Understanding the Changing Economics of Storage
Before addressing the VMware vSAN total cost of ownership (TCO) question, or building a business case for vSAN, it is important that we cover how the storage market is in the midst of serious change and realignment.
Twenty years ago, servers were installed with local attached disks, which resulted in massive islands of storage, poor utilization of resources, and complex operational management. As a result, the industry gradually moved to a shared storage model with SAN and NAS arrays, which was largely accelerated by the increased adoption of virtualization. Though this significantly improved the TCO, these traditional storage arrays have become a large part of IT infrastructure costs and have struggled, in many cases, to continue to meet the performance and feature requirements of the modern virtualized application.
vSAN enables both hybrid and all-flash architectures. Therefore, one of the key design and economic factors in adopting vSAN is hybrid configuration versus all-flash architecture, shown in Figure 4.
Figure 4. Hybrid Versus All-Flash Configuration
As shown in Figure 4, irrespective of the architecture choice, there is always a flash-based caching tier, which can be configured from flash devices such as SSDs, PCIe cards, NVMe, or Ultra DIMMs. This flash caching tier acts as the read cache/write buffer in a hybrid configuration, which dramatically improves the performance of storage operations. In an All-Flash configuration, this caching tier acts as a write buffer only, to provide a layer of endurance flash, with all reads typically being sourced directly from the persistent storage flash layer.
In the hybrid architecture, server-attached magnetic disks are pooled to create a distributed shared datastore, which persists the data. In this type of architecture, you can typically get up to 40K IOPS per server host. However, in an All-Flash architecture, the flash-based caching tier is intelligently used as a write-buffer only, while another set of SSDs forms the persistence tier to store data. Since this architecture utilizes only flash devices, it can deliver extremely high IOPS of up to 100K per host, with predictable and low latency across the vSAN datastore.
A very common question from service providers is “Have we reached that cost tipping point for SSDs versus HDDs for persistent storage yet?” At the time of writing, for most use cases the answer is yes, particularly when you factor in the impact of space efficiency technologies, such as erasure coding and deduplication and compression, which are only available in an all-flash configuration. With the falling prices of flash and the space efficiency technologies accounted for, the All-Flash value proposition for persistent storage is often far more compelling than hybrid. However, that does not mean that All-Flash is always less expensive. A hybrid configuration still maintains a place for some vSAN storage use cases. The key cost factors that must be considered include:
• On a $/GB basis, SSDs are more expensive
• On a $/IOPS basis, HDDs are more expensive
• When building a system, you must consider:
o Usable versus raw capacity (the impact of erasure coding/RAID)
o Space efficiency from dedupe and compression
o Pricing differences between disk vendors
o Power savings
o Reliability and availability (annual failure rates)
In summary, key factors that are pertinent to the TCO of storage and to building a business case for a vSAN hyper-converged model include:
• We are seeing server flash become far more viable in terms of cost and cost per GB/IO
• The cost of rotating magnetic disks continues to fall
• The capacity of flash and magnetic disks continues to increase
• We find we have far more CPU cycles in servers, and as a result we can implement a much higher level of functionality in software instead of hardware
Finally, we are also seeing object storage and cloud economics significantly influence the overall cost of storage.
