4 Component model #
This section describes the various components being used to create a K3s solution deployment, in the perspective of top to bottom ordering. When completed, the K3s instance can be used as the application infrastructure for cloud-native workloads and can be imported into SUSE Rancher for management.
4.1 Component overview #
By using:
Software
Kubernetes Platform - K3s
Linux Operating System - SUSE Linux Enterprise Micro
Compute Platform
Supermicro Supermicro SuperServer
you can create the necessary infrastructure and services. Further details for these components are described in the following sections.
4.2 Software - K3s #
K3s is packaged as a single binary, which is about 50 megabytes in size. Bundled in that single binary is everything needed to run Kubernetes anywhere, including low-powered IoT and Edge-based devices. The binary includes:
the container runtime
important host utilities such as iptables, socat and du
The only OS dependencies are the Linux kernel itself and a proper dev, proc and sysfs mounts (this is done automatically on all modern Linux distributions). K3s bundles the Kubernetes components:
kube-apiserver,
kube-controller-manager,
kube-scheduler,
kubelet and
kube-proxy
into combined processes that are presented as a simple server and agent model, as represented in the following figure:
K3s can run as a complete cluster on a single node or can be expanded into a multi-node cluster. Besides the core Kubernetes components, these are also included:
containerd,
Flannel,
CoreDNS,
ingress controller and
a simple host port-based service load balancer.
All of these components are optional and can be swapped out for your implementation of choice. With these included components, you get a fully functional and CNCF-conformant cluster so you can start running apps right away. K3s is now a CNCF Sandbox project, being the first Kubernetes distribution ever to be adopted into sandbox.
As K3s can be deployed on a single or multiple nodes, the next sections describe the suggested component layering approach.
4.3 Software - SUSE Linux Enterprise Micro #
SUSE Linux Enterprise Micro combines the assurance of enterprise-grade security and compliance with the immutability and portability of a modern, lightweight operating system. The top 4 features are:
- Immutable OS
Immutable design ensures the OS is not altered during runtime and runs reliably every single time. Security signed and verified transactional updates are easy to rollback if things go wrong.
- Security and Compliance
Fully open source and built using open standards, SUSE Linux Enterprise Micro leverages SUSE Linux Enterprise common code base, to provide FIPS 140-2, DISA SRG/STIG, integration with CIS and Common Criteria certified configurations. Includes fully supported security framework (SELinux) with policies.
- Architectural Flexibility
Both Arm and x86-64 architectures are supported so you can deploy edge applications with confidence across multiple architectures.
- Kubernetes-Ready
You can easily combine SUSE Linux Enterprise Micro with the latest cloud-native technologies including SUSE Rancher, Rancher Kubernetes Engine, Longhorn persistent block storage, and K3s, the world’s most popular Kubernetes distribution for use in low resource, distributed edge locations.
As a result, you get an ultra-reliable infrastructure platform that is also simple to use and comes out-of-the-box with best-in-class compliance. Furthermore, SUSE’s flexible subscription model ensures enterprise assurance for any edge, embedded or IoT deployment without vendor lock-in. A free, evaluation copy can be downloaded or if the organization already has subscriptions, both install media and updates can be obtained from SUSE Customer Center.
4.4 Compute Platform #
Leveraging the enterprise grade functionality of the operating system mentioned in the previous section, many compute platforms can be the foundation of the deployment:
Virtual machines on supported hypervisors or hosted on cloud service providers
Physical, baremetal or single-board computers, either on-premises or hosted by cloud service providers
To complete self-testing of hardware with SUSE YES Certified Process, you can download and install the respective SUSE operating system support-pack version of SUSE Linux Enterprise Server and the YES test suite. Then run the tests per the instructions in the test kit, fixing any problems encountered and when corrected, re-run all tests to obtain clean test results. Submit the test results into the SUSE Bulletin System (SBS) for audit, review and validation.
Certified systems and hypervisors can be verified via SUSE YES Certified Bulletins and then can be leveraged as supported nodes for this deployment, as long as the certification refers to the respective version of the underlying SUSE operating system required.
Supermicro servers take advantage of the latest CPU technologies available. The new servers have been shown to produce more work per watt than ever before. Thus, additional workloads can not only be performed in less time, but at a lower cost as well. Supermicro systems can support up to 6TB of memory per socket.
4.4.1 SYS-120C-TN10R Rack Servers #
The Supermicro SYS-120C-TN10R is SUSE YES Certified Hardware.
The SYS-120C-TN10R Rack Servers provide the following attributes:
CPU: Up to 270W and 40 cores
Memory: 4TB DDR4-3200 memory in 16 DIMM slots w/ support of Intel Optane PMEM 200 series
Storage: Up to 10x all hybrid drive bays (NVMe/SAS/SATA) + Flexible internal storage options (dual NVMe M.2 / SATADOM)
Expansion: Up to 2 standard PCIe 4.0 FHHL expansion slots + 2 AIOM for OCP 3.0 NIC; Building block solution for different applications and environment
860W Platinum level redundant PWS
Cost optimized for large volume deployment
Tool-less mechanical design for rapid deployment
Hot-swap storage and PWS for easy maintenance.
IPMI, serial port and service tag for easy management
Compact system design makes no waste of internal space
< 600mm chassis depth
Fully utilized system resource with 12 NVMe, 4 PCIe 4.0 x16 + 2 PCIe 4.0 x8 expansion
Security is top priority
TPM 1.2/2.0, signed firmware, Silicon Root of Trust
Secure Boot, System Erase
FIPS Compliance, Trusted Execution Environment
Balanced architecture between CPUs and optimized for scalable compute, database, GPU, tiered storage and I/O intensive applications
Support open standards like OpenBMC and OCP 3.0
Optimized thermal design
High efficiency Platinum level PWS (AC/DC)
Reduced waste with bulk packaging and customizable accessories
4.4.2 SYS-620C-TN12R Rack Servers #
The Supermicro SYS-620C-TN12R is SUSE YES Certified Hardware.
The SYS-620C-TN12R Rack Servers provide the following attributes:
CPU: Up to 270W and 40 cores
Memory: 4TB DDR4-3200 memory in 16 DIMM slots w/ support of Intel Optane PMEM 200 series
Storage: Up to 12 all hybrid drive bays (NVMe/SAS/SATA) + Flexible internal storage options (dual NVMe M.2 / SATADOM)
Expansion: Up to 6 standard PCIe 4.0 expansion slots + 2 AIOM for OCP 3.0 NIC; Up to 2 FHFL DW GPUs or 6 LP GPUs
Building block solution for different applications and environment
1200W Titanium level redundant PWS
Cost optimized for large volume deployment
Tool-less mechanical design for rapid deployment
Hot-swap storage and PWS for easy maintenance.
IPMI, serial port and service tag for easy management
Compact system design makes no waste of internal space
< 650mm chassis depth
Fully utilized system resource with 12 NVMe, 4 PCIe 4.0 x16 + 2 PCIe 4.0 x8 expansion
Security is top priority
TPM 1.2/2.0, signed firmware, Silicon Root of Trust
Secure Boot, System Erase
FIPS Compliance, Trusted Execution Environment
Balanced architecture between CPUs and optimized for scalable compute, database, GPU, tiered storage and I/O intensive applications
Cost and performance optimized down to component level
Support open standards like OpenBMC and OCP 3.0
Optimized thermal design
High efficiency Titanium level PWS (AC/DC)
Reduced waste with bulk packaging and customizable accessories
A sample bill of materials, in the Chapter 9, Appendix, cites the necessary quantites of all components, along with a reference to the minimum resource requirements needed by the software components.