4 Min Read
TEAM Cloud frequently shares why TEAM Cloud - sovereign Oracle Cloud Infrastructure (OCI) in New Zealand - is often both cheaper and more performant than other cloud options.
Now, while we stand behind that claim, the reasons can sound “too technical” when you’re trying to explain them to customers, stakeholders, or even your own team.
Here’s a real‑world example we’ve published: Inland Revenue repatriated a content workload from AWS Sydney to TEAM Cloud Auckland, and the case study reports a 48% year‑on‑year infrastructure cost saving, alongside benefits like data sovereignty and local support. [1]
Below are two concepts that explain a big part of the price‑performance advantage - in plain English - plus two extra concepts that are worth knowing.
Picture a data centre as a city full of buildings (servers). The network is the motorway system connecting everything together.
In many cloud environments, those internal motorways are built with some level of shared capacity (oversubscription). Most of the time it’s fine - but at busy times you can get congestion. That congestion shows up as slower responses, inconsistent throughput, and annoying performance spikes.
A non‑blocking network is engineered so that internal links don’t become the bottleneck. The goal is that servers can communicate at line‑rate even when lots of things happen at once.
A simple way to understand it is: split the data centre network into two equal halves and ask, “How much total traffic can flow between the halves at the same time?”
If the answer stays high even under heavy load, the network behaves more like a real motorway system with enough lanes - not a single‑lane bridge.
Oracle describes its network fabric (for large superclusters) as designed to be nonblocking and to offer full bisection bandwidth to all hosts. [2]
When internal networking is unpredictable, teams often compensate by over‑provisioning: more compute nodes, bigger instances, extra buffering, or bigger clusters - all of which cost money.
When internal networking is consistent, workloads often finish sooner and scale more efficiently. That can translate into fewer nodes and fewer billable hours to hit the same outcome.
For tightly coupled HPC/AI and large cluster jobs, TEAM Cloud/OCI’s RDMA cluster networking provides extremely low latency (as low as single‑digit microseconds). [8] Those kinds of workloads can be especially sensitive to “network traffic jams.”
Cloud pricing gets expensive when you’re forced into preset instance sizes that don’t match your workload. It’s like being told you can only buy a 4‑seat car or a 60‑seat bus - nothing in between.
If your workload needs more memory but not more CPU, many menus make you buy extra CPU anyway (wasted spend). If your workload needs more CPU but not more memory, you pay for memory you don’t use (also waste).
OCI Flexible shapes let you select the number of OCPUs and the amount of memory when launching or resizing a VM, and OCI notes that network bandwidth scales proportionately with the number of OCPUs. [3]
In practice, that makes right‑sizing easier and reduces waste: you pay for what you need, then tune as real utilisation data comes in.
Two other concepts that also help explain performance (and cost)
In many architectures, the same machine running your application is also doing a lot of the networking/virtualisation work.
OCI’s design includes moving parts of that network virtualisation stack off the compute host and onto dedicated infrastructure (Oracle describes an “off‑box virtualization device” connected to each host). [4]
Oracle also describes isolated network virtualisation using a custom SmartNIC to isolate and virtualise the network. [5]
Layman’s version: you’re shifting the ‘traffic controller and plumbing’ off the machine that’s meant to run your app — which helps consistency and can reduce overhead.
Some workloads want maximum performance, strong isolation, or simpler licensing. OCI offers bare metal compute instances that provide dedicated physical server access. [6]
Oracle also positions bare metal as giving customers full control, and notes that Oracle installs zero software on its bare metal instances. [7]
Layman’s version: instead of renting an apartment in a building, you rent the whole house.
Non‑blocking networking and flexible VM shapes are two of the easiest “why we win” concepts to understand and communicate.
But they’re not the only reasons TEAM Cloud/OCI can be price‑performant — there are also design choices like off‑box/isolated network virtualisation and the availability of bare metal for workloads that benefit from it.
Ultimately, better architecture means you can often get the same outcome with fewer resources, less waste, and more predictable performance — which is what customers actually care about.
Contact us today to learn more about TEAM Cloud.
[1] TEAM Cloud case study: “Inland Revenue Repatriates Content Workloads from Australia to Sovereign TEAM Cloud” (reports 48% cost savings). https://teamcloud.nz/about/resources/case-studies/inland-revenue
[2] Oracle Cloud Infrastructure blog: “First Principles: Superclusters with RDMA-Ultra-high performance on Ethernet” (states the fabric is designed to be nonblocking and offers full bisection bandwidth). https://blogs.oracle.com/cloud-infrastructure/superclusters-rdma-high-performance
[3] Oracle Cloud Infrastructure documentation: “Compute Shapes” (defines flexible shapes; choose OCPUs and memory; network bandwidth scales with OCPUs). https://docs.oracle.com/en-us/iaas/Content/Compute/References/computeshapes.htm
[4] Oracle Cloud Infrastructure blog: “First Principles: L2 network virtualization for lift and shift” (describes each compute host connected to an off-box virtualization device where the network virtualization stack runs). https://blogs.oracle.com/cloud-infrastructure/first-principles-l2-network-virtualization-for-lift-and-shift
[5] Oracle Security page: “OCI - Isolated Network Virtualization” (SmartNIC-based isolation/virtualisation). https://www.oracle.com/nz/security/cloud-security/isolated-network-virtualization/
[6] Oracle OCI documentation: “Overview of the Compute Service” (bare metal gives dedicated physical server access). https://docs.public.oneportal.content.oci.oraclecloud.com/en-us/iaas/Content/Compute/Concepts/computeoverview.htm
[7] Oracle: “OCI Bare Metal Instances” (notes Oracle installs zero software on bare metal; mentions off-box virtualization). https://www.oracle.com/nz/cloud/compute/bare-metal/
[8] Oracle OCI documentation: “High Performance Computing” (RDMA latency as low as single-digit microseconds). https://docs.oracle.com/en-us/iaas/Content/Compute/References/high-performance-compute.htm