Next up at Storage Switzerland’s EMC Briefing Day was the Midrange part of EMC’s Enterprise & Midrange Storage Division, also known as the VNX family; and of course we are going to talk about the VNX2. The EMC VNX2 is the workhorse of the EMC family designed for the mid-range data center looking to support a variety of workloads. The VNX2 is a hybrid array that was, according to EMC, designed from the ground up to leverage flash. Essentially, the original VNX was flash capable, the VNX2 is flash optimized.
Flash Capable vs. Flash Optimized
What’s the difference between Flash Capable and Flash Optimized? While both VNX and VNX2 add flash by adding drive form factor solid state disk drives, the VNX2 software has been redesigned to leverage multicore CPU’s so that it can match the high speed devices with high performance software services.
As we describe in our article, “What is a Storage Controller?”, a storage controller is the engine that drives an array. It provides the processing power to deliver functions like RAID, volume management, snapshots, tiering, etc… The CPUs on these storage controllers are typically all Intel multicore processors.
The original VNX could only leverage those cores by locking a particular function to a particular core. The VNX2 can stripe functions across cores so that the combined processing power of the chip can be applied to all functions. In addition to supporting up to 32 processor cores, the VNX2 can provide up to 160 PCIe Gen 3 lanes for the storage interconnect and leverage 6Gbs SAS for connectivity.
When we use the term Hybrid Array, we typically mean a system with Flash and HDD storage. The VNX2 has at least four tiers of storage performance that it can automatically tier data between. There is a mirrored battery backed DRAM area that can be as large as 128GB, which is ideal for write acknowledgment. There is an up to 4TB SLC based flash cache area which is ideal for de-staging the DRAM writes and acting as a high performance READ cache. Then there is a MLC based flash tier for active primary data and there is an HDD tier that can be either SAS or SATA based for data at rest.
The system is intelligent in the use of these tiers. For example, it knows not to cache data that is already on the MLC tier. This allows the flash cache area to focus on accelerating data being served up from the HDD tier.
Also important is that the HDD tier can still leverage high performance hard drives, that way if there is a cache miss, there are enough high speed spindles to still deliver acceptable performance to mission critical applications. Too often, hybrid arrays only provide high capacity near line HDDs making I/O performance during a cache miss a potentially painful experience.
Storage Swiss Take
If IT planners are going to leverage a Hybrid Array, they should give close consideration to how it is configured. The point of these systems is to balance performance and capacity for an effective cost per GB. The problem is that many hybrid storage systems don’t provide the configuration flexibility required to minimize the chances and the impact of a cache miss. Failure to provide this flexibility limits the ability of a hybrid array to address the wide range of workloads it is designed to support.