Modified from an article originally published @Oracle.com on Monday, March 20, 2017 as :
CLOUD ARCHITECTURE AND IMPLICATIONS
By: Todd Jobson | Sr. Principal Enterprise Architect; MBA
If you’re wondering WHAT “Conditions” led us to Cloud Computing ? .. this brief synopsis is targeted to bring you up to speed rapidly on the underlying themes and industry momentum (in a nutshell).
As you can see from the IDC report and chart below, the 2000’s paradigm shift was a rush to cut costs in a down global economy, where Capital IT Expenses (CAPEX) were the first to go. Demand for expensive, high-performance HW (with 99.999% available RAS) became a luxury, and the new phrase “Good Enough” became the gauge that so many came to use (especially given that the majority of systems in place at the time where underutilized without virtualization, if not sitting idle as HA/Fail-over systems).
The rise of Virtualization technology in many areas :
–> Consolidation of Workloads & Systems : together, Virtualization technology allowed companies the ability to effectively consolidate workloads (and Tenants) from many systems down to fewer (also increasing the utilization rates of those systems).
*For a more thorough examination of “enabling technology”, see my Cloud Architecture 101 article.
The missing piece within these “commodity“, (Custom-Integrated so called BYO (“build-your-own”) or BOB (“best of breed”) solutions was robust unified full-stack Lifecycle Management/ Orchestration of App SW, OS, System Provisioning, & Virtualization, NTM full-lifecycle management.. where VMware became the dominant player (at an ever growing, though initially hidden cost).
While a little dated, the graph below clearly depicts the historic shift to fewer physical servers AND greater numbers of VM’s (Logical servers).
The irony is that what was initially viewed as a low-cost “panacea”.. (when dozens of ~FREE/FOSS KVM/Xen VM’s and x86/Linux systems need to be managed), quickly becomes a System/VM Management & Maintenance challenge as environments rapidly scale up & out, which adds more significant Linux-RHEL, & VMware/ESXi+ licensing costs (initially thought to be FREE & reduce cost).
These hidden Support Costs for the Linux OS & VM licensing, coupled with the time & effort to maintain these growing islands of workloads running on several-fold 1-2 socket x86 Servers (without SPOG & automation, can incur much more downtime to configure/patch 2x, 4x, 10x the number of systems and resulting VM’s), were the catalysts that resulted in the realities of the chart below. The task became so great, that VMware’s dominance and traction grew even further, noting it’s advanced management/automation capabilities to fill the SPOG void that existed. Today, the IaC (Infrastructure as Code, eg. Chef, Puppet, Ansible, Terraform) competition is squarely targeting this space, in addition to OEM frameworks/toolsets (eg. Dell OMEnt, Oracle OEM, etc).
For most Virtualized x86 (eg. KVM, VMware) deployments, over-provisioning is needed to accommodate for overhead (eg. each VM requires an independent full copy of the OS & CPU/Memory resources, +VM scheduling and layers of emulation/translation ..) inefficiencies (~12%+ per VM is not uncommon). This is why modern Enterprise deployments of VMware need to utilize para-virtualized drivers to bypass as much overhead as possible, also one reason many are deploying/exploring Containers, which share only 1 copy of the underlying OS kernel and can run on BM without a Hypervisor layer.
Without Vertical Scalability (> 2 CPU’s) as a high-performance “tightly-coupled” & cost-effective option for larger workloads, an additional penalty for Horizontal Scalability GROWS with greater network latencies in larger “loosely-coupled” distributed environments, incurring more network hops and inter-system traversing of TCP/IP stacks and protocol hand-shaking (several fold vs. all happening previously within a single system).
For many of these reasons, Converged Infrastructure & Engineered Systems began to gain some traction to reduce these costs & complexities.
However, granted that most of these commodity Converged x86/Linux/Virtualized IaaS/SDN deployments have continued to grow.. and grow, the cycle continues .. further fueling a justification to explore alternatives to On-Premise Infrastructure.. aka Hybrid or Public Cloud deployments & services.
This section is an excerpt from the article : Cloud Architecture 101: The Road to Cloud Services (IaaS/PaaS/SaaS) & Deployment Models (Private, Hybrid, Public)
Before your organization boils the ocean and does detailed Discovery & Requirements Analysis, the 4 Key areas in the chart below highlight the Key Questions that you need to first explore more thoroughly in order to determine cloud “viability”, let alone select the appropriate Cloud Service (IaaS, PaaS, SaaS) & Deployment Models (Private, Hybrid, Public) :
On top of the items noted above, having a greater density of an organization’s SW running on less reliable (lower RAS/ less redundant) HW, lends itself to other significant business issues (having all your eggs in fewer baskets) :
Combined –> = Customer Business RISK, Overhead, Complexity, COST, & Control
Collectively, for all of the reasons above (Reduced Complexity, CAPEX reductions, reduction in facilities, staffing, etc), most organizations are today exploring the MANY private, public & even more prevalently, the Hybrid/ Multi- Cloud Deployment models & Service Offerings. Public Cloud is still going to always be an option (AWS, Azure, Oracle OCI, Google, ..), however today the momentum is definitely targeting the rapidly growing On-Premises Private & Hybrid/Multi-Cloud space (~75% of deployments !), see below.
Additionally, Public Cloud providers are extending their deployment models on-premises with certain Hybrid-Cloud offerings (eg. AWS Outposts, Google GCP Anthos/ GKE, Azure Stack / ARC). Also realize that most Hybrid/ Multi-Cloud solutions leverage Kubernetes (K8S). Additionally, one of the most widely adopted solutions today for Private/ Hybrid/ Multi-Cloud is VMware’s vSphere v7 (now includes Containers/ Kubernetes) and VCF v4 (Vmware Cloud Foundation) solution offerings.
However, in the interim we should expect to see most Enterprise Mission Critical, Highly Secure, and/or Extremely High performance workloads (~70%-80+% reside On-Prem today) to either stay On-Premises, or potentially extend to Hybrid/ MC Deployment models (if able to meet compliance/regulatory/SLA/SLC requirements), as the current trends depict.
** See my other blogs for a more thorough examination of Cloud Architecture and Solution Offerings **.
Let us know what you think. Please Comment and/or Retweet Below with any Q’s, thoughts, or requests for future content.
All content and comments made are that of TechThought.org, the author, and not of current nor past employers.