The New Media Evolution, Memory-Centric Architectures and Deep Learning: 2017 Predictions from the CTO

As we approached the end of 2016, I reflected on the storage trend predictions we identified – Containers, 2-Tier Storage, Cloud Portfolio Management, New Media technologies and IT Skills focused on Cloud-native development.

While predictions are mostly educated guess, our 2016 predictions played out favorably.

So now it’s time to look ahead for 2017… what can we expect this year?

1. Cloud-Native Stacks and Technologies will continue to absorb more of the traditional enterprise workload base. At the end of 2016, we saw that Java Apps could move to systems like Cloud Foundry and get huge operational advantages. We expect this to continue and we also expect that in 2017 we will see other traditional enterprise environments become re-factored into a cloud-native model. Systems of record such as SAP and Oracle will shift to cloud-native application frameworks and in-memory databases more aggressively in 2017. This trend is making its way in IoT and Industrial systems from companies such as GE with Predix and Siemens. These new platforms and applications will still require cloud-oriented IaaS, such as VMware, under them. The enterprise applications will become more separable, flexible and automated by leveraging the work originally done for greenfield cloud-native apps.

2. Cloud Data Protection will become a real concern. In 2016, cloud adoption both on- and off-premises accelerated but the enterprise resiliency of the legacy systems has not kept pace. One of the most significant gaps is how to protect data in the cloud era. Most customers are unaware of the liability and indemnification limits of SaaS and public clouds. We have not yet seen massive data loss and consequently the focus has not been on how we handle these situations. In 2017, as multi-cloud enterprise IT designs become more commonplace, the need to protect that data in SaaS systems or across multiple on and off-premises clouds will become critical. Dell, as the leader in data protection, already delivers this service to on-premises clouds and some SaaS applications, but 2017 will be a year of expanded capability including retention, compliance, and recovery and business continuity. Customers can be sure that even if a public cloud or SaaS provider loses, has compromised, or corrupts their data that there is a recovery path.

3. The new media evolution will continue. We already see early details of non-volatile memory such as Intel 3DXP but in 2017 we’ll see early access to this and other technology. While we don’t expect mass adoption or a strong applications ecosystem to form around such new memory in 2017, it will be a year where the reality of these media starts to become evident. Performance, endurance and functionality will become characterized in real systems rather than marketing material. The limitations and issues with this shift in a foundational building block will surface and drive innovation that will be required. A great early example is the need for error recovery when using persistent memory – you can no longer just reboot your system to clear the run time memory when using these technologies. How will we reset, recover or roll back to a safe state? Seems like a great area to innovate across recovery, data locality, and media management along with enhancements in the software stacks to leverage this new media.

4. The first real availability of post-flash memory in 2017 will start the acceleration of software to exploit massive pools of persistent memory. We call this the era of Memory Centric Architectures. We expect that over time more of the semantics used to access data will shift from storage instructions such as “Read and Write” to memory semantics such as “load and store”. The implication is that our applications will no longer see the memory and storage pools as separate but instead more of them will see data as stored in huge tiered memory centric pools. In-memory processing architectures of databases and applications frameworks will become much more common. The functions of the infrastructure to manage this diverse set of media as reliable scalable memory pools will become the new frontier of innovation in storage technology.

5. Multi-Cloud enterprise architectures dominate. We expect 2017 to be the year when enterprises realize that not all clouds are equal. Given the variety of workloads and use cases, the modern enterprise will need a diverse set of infrastructures that deliver the correct set of performance, cost, compliance, security and scale to serve their needs. This means that multiple clouds will be needed; the default objective will be to select the right set of clouds but – more importantly – to unify their operations in ways that enable a coherent set of services to the enterprise users. This will result in a huge shift where innovation will be more focused on how to operate a multi-cloud enterprise rather than how to build or use any one specific cloud.

6. Machine Learning and Deep Learning will be the new bright shiny technology for enterprises. While we don’t expect mass adoption of ML/DL due to its early nature and lack of IT skills to exploit, we will see in 2017 that the dialog around using new artificial intelligence engines to create better data-driven outcomes will start to be more dominant in forward-looking IT strategies. This is likely to originate first in entirely new cloud-native application stacks and IoT use cases due to the need to complement relatively unintelligent endpoints with a level of infrastructure artificial intelligence. We are confident that this can happen over time and expect that 2017 will see continued public cloud based ML/DL engines like Google Tenserflow evolve. We are now also seeing dedicated on-premises ML/DL stacks enter testbeds and be leveraged in specific industries; the Toshiba/Dell deep learning is an excellent example – recently approved by the Industrial Internet Consortium.

While there will be many other activities in 2017, the six highlighted predictions above indicate that we are settling on a multi-cloud future for IT that will be driven by innovation at the lowest level physical media and will have a significant impact on core functions such as processing and storage. Above that we expect to see a blurring of existing and cloud-native technologies to create more efficient hybrid cloud operating models and development frameworks. Finally, we expect to see some new elements of technology such as machine learning and deep learning make their way into our thinking about IT architectures. The pace of change continues and 2017 is likely to be characterized by expanded innovation driving another very interesting year for IT.

About the Author: John Roese