Can HCI Replace a SAN?

I speak with two to three IT leaders daily looking for a VMware Alternative, and while most are open to using a hyperconverged infrastructure solution (HCI), they rightly wonder, can HCI replace a SAN? There are several very legitimate questions about the storage capabilities of HCI compared to a dedicated storage area network (SAN) or network-attached storage (NAS) system:

1. Can HCI Scale?
2. Is HCI Flexible?
3. Will HCI protect data?

The Advantages of HCI Storage

In theory, HCI storage should have a tremendous advantage over a dedicated storage array, SAN or NAS. First, in theory, it should be significantly less expensive. HCI uses internal server storage. A server-class flash or hard disk drive is typically 5-10X less expensive than the same drive in a dedicated storage array. Second, in theory, HCI should perform better than a dedicated storage array. It has the advantage of data locality, meaning that data is local to the virtual machine (VM) requesting it.

Unfortunately, neither one of these theories manifests itself in today’s HCI solutions. In reality, the cost of an HCI solution from companies like Nutanix or VMware (vSAN, vSphere, NSX) rivals the cost of a three-tier architecture, and performance is sub-par. In addition, traditional HCI solutions fall well short of IT’s expectations for scalability, flexibility, and data resiliency.

HCI vs. SAN – Scalability

There is a belief that HCI is more scaleable than a SAN or NAS because adding additional nodes (servers) to the HCI cluster should increase performance or capacity. There are two reasons why data center reality proves this belief wrong. First, most HCI solutions use a standard IP protocol for internode communication. The HCI’s backend network is responsible for coordinating data placement across the cluster. Unlike a SAN or a NAS, it is also responsible for other cluster synchronization activities, like virtualization and network services. Using standard IP protocols combined with the services’ load creates a storm of east-west traffic. The weight of this east-west traffic is noticeable even when node counts are relatively small (3 or 4 nodes).

The second belief in HCI “hypes” convergence, but it doesn’t actually converge anything. HCI solutions stack the three tiers, via software onto servers, or nodes within a cluster. Each tier, networking, compute, and storage are still present in the form of three separate software packages. These are often from different vendors and have no integration between them. The result is a lot of redundant code and metadata, which leads to relatively poor resource utilization.

HCI vs. SAN – Flexibility

As they scale, HCI solutions aren’t any more flexible than SAN or NAS solutions. Both architectures are, for the most part, locked into using very similar types of storage media. HCI compounds this by also locking in the type of processor. As you scale, it is very difficult to introduce other types of media or processors within the same HCI or SAN instance. Mixing different types of processors, leveraging GPUs, or adding compute-only and storage-only nodes within HCI requires another HCI instance. This brittleness on the part of both architectures means that customers must stand up separate silos for each use case and go through arduous infrastructure refreshes.

HCI vs. SAN – Data Resiliency

Both HCI and SAN solutions lack adequate data resiliency, something that the rise in ransomware attacks exposes in great detail. IT leaders can’t count on the production infrastructure to protect itself, so they must deploy stand-alone backup and high-availability solutions. These secondary solutions dramatically increase data center costs and complexity while introducing “data loss windows” to accommodate time to transfer data and services to secondary systems.

Ultraconverged Infrastructure vs. HCI vs. SAN

An alternative to HCI or SAN is for IT to select an Ultraconverged Infrastructure (UCI). This architecture solves the limitations of HCI, enabling it to deliver on the full promise of a SAN or NAS replacement. VergeIO’s UCI solution, VergeOS, tightly integrates the legacy three-tier network, compute, and storage architectures into a single-tier, cohesive data center operating system (DCOS).

The efficient DCOS uses a proprietary communication method between nodes that intelligently distributes I/O packets minimizing crippling east-west traffic and enabling the linear scale of capacity and performance to hundreds of nodes. Its Virtual Data Center (VDC) technology enables customers to mix nodes of dramatically different types within the same cluster while still controlling how resources are allocated to ensure service levels.

VergeOS is built on a foundation of global inline deduplication, which is the key enabler of its advanced data resiliency capabilities. Global inline deduplication enables IOclone, which clones data instead of using traditional snapshots. Clones are independent copies that have no impact on performance no matter how many of them there are or how long the IT needs to retain them. However, because of global inline deduplication, they are capacity efficient. Lastly, VergeOS uses global inline deduplication to assist in detecting ransomware attacks by monitoring even the slightest changes to deduplication ratios.

In addition to data resilience, VergeOS is innately highly available but not just for storage. All services, including virtual machines and network services, are protected and highly available.

UCI Delivers the HCI Promise

UCI delivers on the HCI promise of lower cost and superior performance without compromising areas where legacy SAN or NAS solutions have an advantage over legacy HCI. By adopting UCI, customers repeatedly reported a total cost of ownership reduction of up to 80% compared to VMware or Nutanix HCI solutions or legacy three-tier architectures built using SAN or NAS.

Best of all, you can try VergeOS without having to scrounge up hardware. You can create your virtual test environment using VergeOS’ VDC technology, including creating VMs, executing clones, and even creating sub-virtual data centers.