This post is part of a sponsored Intel blog post series on Intel Optane. To learn more about Intel, please visit intel.com.
In our previous posts, we covered how VMware VCF delivers more value with Intel Optane solutions, and we also performed an architectural deep-dive of an Optane-enabled VCF environment.
As highlighted before, Intel has built a solution based on the Intel Optane technology which seamlessly integrates with VMware VCF hybrid cloud platform. This integration happens at two layers: Optane PMEM for memory capacity expansion at the hypervisor layer (vSphere), and Optane SSD as a fast cache for the storage layer (vSAN).
Organizations want to reap the benefits of a hybrid cloud operating model: on-demand capacity provisioning, workload mobility, capacity bursting, data protection / long-term retention and disaster recovery purposes.
But they also know that cloud spend can quickly grow out of control and needs to be carefully looked after. On-premises infrastructure still remains the most cost-effective portion of a hybrid cloud architecture.
And this is where the value proposition of VMware VCF with Intel Optane becomes compelling for organizations.
On-Prem Infrastructure Challenges
Current on-premises hybrid cloud architectures are primarily bound by two kinds of bottlenecks: performance bottlenecks at the storage layer and memory capacity bottlenecks at the compute layer.
In current CPU ranges, CPU core count is no longer a limiting factor. Instead, capacity bottlenecks are primarily caused by high-cost, low-density DRAM modules and limited amount of DIMM slots per server.
To ensure optimal memory per core density, organizations would have to invest in 4-socket systems, whose price is more prohibitive than 2-socket systems. And of course, populate DIMM slots with low-density, high-cost DRAM modules.
Hybrid cloud architectures also require a storage layer able to support the most demanding workloads. It must not only provide good storage performance but also sustain I/O demand during peak times and provide consistently low latency. Without an optimized storage architecture, the outcome is that more drives are required to meet the desired throughput and latency requirements.
These bottlenecks lead to horizontal scaling challenges, where more servers are needed to meet the expected capacity and performance requirements. This has a direct impact on CAPEX and OPEX: more hardware needs to be purchased and in turn more software licenses are also required.
This goes hand in hand with additional hardware and software support to be paid for during the entire hardware life cycle. And it doesn’t end there: environmental requirements are also to be taken in consideration (power consumption, additional UPS load, cooling, data center real estate, etc.).
Delivering Value Across the Infrastructure Stack
A VMware VCF hybrid cloud built with Intel Optane solutions truly delivers value across and beyond the infrastructure stack.
Optane PMEM enables significant VM density increase on compute nodes. This enables organizations to meet the same capacity and performance requirements with a lower node count. This leads to a decrease in CAPEX and OPEX: less hardware and licenses to be purchased, lowered hardware and software support costs, lower environmental costs.
At the storage layer Optane SSD acts as a cache. The outcome is predictable and sustained throughput combined with very low latencies, and a reduction in the count of drives needed to achieve those two requirements.
Proper Architecting Is Key
Building a VMware VCF environment with Intel Optane isn’t complicated, but as with every infrastructure it must be properly planned.
Points to take in consideration are:
- What workload type will the VCF domain support?
- Is the workload type core-optimized or frequency-optimized?
- How much usable memory (DRAM + Optane PMEM) a node should have?
- What is the desired consolidation ratio?
In terms of existing configurations, Intel supports nodes with either 1 TB or 1.5 TB of usable memory. Those configurations consist of a DRAM cache of 384 GB RAM, plus either 1 TB or 1.5 TB of PMEM. The PMEM is seen by the hypervisor as the main system memory, while the DRAM acts as a cache for PMEM and is not visible as usable capacity.
Designing around the consolidation ratio also helps mitigate the blast radius of a potential node failure. High consolidation ratios might lead to a higher count of VMs impacted during a node loss.
Thankfully, the presence of Intel Optane SSD technology at the storage layer eliminates the adverse impact of performance bottlenecks during HA events by delivering high-throughput and low-latency, which ensures VMs are restarted without delay on surviving nodes.
Increasing the core count is not always a rule of thumb. A VCF domain with lower-frequency, higher CPU core count processors is optimized for general virtualization workloads, however a VCF domain supporting Oracle databases would better be architected with a frequency-optimized CPU (lower core count, higher frequency) to meet performance demand while keeping licensing costs under control.
Real-world Use Cases
More and more organizations are adopting Intel Optane solutions to improve VM capacity and reduce infrastructure costs. Softbank, a Japanese communications carrier also operating in the Banking and Finance verticals is one of these.
With more than 10,000 VMs used by 45,000 employees and a 1,000+ VM growth year-over-year, Softbank turned to Intel to address those challenges. The switch to servers based on Intel Optane Persistent Memory technology allowed Softbank to deliver up to 3x improvement in VM capacity and 41% cost reduction on the infrastructure supporting Softbank’s communication business.
The good news though is that organizations do not need to be the size of SoftBank to reap the benefits of Intel Optane. An Optane-optimized VCF infrastructure can start delivering value even with the smallest clusters, starting at four nodes only, as can be seen in the diagram below.
Beyond the technical prowess and innovation brought forth by Intel Optane solutions, perhaps the key takeaway from our analysis of VMware VCF solutions built with Intel Optane is about how important it is to adopt a solution-based thinking approach instead of analyzing components in isolation.
Individual components in a solution are certainly essential but looking at them in isolation may end up being counterproductive. That is certainly the case with Intel technologies.
Intel has worked hard to build and end-to-end solution that connects the dots between compute, memory and storage. Creating – and supporting at the CPU level – a persistent memory technology that non-disruptively fits into the x86 architecture has great engineering merit.
But it is the practicality of the synergies enabled by this end-to-end integration that truly delivers value across the stack. At comparable performance, organizations get increased compute power, more memory capacity, better storage performance and higher consolidation ratios with less physical hardware. And lower costs, both from a CAPEX and OPEX perspective.
With Intel Optane solutions, organizations can deliver on the promise of hybrid cloud infrastructure without having to compromise on performance or costs.