We have seen a significant shift in enterprise IT over the last year–specifically as it relates to application data and how it is provisioned, stored, processed, and managed. But this shift is not only as a result of the COVID-19 pandemic. IT Organizations appreciate the benefits of the cloud experience–applications up and running quickly while underlying infrastructure is maintained for users ‘behind the curtain.’ Unfortunately, when it comes to providing a cloud experience for data and related infrastructure that needs to stay on-prem, organizations can struggle to deliver it because of a variety of factors:
- Demand on expert IT staff which are scarce and difficult to retain
- Separate infrastructure silos for bare metal, virtualized & containerized applications complicate infrastructure management and automation
- The expense of external storage and restrictions of SDS/HCI constrain any significant shift to a cloud experience
So how will it affect the direction of IT and the cloud in 2021? For data than cannot or should not go to the cloud, the cloud needs to come to your data.
PREDICTION 1: On-prem infrastructure management will move to the cloud
Speak with any IT Organization and most will tell you that they have data that cannot or should not be moved to the cloud for security, cost or compliance reasons. Roughly 78% of mission critical data is on-prem in core and edge data centers serving a wide variety of applications. The various management silos created by a unique mix of legacy infrastructure per application is further complicated by the new remote workforce norm due to COVID-19.
A shift to cloud-based management for on-prem data has a number of benefits, and is a success story we have seen over and over again on the consumer side. One such success story that comes to mind is NEST which places a physical thermostat in your home, but is completely managed in the cloud. In this scenario, NEST users always benefit from the latest features that NEST offers as software is updated without any action required by the homeowner whatsoever. Plus, remote management features allow the homeowner to adjust the HVAC settings for their home from anywhere in the world and simple AI-assisted device automation means that HVAC operation can respond to a variety of events without any manual intervention.
So why can’t this be done in the data center? It can, but enablement requires an architectural separation of the control plane from the data plane–control plane moves to the cloud while the data plane remains on-prem. With traditional storage software, 75% of the code is related to management, meaning I/O path updates tend to be more infrequent. Moving the management layer to the cloud means most software updates and delivery of new features happen automatically, like Gmail or the NEST thermostat. In addition to automated software updates, organizations can leverage AI-assisted remote management for global reporting and proactive recommendations to improve performance and get better utilization rates.
PREDICTION 2: Self-service models will be a critical path to success for on-prem environments
According to an article published by a Gartner analyst last summer, 90% of enterprises will fail to scale DevOps initiatives if they don’t move to a shared, self-service platform. One reason is DevOps teams lack the time and occasionally the expertise to provision, administer and support on-prem infrastructure. The frequent infrastructure set-up and tear down necessary in many DevOps environments is huge, but with a more cloud-like self-service infrastructure model, users get the rapid composability needed in these fast moving, rapidly evolving environments.
Self-service provisioning of on-prem application infrastructure, similar to the cloud-native experience, makes application deployment as simple as creating a meal in a buffet line. Application owners are first presented with an application service catalog where they can select a curated recipe for the application they want to provision. After that, they simply select the application servers needed, and voila! Servers, operating system and application software, boot and data volumes, etc. are provisioned in minutes. No frustrating DIY, no recabling, no request forms, no meetings, no emails between teams–DevOps teams can get what they need, when they need it, in a way that is certified by an organization’s infrastructure experts…FAST.
PREDICTION 3: API-first infrastructure will be a requirement for IT managers
Enterprises are hitting the gas on automation initiatives–according to Gartner, automation is so strategic to IT orgs that by 2025, more than 90% of orgs will have an automation architect. 2021 will be a turning point for API-first infrastructure solutions. But what does that mean? Products and services which are API-first, start with the API and then build the necessary tools & UI on top so that there are no hidden systems calls and no delayed access to new functionality. Think of it as a fully programmable end point in the cloud with full access to all devices across your environment with full functionality.
In order to do be API-first, a product should come with a variety of SDKs and include integration with popular integration networks. API-first products should allow IT orgs to develop simple automation that scales across the entire organization and allow interesting queries about the entire data center environment, globally, for appropriate insights. If done right, you will also benefit from one-click fleet management and easily maintain endpoint elements such as server SSD firmware, boot images, and more, across your enterprise.
PREDICTION 4: Data services move off general purpose server CPU’s and onto smart devices in the server.
The sad reality is that the agility of an enterprise is dependent on where it’s data services run. For example, if your data services run on an array controller, you naturally have the complexity and expense associated with 3-tier architectures. Moving data services to your server is a great alternative, but solutions like Hyper-Converged Infrastructure consume valuable server resources requiring a 25-30% higher server investment, and tie your data services to a specific OS or hypervisor, limiting flexibility to support of a variety of workloads. This is a problem especially if you need to pivot to Linux for bare metal applications or to a containerized deployment.
We have seen the rise of SmartNICs and DPU’s for certain network workloads, and it follows that moving your data services to a smart IoT endpoint in your server eliminates the costs associated with 3-tier architectures as well as the restrictions posed by Hyper-Converged solutions as it doesn’t take CPU, network or memory resources from your server. And because data services are decoupled from the OS, it is supported by any OS/Hypervisor and allows for boot drive management, a key ingredient for self-service provisioning across your environment.
PREDICTION 5: growth at the edge with drive increase for cloud-managed infrastructure
According to Global Market Insights, the Edge Data Center Market will grow to $16 Bn by 2025. The challenge this creates is a management nightmare of a broadly distributed environment which is truly only addressed with cloud-based management over time. With a cloud-managed approach to an Edge Data Center deployment, enterprises can enjoy the benefits of zero touch remote administration–a must for remote sites where there is no on-site personnel. With fleet managed software updates, IT organizations can update firmware with a click of a button across their entire worldwide data center footprint, which could be 100s or 1000s of sites. A cloud-managed infrastructure also allows data center to take site automation to a new level through a single API endpoint in the cloud. And there is the added benefit of common management and a common approach to storage within the core data center.
With the growth of 5G and IOT, edge initiatives (and challenges) will define success for a growing number of enterprises, and a shift to cloud-based management holds the key.