HCI vs Traditional Storage: 7 Common Misconceptions

I’m writing this post because the topic frequently arises in my consulting work. Some of my clients recognize the benefits of Hyper-Converged Infrastructure (HCI), while others only focus on its drawbacks.

Currently, many people perceive HCI technology as less powerful than legacy or Traditional Storage systems.

In this post, I aim to address common misconceptions I usually encounter.

Introduction

In this introduction, I aim to provide essential context for those new to the realm of storage solutions, shedding light on the distinctions between Traditional Storage and Hyperconverged Infrastructure (HCI) Storage.

Traditional Storage or Legacy Storage, as referenced in this discussion, encapsulates all legacy storage systems hosted on storage arrays, typically necessitating a Storage Area Network (SAN).

In this post: this classification excludes Direct Attached Storage (DAS) or standard internal storage managed behind a RAID controller.

On the other hand, Hyperconverged Storage refers to HCI solutions, which revolutionize storage architecture by distributing storage across a unified storage object over IP network.
Redundancy within this solution is achieved through HCI software and the number of host components within the HCI cluster. Notable vendors offering HCI solutions include VMware vSAN/DELL VxRail, Nutanix, CEPH, Dell Power Flex, …

Now we will enter the main part: The Misconceptions

“HCI is less performant than Traditional Storage!”

Contrary to this belief, HCI is not inherently less performant. Initially, when compared to Traditional Storage, HCI might seem less performant.
However, as you scale out your HCI cluster by adding more members, its performance increases; by distributing IOs across the cluster and enabling more compute servers to handle them. HCI becomes more powerful.
Additionally, unlike with Traditional Storage where you’re bound by a guaranteed IO limit, HCI offers flexibility. With HCI, you can enhance IO processing capacity through scaling up (adding disk/CPU/memory/NICs and bonding) or scaling out (adding hosts to your cluster).

“There is less latency on Traditional Storage than HCI!”

In the performance battle, IO processing time is crucial. Latency largely depends on the number of elements between the reader/writer and the location of the read/write operation.

In Traditional Storage, the IO path involves more elements, thus increasing latency.
However, HCI simplifies this process significantly; VMs read and write operations are directly handled by the Hypervisor’s driver on the disks, which minimizes the latency.

“HCI is more complex to manage as the management is not centralized.”

While it’s true that HCI management is typically linked to the cluster, unlike the centralized management of Traditional Storage, this doesn’t necessarily imply greater complexity.

In a Traditional Storage environment, you deal with centralized management for tasks such as allocating LUNs, managing disks, managing RAID, managing pools, managing tiering and masking LUNs.
However, HCI streamlines management by consolidating tasks within the cluster, providing a unified interface for administration and configuration.

“Doing HCI monitoring is a nightmare compared to Traditional Storage”

While it’s a complex topic, in many cases, purchasing legacy storage entails acquiring specific monitoring tools from the manufacturer. This often requires significant investment in tuning the tool or customizing it to meet your monitoring requirements.
On the other hand, HCI typically involves adding management packs or plugins to your existing monitoring tools : you are already monitoring it as you’re monitoring the hypervisor cluster.

For instance, with solutions like vSAN or VxRail, you can simply configure Aria Operations to collect data from the vSAN cluster. Aria Operations can then be set up to send alerts and data to your standard monitoring tools, all of which can be fine-tuned to suit your specific needs.

“The only real use case of HCI is for VDI.”

While HCI indeed offers significant advantages for VDI environments, particularly in distributing IO across multiple hosts to mitigate the morning start “VDI storm,” its utility extends far beyond VDI.
One emerging need where HCI shines is in Cloud Native applications. The challenge of providing persistent storage to Worker Nodes in the Container world is significant.

Traditional storage solutions often require band-aid approaches, such as deploying Low Latency VMs for file storage on block storage, which can be complex and limited in multi-tenancy scenarios. With HCI, however, the solution is much simpler. Most HCI solutions are compatible with Cloud Native Storage plugin, supporting both ReadWriteOnce (block) and ReadWriteMany (file) modes.

HCI can accommodate a wide range of workloads, from small to large VMs, and offers features like compression, deduplication, and encryption to enhance space efficiency and security.

“HCI is less resilient and redundant than Traditional Storage”

This statement stayed true only when comparing individual components. A storage array box may appear more redundant than an HCI node due to its multiple processing nodes/service processors, power supplies, network interfaces, and cards.

However, when considering the overall solution, HCI often proves to be more redundant. This is primarily because HCI distributes IO across multiple nodes, allowing for seamless redistribution and potential storage rebuild in case of node failure.

In contrast, losing a service processor in legacy storage can result in a significant loss of IO processing capacity.

From a resilience standpoint, both solutions can provide stretch solutions with a 0 Recovery Time Objective (RTO).

“HCI drain hypervisor performance”

This statement is often true due to misknowing during HCI deployment.
For traditional storage, the IO Read and Write are usually handled by the CPU and the memory of the storage processor.

In HCI, it’s the Host CPU that handles this workload! During design, allocate powerful CPUs and slightly more memory, reserving about 10% of CPU for vSAN operations.

Remark: by ignoring features like encryption, deduplication, or compression, performance concerns can be reduced.
Additionally, future VMware vSAN versions with Express Storage Architecture (ESA) models and Data Processing Units (DPU) are expected to improve performance further.

Conclusion

I’m quite convinced that HCI, especially VMware vSAN, represents the future of enterprise storage solutions.
These days, I’m noticing a significant shift among customers towards vSAN or HCI, primarily due to their simplicity and built-in automation features.
While some still maintain a small Traditional Storage setup for legacy applications tied to Solaris/Oracle appliances, this trend is gradually fading away as more applications transition to containers.

If you’re considering adopting an HCI solution, I recommend seeking an expert to design a solution that perfectly aligns with your needs.

And if you’re concerned about the complexity of designing it, consider solutions like DELL VxRail. They offer standardized, validated designs that streamline the adoption of VMware vSAN, providing a straightforward path to implementing HCI.

If you have any questions, do not hesitate to contact me,

Cheers Valentin