Kubernetes’ extraordinary resilience tends to change the emphasis of monitoring from alerting to resource and performance management.
You are here
Is IT Hardware really getting a second life?
On Monday of last week, Katy L. Huberty, head of Morgan Stanley’s North American Technology Hardware Equity Research unit, published a research note with a counterintuitive prediction. Huberty and her colleagues had identified a "perfect storm" of conditions that could drive double-digit earnings growth for big-brand enterprise IT hardware in 2018. Simon Sharwood, APAC Editor of The Register, drew our attention to Morgan’s prediction in a Tuesday article -- one whose deadpan tone might suggest he agrees with Huberty's reasoning.
With public cloud growth still accelerating and dominating the news, and with IT on-premises hardware revenues now static or declining for most of a decade, Register commenters were mostly not buying Morgan's prediction. “Seems like a typical, mindless, Morgan Stanley BTFD call, Simon,” opined reader amanfromMars 1, who added: “Clouds are not hardware-critical. They are software-formed and driven.”
Other Register readers pushed back by noting that the trend towards cloud (and, presumably, away from on-premises datacenters), though initiated by developers and end-users, has now become executive orthodoxy. As reader Olaf wrote: “There’s still a very strong strategic push to move stuff into the cloud. And that push is happening, irrespective if cloud-based services are the right choice or not.”
Is Morgan just nudging investors to ‘buy the (effing) dip?' If not, what factors combine to produce Morgan's "perfect storm?"
Plenty of Cash
Morgan makes a solid case that cash is now available for private cloud investment. The largest US enterprises are being nudged by a range of financial dynamics influencing cash availability and promoting aggressive spending, including:
- Lower US corporate taxes, freeing up cash.
- New incentives to repatriate cash sheltered from taxation overseas, including accelerated depreciation opportunities.
- A weaker US dollar, against a background of overall economic growth.
Outside the US, a generally-positive outlook prevails as well. In China, renewed global IT hardware spending would buoy the top end of the manufacturing sector (now flagging at the low end), and might encourage increased Chinese spending on enterprise hardware too -- already predicted by Forrester to grow at 6%. Rapid 2018 growth in enterprise IT spending is also expected in Japan (the largest IT market in APAC) and India.
Hybrid, not Private
The main thrust of Morgan Stanley’s argument is that reduced enterprise spending for on-premises hardware reflects not so much a desire to abandon private datacenters, but just to pause further capital investment while figuring out where public cloud fits into enterprise IT strategy. Now, they say, several years of experience has taught enterprise IT leaders that hybrid clouds offer the best combo of flexibility and cost advantage, so spending on private clouds will now resume.
Is this an out-take from the Private Cloud and Datacenter Hardware Wish Fulfillment Playbook, or does it (also) make sense? Yes, enterprise use-cases are diverse: from delivering predictable, commodity IT services with long lifecycles, to responding with agility to fast-changing market and seasonal demands, competitive pressures, and the requirements of hair-on-fire skunkworks development projects. Public clouds can answer well (so goes the conventional argument) where demands are variable, intermittent, unpredictable, and/or novel. But where things are more predictable, private clouds are often cheaper; especially at the large scales at which global enterprises work. As one Register reader noted: "For fixed demand, cloud is just hosting, and we know we can do that cheaper than Amazon.”
Is the marginal cost advantage of hosting predictable apps on private clouds sufficient to compel increased enterprise hardware spends? We’re not convinced that this outweighs the notion – long used to promote IT outsourcing, public cloud growth and related phenomena -- that enterprise IT needs to justify its existence by producing differentiating business value. In a truly biz-value-forward IT cosmology, you don’t put boring, predictable, commodified (but also mission-critical) applications on private clouds. You rent them from SaaS providers.
Helping Morgan Stanley Make a Better Case
Morgan ends its brief by suggesting – without elaboration -- that containers, automation and the Internet of Things (IoT) may help drive enterprise demand for on-premises hardware. Though the Register’s commenters mostly ignored this assertion, we think that if Morgan had built on this idea, it would have made their case much more compelling.
Public cloud costs are, as Register reader Anonymous Coward (no relation to Noel Coward) puts it: “horrifically bad once you are a company of a decent size. I know a decent list of companies who have moved a lot of infrastructure to the cloud – successfully – and then moved lots of it back once they saw the seven-figure bills from Amazon, every month.”
But for private clouds to grow – as Morgan says they will – they need to find a greater purpose than just hosting predictable, commodified applications more cheaply than Amazon. To be fully viable in enterprise hybrid clouds, private clouds also need to offer a public cloud experience in the on-premises datacenter. At minimum, they need to support the same development workflows and run all the same, high-value applications public clouds do; offer similar (or identical) lifecycle management, operations, and monitoring options and interfaces; and support key hybrid cloud use-cases, including bursting on demand and workload portability.
As Morgan low-key suggests, containers, automation, and other innovations are now making this happen, and IoT and other novel technologies are accelerating and enabling the change.
IaaS Solutions for Hybrid Cloud
Large enterprises with strong commitments to Microsoft and/or VMware may look no further than products like the newly-delivered Azure Stack (premises-based Azure regions) or the VMware/AWS partnership (VMware regions running on AWS-managed cloud infrastructure) to turn on-premises and public clouds into a continuous Infrastructure-as-a-Service substrate, managed through single, familiar panes of glass. Though proprietary and comparatively costly, these solutions provide full lifecycle management of the cloud framework, and can (if configured to do so) deliver the full gamut of hybrid cloud benefits, including seamless workload portability and bursting,
Similarly-conceived, but more open solutions, available from providers like Platform9 and ZeroStack, host highly-available control planes for OpenStack and Kubernetes and handle lifecycle management, while letting you deploy compute/worker nodes in private datacenters and on popular public cloud platforms. These, too, enable single-pane-of-glass management and portability of workloads between private and public-cloud regions.
Hybrid Container Cloud
More generally, Kubernetes is emerging as a leading tool for standardizing container workload hosting across private and public datacenters. Every public cloud provider now offers Kubernetes-based Containers-as-a-Service, and numerous third parties (such as StackPoint.io) offer ways of automating and lifecycle managing Kubernetes clusters on public clouds. While it’s still somewhat challenging to deploy production Kubernetes in-house, projects like Kubespray are making things easier: providing a generalized solution for getting Kubernetes on premises-based or public-cloud-hosted IaaS virtual machines or on bare-metal nodes.
With Kubernetes running on premises and on one or more public cloud providers: problems with workload portability mostly go away. You can deploy the same app to any compatible Kubernetes cluster using the same automation and manage it via identical lifecycle management code, CLI commands, and WebUI operations. Using Kubernetes federation, you can sync resources across public and private clusters, enable cross-cluster discovery, and thus easily automate key hybrid cloud use-cases, like bursting.
Climbing the Value Stack
If you prefer (and this is where IoT comes in), you can climb higher up the value stack and deploy serverless compute (aka ‘Functions as a Service’) solutions like OpenFaaS on Kubernetes, using it to create and manage even lighter and more-portable functional solutions for (among other things) high-throughput realtime processing of time-series or other metrics from sensors and other IoT devices. Solutions like OpenFaaS provide a more open, less proprietary alternative to AWS Lambda, that runs on Kubernetes (or Docker Swarm) orchestrators running on private and public clouds.
At Opsview, we’re investing in automatic monitoring and predictive analytics, making it easier to implement monitoring in Kubernetes, serverless compute and other complex frameworks, keep track of their health, and gain rapid, useful insight into the often-transitory workloads they execute. We think we can significantly reduce the effort, skill level, and headcount required to manage these platforms proactively, along with the critical business services they support.
With all these opportunities and trends in play, Morgan Stanley’s prediction that a bounce is coming for enterprise datacenter hardware seems more credible. Whether it will happen in 2018 or somewhat later – perhaps awaiting better solutions for auto-deploying production Kubernetes in private datacenters – is an open question.
More like this
Opsview's Bill Bauman and John Jainschigg attended Percona Live 2018 -- to talk about serverless computing, database monitoring, and catch up with...
Whether you prefer Ubuntu, Centos or RHEL, monitoring the performance of your Linux hosts in the public cloud is not as straightforward as it...