. The company’s technology holds metadata in quick NVMe flash storage providing fast data access for Big Data analytics apps. Iguazio launched its analytics software and storage hardware in 2016. It provides NVMe-supporting storage arrays with high IOPS and low latency, but these arrays are in place to get its software running fast, the hardware being subservient to the software.

A diagram shows Iguazio’s ideas in a retail environment.

Iguazio cloud-native integrated cloud to edge (ICtoE) scheme. The Google Cloud ‘Outpost’ is our terminology

The white arrows in the diagram shows selected and filtered data flowing upstream to the Google cloud.

The grey arrows shows control commands and containerised software, developed centrally, cloud-native and orchestrated by Kubernetes, flowing out to the stores, where it executes. Machine learning models, developed centrally, are also shipped out to the stores to be used in the analytic routines there, as are serverless functions.

These analytics routines determine simple things, such as how many packs of 300cl Cola were sold per shelf space unit per hour, to more complex points. At which shelf level should the Cola be presented, and how many cans should be in a row on the shelf? How do end-of-aisle notices affect sales? Is it more profitable to sell five cans of branded Cola or ten cans of own-brand Cola? How effective are special offers?

The Wisdom of Clouds

Retail has become an intensely data-driven environment.  Bricks and mortar retailers need to offer the best shopping experience they can, to match Amazon, and to be as fast-moving and flexible as Amazon in their stocking, pricing, delivery and operational measurement and analytical operations.

Iguazio asserts that at a central level this can mean using the public cloud because it can provide the speed and flexibility needed.

But operators can’t maximise individual store-level operations with just a central public cloud, because there’s too much data to send upstream to the cloud, and decision-making takes too long.

They need mini- or micro-data centres in each store, for real-time store-level optimisation, and a central core data centre operation, which could be in the cloud. For example, Google’s in this case. It is used for business optimisation across the bricks and mortar estate, and for application and machine learning model development to continually improve individual store operations.

Superstore as IOT edge device

Iguazio would have us think of a superstore branch as an Internet edge location, needing real-time optimisation of its activities, with analysed and filtered store data fed to a central location for analysis and business optimisation across an entire retail branch estate, with hundreds of stores.

Iguazio claims that the Google Cloud Platform (GCP) is the best central location and says it can effectively provide an on-premises presence for CGP with data upload for central processing. Further, the Iguazio/Trax offering has a GCP presence and so provides a common environment across the store branches and the central, head-quarters operation.

The GCP advantage is based on it being a public cloud, and so scalable, elastic, with usage-based pricing and cloud-native SW development orchestrated by Kubernetes and supporting serverless computing.

Retail branch system

A retail branch Iguazio system has 4 Intel servers, each with an additional GPU for image and AI processing, and 24 x NVMe disks in an off the shelf 2U enclosure. This has more than 100 CPU cores and 100TB of usable data, he says it performs like a rack of traditional servers.

“This solution is way faster and denser than Amazon Snowball, and unlike Snowball can run any service on the embedded Kubernetes, it supports 100x faster serverless functions with Iguazio Nuclio and an extremely fast flash-optimized database and file system.” 

Nuclio is Iguazio’s own serverless computing technology.

Containers and Kubernetes are the best pairing for application development and deployment because applications can be developed and distributed quickly. 

Iguazo says the combination of its managed platform and Google Cloud services enables the collection and analysis of large volumes of data at the edge, while using Google Cloud and applications there, such as Trax, for deep learning, AI, data aggregation and central control. 

Aparna Sinha, Group Product Manager for Kubernetes and GKE, Google Cloud, sings off the same hymn sheet: “We are excited to collaborate with Iguazio to deliver a solution that enables real-time analytics of store data, all centrally managed from Google Cloud.” 

Shopping for the storage lesson

The storage lesson here is that complex edge environments need much more than basic commodity storage hardware and software. At retail branch level the computing environment is dense, intricate and rich. Real-time responsiveness is needed and only; this is Iguazio’s message, only specialised   database storage software and hardware can support the retail edge application and sensor-driven environment, and enable its real-time operation.

Iguazio, in this respect, has a workflow integration characteristic in common with Quantum and its StorNext product. The workflows are so distinctive though, that neither company’s products will work in the other’s environment. The two companies’ products don’t overlap. 

Iguazio’s strategy is to prevent other less workflow-focused storage suppliers entering its Internet edge markets by integrating its product with the data-generating devices and applications there better than competitors. It’s found a way to make its fast storage array offering differentiated from virtually all the other suppliers and that is no small feat.