The move to disaggregated components in servers is well received by cloud providers. However, ironically, the disaggregated servers (compute/storage/networking switches) doesn’t work well in marketing/messaging to enterprise customers, who would like to buy servers as workload-based solutions.
What’s interesting here is that the same technology is appealing to both buying communities – enterprise IT and cloud providers. But they’re asking system vendors for different packaging and different messages about the products.
Case in Point
Last week, I chaired the Future of Open Servers panel discussion – the closing session at the Open Server Summit in Santa Clara, CA. One customer, two analysts, and three vendors all came to the same, overriding conclusion, through that panel discussion: Data center technology is getting better and faster, but it’s getting more complex for enterprise IT to acquire and more difficult to manage across platforms.
The Quest for Simplicity in Enterprise IT
What to do? Speakers suggested vendors need to provide better ways of packaging the technology, so that it can be consumed more easily. This is especially important for enterprise IT customers, who are deploying a new layer of scale-out architecture to support new workloads in analytics, cloud computing and mobility in the datacenter. Software-defined infrastructure (SDI) and orchestration software to manage large numbers of server platforms.
When it comes to acquiring systems, IT managers love the faster feeds and speeds, reducing network travel time (latency), and improving end-to-end performance. But opinions are mixed about how best to package this scale-out technology (capex).
Here are some examples of package options for enterprise IT:
- Purpose-built appliances, as “on-ramps” to public clouds.
- Engineered systems optimized for specific workloads
- Converged/hyperconverged systems combining compute, storage and software.
Inside the datacenter, customers are focused on controlling IT costs (opex) – and would like to do more management tasks via a single-pane-of-glass with a global view of resources. And, they are beginning to adopt software-defined infrastructure (SDI) to manage workloads, VMs and containers across the data center. At the same time, customers are pushing more workloads into public cloud resources, when possible to do so.
Now, What do the Cloud Providers Want?
In contrast, the cloud providers enjoy the process of “disaggregation” of systems. Why? They have enormous IT resources, and lots of IT staffers, to assemble their own brand of systems infrastructure as they work to optimize cloud services they deliver to end-users. CSPs write their own customized software to manage thousands of servers. They are pioneers in deploying SDI for orchestration/management.
The process of disaggregation allows CSPs to balance the “mix” of compute, storage and networking, as needed – and as workload demand for IT resources shifts and business conditions shift.
In this brave new world of scale-out web and cloud infrastructure, much of the attention for AppDev and DevOps goes to NoSQL databases (e.g. MongoDB and Cassandra); scale-out analytics (e.g. Hadoop, Spark) and OpenStack open-source software; these are now the hot topics of new IT workloads. This approach may be expensive, given high staffing costs, but it pays off in revenue and profits from cloud services they sell.
Disaggregated systems will increasingly be deployed in “open racks” as IT enterprises continue to migrate more applications to cloud providers for deployment and management. Expect a “cross-over point” by 2020, when more than 50% of enterprise IT workloads will be hosted on cloud resources.
Given the outlook of these two distinct buying communities – enterprise IT and cloud providers, I expect that systems packaging and marketing is bound to take two different tracks.
These parallel conversation streams neatly illustrate the dichotomy of computer systems marketing – presenting systems vendors with two buying audiences that are apt to talk past each other, most of the time.