Nutanix web-scale solutions employ a variety of advanced mechanisms to drive the highest possible performance for a wide range of virtualized workloads. Server-attached flash storage is used both as a high-performance read cache, as well as a persistent data tier in a Nutanix hyperconverged system. Performance acceleration capabilities include:
Intelligent Distributed Data Tiering
Application and user data is automatically and intelligently placed in the optimal storage tier — flash or HDD — to yield the fastest possible performance and to optimize capacity in the cluster. The most frequently accessed data (‘hot’ data) is placed in the SSD or cache tier for near instantaneous access by VMs. As data becomes ‘cold,’ it is demoted into the higher capacity HDD tier for more efficient storage. This ensures that SSD and cache capacity remains available for new ‘hot’ data. MapReduce technology is leveraged across all nodes in a Nutanix web-scale cluster, and manages data placement as a distributed, background process to ensure negligible impact to real-time performance.
A core web-scale design principle of the Nutanix platform is data localization. It keeps data proximate to the VM and allows write I/O operations to be localized on that same node. Each Nutanix Controller VM (CVM) continually analyzes whether I/O requests are satisfied by storage resources local on that node, or from another node in the cluster. For example, If a VM migrates to another host for load balancing in an active migration scenario such as vMotion in VMware environments or PRO and live migration with Hyper-V, hot data automatically follows the VM so it maintains the highest performance. After a certain number of read requests made by a VM to a controller that resides on another node, Nutanix data localization transparently moves the remote data to the local controller. The read I/O is served locally, instead of traversing the network. Data localization is managed as a MapReduce job to take full advantage of all cluster resources, and not impact system performance.
Flexible Block Sizes
End-to-end support for a wide range of I/O sizes, from 512 bytes up to 1 MB ensures optimal performance across wide range of workloads and delivers application-tuned metadata efficiency. Unlike traditional storage systems that may need to break up and store large incoming I/O requests into fixed 4KB chunks, Nutanix efficiently handles block sizes of variable length across the entire write path.
VM Flash Mode
For applications that require consistent, fast storage performance with no penalty regardless of data access, Nutanix provides the ability to keep the data in vdisks in the cluster-wide SSD tier for highest performance regardless of data access. VM Flash Mode provides the ability to mix IOPS-sensitive workloads with regular workloads in the same Nutanix cluster without creating resource silos. This feature can be enabled and controlled at VM or vdisk granularity, and provides fine-grained control over storage performance in a hybrid cluster.