VMware Cloud Foundation with Intel Optane – an Architectural Deep-Dive

This post is part of a sponsored Intel blog post series on Intel Optane. To learn more about Intel, please visit intel.com. First post is available here: VMware VCF Delivers More Value with Intel Optane Solutions

Hybrid Cloud is the new standard in infrastructure deployment and consumption. With VMware Cloud Foundation (VCF), VMware builds a bridge between the on-premises private cloud environments and the public cloud, while delivering the benefits of the Software-Defined Data Center (SDDC): policy-based driven capabilities, resource abstraction and automation.

Even though VCF remains the best way to implement and deliver a true hybrid cloud experience, organizations may face efficiency and cost challenges when the solution is architected as-is.

Intel can help organizations achieve better efficiency and a lower TCO when their technologies are used as the foundational stones of a VCF environment. Intel Optane will be particularly looked at, but the new 3rd Gen Xeon Scalable processors will also be touched upon.

To understand how Intel Optane helps deliver better value, we should first look a bit further into the VCF architecture. VCF consists of three building blocks that have been traditionally operated as independent silos before the SDDC era:

  • Compute
  • Storage
  • Network

VCF abstracts those infrastructure silos by providing the customer with an easier way to manage the platform as a whole.

Figure 1 – Intel Optane Solutions in the VMware VCF context

Public Cloud Providers like AWS, Azure and GCP provide their worldwide infrastructure as a service to customers. VMware on the other hand provides its customers a solution that enables them to create a similar service as the cloud providers, but in their own datacenters. More than that, they even provide the tools to incorporate the public cloud services as well.

We’ll look at the key infrastructure components that define VCF and the added value of Intel Optane technology.

The VCF Compute Layer

VCF’s compute component is the VMware vSphere hypervisor. This component relies on CPU and memory resources to execute workloads. There is an interdependency between those two resources: workloads require computational power (CPU cores) and memory.

New processors (especially the recently introduced 3rd Gen Xeon Scalable architecture) have a significantly higher core density than the previous generations. A high core count leads to an increased VM density on the hosts, which in turns leads to a higher memory demand per server.

Beyond this correlation, memory demand is on the rise: not only normal applications increasingly demand more memory, but the footprint of memory-intensive workloads is significantly growing in the enterprise space.

Intel Optane Persistent Memory (Optane PMEM) can cost-effectively address these additional memory needs for systems requiring 1 TB of memory or more. The Intel 3rd Gen Xeon Scalable CPU architecture is able to drive up to 6 TB memory per socket with Intel Optane 200 Series, ensuring that even the most demanding requirements are met.

Figure 2 – A brief overview of Intel Optane PMEM 200 Series performance improvements.

Workload Domains

One of the key architectural differences between VMware vSphere and VCF is the concept of workload domains. Workload domains ensure automated workload placement based on business or technical policies.

Organizations can rely on a generic workload domain for standard workloads, and create additional workload domains optimized to address architectural, performance or licensing constraints. Such is the case with specific applications such as Oracle and WebLogic, where the licensing model leads architects to prefer CPUs with lower cores and a higher frequency.

The broad range of Intel Xeon processors allows the design of efficient workload domains that help organizations mitigate licensing costs while delivering the expected performance requirements. All while making sure that memory demand is met with Intel Optane.

Figure 3 – The new 3rd Gen Intel Xeon Scalable delivers up to 20% higher performance/core, but also more memory channels to further increase the memory density with Intel Optane PMEM 200.

Storage (vSAN) and Optane

vSAN provides storage to the VMware environment by leveraging, aggregating and pooling the capacity of local-attached Solid-State Drives (SSD) present in each ESXi physical host. The vSAN architecture consists of a cache tier (which accelerates write operations and provides caching of “hot” data for read operations), and a capacity tier, where the data is stored persistently.

SSDs usually consist of 3D NAND memory cells. These cells can be either TLC (Triple Layer Cell) or QLC (Quadruple Layer Cell). TLC has a better endurance and performance than QLC drives, but they are also more expensive. QLC on the other hand as a rather mediocre endurance but provides better capacity. Unfortunately, neither are adequately suited to sustain the high amount of write operations needed for a cache tier.

Intel Optane SSD provides very low latencies and very high endurance, making it best suited to operate in the cache tier. Writes are acknowledged with sub 10 microsecond latencies, and endurance is 10 to 20-fold the endurance of a best-in-class TLC 3D NAND SSD.

Figure 4 – An overview of the Intel Optane SSD P5800X model, gains compared to 3D NAND drives, and capabilities. The central part (Caching) shows the density improvements with VMware (also applies to VCF).

VMware Networking and Optane

While the NSX layer is not directly impacted by Intel Optane, the presence of NSX in the VCF stack implies that compute and storage resources will be drawn from the resource pool to satisfy NSX requirements.

There is also a correlation between the network speed and the impact on the storage layer: increasing the network interface speed from 10 GbE to 25 GbE puts additional pressure on the storage layer.

Providing a sizeable pool of usable memory and consistent performance on the storage backend alleviates the pressure from management components on the entire VCF stack. This in turn provides better outcomes for customers, as ample memory and bandwidth remains available for business workloads.

Although this blog post is focused on Intel Optane solutions, it’s worth mentioning that organizations can benefit from synergies in Intel’s solution portfolio. Xeon processors have specific network acceleration optimizations, Intel ethernet adapters can also offload some of the operations and thus alleviate even further the overhead on the NSX layer of a VCF infrastructure.

Figure 5 – Intel Ethernet 800 Series NICs are part of the broader Intel product ecosystem and also include various optimizations and offloaded capabilities to free up CPU usage for application workloads.

Conclusion

Proper architecting is key when building a platform such as VCF, because of the sheer implications of the overall design on performance, efficiency and total cost of ownership.

With Intel solutions, organizations can achieve better value across the entire VCF stack:

  • Intel Optane PMEM helps achieve significantly higher memory densities
  • Intel Optane SSD bolsters the vSAN storage layer by providing best-in class caching capabilities.
  • Intel 3rd Gen Xeon Scalable models allow better consolidation ratios and enable organizations to build workload domains that closely fit their requirements

From a CAPEX perspective, Optane PMEM helps keep memory costs under control and achieve densities that would otherwise possible only by scaling out i.e., by adding more servers, with a detrimental cost impact. Xeon processors and their increased core count also help consolidate the server footprint. This in turns reduces licensing costs for VCF and other applications.

From an OPEX perspective, organizations will achieve cost savings in terms of support (both for hardware and software licenses) as well as environmental cost savings (datacenter space, power, cooling, etc.).

The VMware Cloud Foundation platform delivers an end-to-end hybrid cloud experience to organizations, across compute, storage and network layers. Intel’s end-to-end vision of the modern datacenter and the breadth of its portfolio enables them to consistently deliver better value on each of those VCF layers.