One of the key components of SimpliVity’s hyperconverged infrastructure solution is data efficiency: deduplication, compression and optimization of data before it hits the disk. Inline deduplication and compression are critical to providing efficient usage of hyperconverged infrastructure resources. When the processing is done inline, the amount of data being written to disk can be dramatically reduced. Eliminating the write (or the read) before it impacts the HDD is one more way SimpliVity improves efficiency. Although they are processor-intensive activities, SimpliVity is able to provide hardware-assisted deduplication and compression inline with zero impact to performance.
To achieve deduplication, compression and optimization of data at inception, more processing power is required, as well as the ability to persist the data quickly in its lifecycle (to offer very low latency). SimpliVity could have loaded up its solution with more processing power (i.e., Intel processors); however, that would likely impact per-core licensing for the hypervisor and any other applications licensed for the hyperconverged platform.
OmniStack Accelerator Card
Instead, SimpliVity offloads processing to a specially designed PCIe OmniStack Accelerator Card—a standard component in its hyperconverged infrastructure. The card offers specialized processing power (FPGA and lots of SimpliVity’s deduplication, compression and optimization algorithms). In addition, the accelerator card enables very low write latency given the NVRAM module that’s on the card. All data written to the OmniCube hyperconverged infrastructure datastore first passes through the accelerator card. Deduplication and compression happens once and data remains in its optimized state across all stages of the data lifecycle. There is no need to “rehydrate” data.
The accelerator card delivers the required processing power, but without high costs. It also leaves the Intel CPUs to run customers’ business applications. Any technology that does this processing on the Intel CPU is competing with the business applications and consuming a portion of customers’ hypervisor licensing, as well.
The card contains flash and is also protected by super-capacitors to allow DRAM to be saved in the event of a power loss, making it extremely reliable. It is important to note that the OmniStack Accelerator Card never contains the only copy of a piece of data within a data center. SimpliVity can handle the failure of an entire hyperconverged building block and can continue to serve data at a level of performance that allows mission-critical enterprise applications to continue to run.
Benefits of Inline Deduplication, Compression and Optimization
A key result of real-time inline deduplication and compression is that SimpliVity has data granularity at the foundation. Within a federation (all instances of hyperconverged infrastructure within and across data centers that are managed from a single interface as one massively-scalable resource pool), all data elements are between 4KB and 8KB in size. The data in the hyperconverged infrastructure environment is thus “born” to be mobile from the beginning, and remains so throughout its lifecycle within the SimpliVity hyperconverged infrastructure federation.
Within a given hyperconverged system, the data is processed once and every tier of the system benefits from this more efficient data. This is in contrast to post-processing deduplication that writes data in its original state and later introduces additional read/write IO to analyze the data for redundancy.
Inline deduplication, compression and optimization is a core tenant of SimpliVity’s platform—since its initial design—while competing hyperconverged infrastructure vendors have attempted to engineer deduplication into the system after the fact. Inevitably, competing solutions force their customers to make tradeoffs (either deduplication or compression; post-processing dedupe; dedupe only across certain media types, etc.).
Next, the series examines the benefits of SimpliVity’s data efficiency.
Learn more about the enabling technology for data efficiency in SimpliVity’s hyperconverged infrastructure.