Information technology agility today is key to running the next generation of applications and programs. 

For too long a business has relied on the traditional 3-tier architecture approach of servers, storage and networking.  The next step was converged infrastructures where certified and approved hardware was connected using a software layer to manage various aspects of the infrastructure. 

The next phase is HCI (hyper-converged infrastructure), this uses a software defined layer to manage the whole hardware and software stack to including switch management, VM creation and deployment, data protection and security.  

HCI is a far complete solution than CI, converged infrastructure, as it scales better, is more flexible and can utilise standard x86 servers.

Composable Infrastructure

She’s a beauty

The next step in the evolution of I.T. is Composable Infrastructures, unlike HCI where you create nodes and pool them together to provide compute, network and storage, composable treats each of these components as a pooled resource.

Composable infrastructures provide the ability to automatically create or remove these pools within minutes.  No longer do you need to provision nodes to perform specific tasks.

Moving to a composable infrastructure conveniently turns your datacentre into an elastic bare-metal cloud and allows you to instantly create the necessary resources your applications need. 

Build instantaneous workloads for your applications and when you’re finished, the resources are returned to pool ready for the next application.  Due to the way a Composable Infrastructure is, they’re highly scalable allowing you to manage 1000’s PB’s from a single console.  

All good in theory, but what about practicality?

The number of applications and uses that a composable infrastructure could be used for is immense, below we list some of the more common ones.

The Cloud

Purely from a financial standpoint the cloud is expensive and slow when uploading and downloading large data sets of information to be analysed.  The high cost of cloud computing and storage services are a barrier to building a cost-effective infrastructure, whereas a composable infrastructure could provide the elasticity a business need.

Big Data and Data Mining – Employs Hadoop, NoSQL, massively parallel databases, and other applications for advanced analytics against very large data sets of structured, unstructured and semi structured data that resides in data lakes and data warehouses.  This is where composable infrastructure wins as it allows to scale when the data becomes available.

Machine Learning – This is evolving at a phenomenal rate! It seems everyone father and his son are having a crack. The technology and infrastructure required to deploy the necessary component parts are expensive and complex.  Compute and data-intensive deep neural networks require much larger data sets and high-performance technologies such as GPUs and low-latency network fabric. Production deployments, also called inference, are less computationally intensive with lower requirements for compute and storage yet may still require low latency.  Having the complete flexibility to manage resources on-the-fly provides an ideal solution for machine learning applications.

Kubernetes Containers – It groups containers that make up an application into logical units for easy management and discovery.  Designed on the same principles that allows Google to run billions of containers a week, Kubernetes can scale without increasing your ops team.

Kubernetes provides the tools needed to build and deploy reliable, scalable, distributed applications. And with the addition of storage volumes, Kubernetes solves the need for storage availability by requesting and attaching volumes to containers making the deployment of stateful applications, such as databases, in containers a popular solution.

With Kubernetes, IT can deliver advanced analytics, AI/machine learning and NoSQL databases using
lightweight, flexible containers with the ability to deploy, scale and manage them seamlessly.  This was one of the first applications that composable storage was designed to use and offers instantaneous deployment of pooled resources.

Tiered Data Storage – Tiered storage assigns different categories of applications and their data to a range of storage media types. Tiering for online data ranges from tier 0 for hot data, such as transactional databases and fraud detection, to tier 3 for cold, infrequently accessed data. Tier 4 is generally for archived or regulated data stored on tape or at a cloud service. No matter the applications or how you tier your data, finding a server and storage infrastructure that adapts to your requirements is paramount.  Composable infrastructures provide the ability to automatically migrate data from HDD – SSD – NVMe or back again as and when required.

Cold Data and Object Storage – Enterprises and cloud companies are employing big data analytics and AI to gain operational insights, make better business predictions and personalise content for customers. The continuous growth of data for these uses, and many more require companies to build larger stores, keep data online longer and bring archived data back online. Much of this data is infrequently accessed, or cold data, and requires economical and efficient storage at scale.

Accurately planning how much compute and storage is needed for cold data or a growing object store, and efficiently provisioning compute and drives is an ongoing challenge.

As a result, there is growing demand for infrastructure solutions that can be flexibly designed, scale seamlessly and cost-effectively while ensuring quick and easy access to cold data whenever it is needed.  As composable infrastructures can grow from TB’s to PB’s utilising differing storage technologies makes this an ideal choice.

NVMe-oF Infrastructure – Data-driven companies are increasingly using NVMe storage for performance-hungry workloads that require low latency and high I/O throughput.

 These applications include analytics, predictive modelling, streaming, machine learning and large-volume transaction processing.  There are many advantages to flash storage, most notably high IOPS in a small footprint – a single 2RU all-flash array can replace one or two racks of HDDs when rightsizing the capacity to the required performance.  With the ability to carve flash drives into slices that can be attached to individual compute nodes, flash utilisation can be increased dramatically. In addition, reduced energy requirements and higher reliability make flash an ideal data centre solution.

NVMe over Fabrics (NVMe oF), a new high-performance networking solution, eliminates the performance issues that required you to isolate flash inside server nodes.

The disaggregation of compute from storage creates the opportunity for composable infrastructures to orchestrate or recreate the server and storage platform in a completely new way.

Future Possibilities

Today, most data-intensive computing workloads are run on direct-attached storage, to provide performance and capacity.  Whilst this is ideal for handling large data volumes it is complex to install, manage and deploy, and normally over provisioned or underutilised. 

The complexity of adding additional DAS-Direct Attached Storage to the environment causes issues, even more if adding DAS from a different vendor.  The flexibility a composable infrastructure provides massively simplifies this infrastructure by allowing you to simply add additional resources as and when required.

Daily, data is being created to un-paralleled levels, the data needs to examine, analysed, mined and acted upon and ultimately stored. This is where composable allows for simple organic growth as and when required all controlled from a single GUI.   Composable infrastructures provide the agility and ability to manage any modern infrastructure, efficiently, and whilst being secured using data encryption and the ability to reduce TCO using commodity elements that scale.

Summary

If you work for a business that is producing huge amounts of data, or own huge amounts of legacy data, deploying a composable infrastructure would be the best move you can make.   

At Fortuna we can also offer your company with data mining as a service, for either a flat fee or a subscription.

At Fortuna Data we specialise in CI, HCI and composable infrastructures, talk to us now how this next generation of technology is really making a big impact in the way businesses work.

If you want to find out how we can help, call us on 01256 331614 or email solutions@fortunadata.com

Visit our web page for more information https://www.data-storage.uk/composable-infrastructure/