When working with clients, we tend to hear a lot of misconceptions around flash and virtualization. One common misconception is than having an All-Flash Array (AFA) is a requirement when architecting a virtual infrastructure.
AFAs certainly have many benefits, such as more consistent performance, higher VM Density, and enhanced deduplication results. But all-flash arrays are expensive and the technology is still relatively new. So is going all-flash really a necessity when designing a high-performing virtual environment?
There are some virtual machines that should always reside on flash. For these VMs, AFAs can be appropriate. But a properly architected hybrid system can provide a similar level of performance.
If there are only a handful of performance-sensitive VMs in the data center, a hybrid array with flash pinning is a more affordable option, while still providing the necessary IOPs. With flash pinning, the most active VMs permanently reside on the flash tier, so performance is never an issue.
VMs that only need high performance occasionally are known as trouble makers. In an AFA, troublemaker VMs will always reside on flash, even though they only need flash-level performance intermittently.
In a hybrid array, on the other hand, they will reside on spinning disks most of the time and only be promoted to flash when their access activity increases. Depending on how long it takes for that VM’s data to be promoted to flash, the user will experience a performance difference.
But for many VMs, the change happens so fast that it doesn’t have a noticeable enough impact to affect productivity. And the entire VM does not need to be promoted to flash, just the parts that are being most frequently accessed. For VMs where promotion time does create a productivity issue, the data can be pinned to flash or pre-promoted.
The key to managing troublemakers is having a robust reporting tool that details when performance spikes occur, and for how long. This allows storage administrators to know which VMs should be locked into flash, and which need to be pre-positioned in flash prior to a spike.
By essentially solving any performance issues permanently, All-Flash Arrays can certainly make life easier for a storage admin. If you have the budget for an all-flash array that meets your capacity requirements, it can be a way to stave off incremental upgrades for years in one fell swoop.
But the truth is, most VMs don’t need anything close to flash-level performance. So if you’re like many organizations who can’t afford flash storage at current prices, investing in a hybrid array with as much flash as you can afford is a suitable substitute.
If you’re considering options for either all-flash or hybrid storage, our experienced engineers can help you weigh your options and find a solution that makes sense for you. Reach out to our team of storage experts or give us a call at (404) 551-4534.