Stop Over-Spending on Underutilized Resources: The Future of the Data Center is Composable Disaggregated Infrastructure

CDI lets organizations respond to changes almost instantly while also reducing costs—meaning IT can allocate resources on the fly.

istock 1185894817
undefined undefined

The only thing constant is change, and data centers are no exception. As data architects attempt to anticipate future data center needs – while delivering the required SLAs to the business – the solution is often to over-provision resources so that the infrastructure can absorb any changes or periodic spikes in demand.

But today change happens much more frequently, whether it’s onboarding new applications or reaching new heights in data growth. Most often, organizations expect immediate implementation of those changes In today’s environment of flat or declining budgets, IT can no longer afford to over-provision its way.

Composable disaggregated infrastructure (CDI) enables organizations to respond to changes almost instantly while at the same time reduce costs. This in turn helps IT better align changing business needs and allocate IT resources on the fly.

Do any of these scenarios sound familiar?

  • It’s difficult to anticipate storage, network and compute needs in an evolving data center, so you take a scale-out approach.
    As new applications are onboarded, you just buy more servers with more storage (and more compute). Very quickly you wind up with a “SKUnami” – a multitude of processing platforms (SKUs) and software licenses that, while required for peak demand, go mostly unused. This is classic over-provisioning with a multitude of customized platforms that don’t work together and are highly inefficient. A one-size-fits-all general-purpose approach does not fit all workloads and cripples the interoperability matrix.
  • Standardized platforms don’t always match application needs.
    As a result, you end up with stranded capacity on different silos. Resources go underutilized (such as idle CPU cores, unused DRAM, or untouched SSD capacity) and you cannot reassign resources where they are needed because they are trapped in a hyper-converged node.

With today’s unprecedented data growth and unpredictable future needs, you can no longer afford to buy infrastructure just for the “peaks.” Hardware deployment decisions must be made long before the hardware is actually used. Applications and data are guaranteed to grow but the hardware capabilities are fixed. In traditional infrastructures, you cannot easily grow or shrink resources to adjust to growth rates.

You need a next-generation infrastructure that aligns and optimizes storage resources, while meeting SLAs and without incurring unwarranted expense, to address today’s dynamic data requirements.

Emerging Next-Gen Technology: Composable Disaggregated Infrastructure

The industry is evolving toward CDI, which takes physically aggregated devices (compute, networking, storage, and GPUs) and pools them together as needed for a specific application, without the need for physical configuration. This allows for composing disaggregated system resources – compute, networking, and storage – into virtual systems. These virtual systems are then dynamically provided to the right application at the right time, ensuring SLAs can be met.

In the past one would just purchase new servers, but CDI can deliver greater productivity and agility, improved utilization and faster provisioning while at the same time improving availability and performance. When a new workload is ready to move into production, instead of purchasing new servers, new networking devices, and a new storage system, the IT operations team provisions the resources dynamically to the workload from an available pool of resources.

A Shared Economy

It’s like having all of the needed components on a virtual shelf. The quantity, speed, and power of the resources (such as compute) can be selected independently of the other resources (such as network, FPGA, GPU, flash, and disk). Imagine a large organization with fifteen divisions. Each division would like to use some GPUs for analytics, but only needs them 10-20% of the time. As an alternative to purchasing 15 GPUs, using a shared pool of resources the company can invest in fewer GPUs – perhaps four to six – that are in use more of the time. They can be scaled independently into a complete solution for the workload, and then put back on the digital shelf when finished, ready and waiting for the next workload that requires them.

IT organizations are wasting significant costs on underutilized resources and not getting an appropriate ROI for these investments. Workloads that need to optimize their usage of high price-tag assets, such as flash storage or GPUs, include those that need to deliver high-performance and low latency, such as: real-time big data analytics, AI/machine learning, NoSQL databases, high-performance computing, and more.

Some leading industry analysts have placed composable infrastructure at the peak of their hype cycle for Infrastructure Strategies. IT is ramping up for the future data center. The need for CDI to be an open, industry-wide initiative cannot be stressed enough. To deliver on the promise of a composable infrastructure, the foundation beneath it needs to be built around an open architecture and open API. This ensures the various physical resources can be logically configured and composed into virtual systems without vendor lock-in. 

NVMe-over-Fabrics: The On-Ramp to CDI

Making a CDI work requires a flexible, yet high-performance network with very low latency. NVMe-over-Fabrics (NVMe-oF™) is an ideal network for CDI because it can deliver both high performance and low latency while remaining industry standard.

NVMe-oF, which is increasingly being adopted, serves as an on-ramp to CDI. It allows IT to compose, orchestrate, or share flash storage over fabrics. Using NVMe-oF you can optimize your flash storage investment for dynamic workloads. NVMe-oF will vastly improve server and storage utilization, performance, and agility in the data center.

IDC predicts CDI will be a $4.7 billion business by 20231. Western Digital estimates that by 2025 most data centers will have adopted NVMe-oF in some form2. To get on board the composable train, we recommend starting by sharing flash storage to get some CDI experience. What you can expect is a higher percent utilization of flash without any decrease in SLAs or customer satisfaction.

The Data Center of the Future is All About Shared Resources

Data is not rigid and the infrastructure in which it lives cannot be rigid either. The future data center will be built around the data – to ensure the data is enabled to come alive and drive actionable insights. CDI is truly the future of the data center. By pooling resources and tapping into them as needed, you can create a flexible architecture that is relevant for today and extensible to support future data center needs. If you rethink data centers with data at the center (as we discussed in a previous article), it’s not the physical locations that matter. Composable infrastructure provides the building blocks of compute, network, and storage that can be easily composed, deployed, and even torn down on the fly. This is the flexibility required for the future data center that simultaneously avoids wasted resources and can provide big cost savings.  

1 Worldwide Composable/Disaggregated Infrastructure Forecast, 2018–2023, IDC, Doc # US44224418

https://blog.westerndigital.com/from-flash-to-fabric-nvme-of/

post 6 image wd Western Digital
Related: