Infinidat has doubled its all-flash SSA II array’s capacity, added entry-level models that can scale up, and enabled the SSD cache in its hybrid flash/disk InfiniBox array to function as an embedded all-flash array.
Infinidat’s hybrid InfiniBox array is primarily disk-based and has a flash cache with Neural Cache software that loads read data into memory data for the lowest possible latency access – down to 35 microseconds. SSA Express software turns the hybrid InfiniBox flash cache into an internal all-flash array (AFA). The vendor also supplies an F4308T SSA II all-flash array, introduced in April 2022 with up to 2.624PB of effective capacity. The F4316T variant has doubled this to 6.64PB of effective capacity in a 42U rack.
CMO Eric Herzog said: “SSA Express software reimagines the InfiniBox hybrid platform, ensuring that critical applications and workloads have rapid, low-latency response rates through direct access to the flash layer, while simultaneously reducing costs and simplifying storage management.”
In effect, with the SSA Express software, applications can be pinned into the flash cache. Admins can select specific datasets, applications, and workloads to reside in the SSD layer of the InfiniBox hybrid with near 100 percent read cache hit rate. This eliminates the need for enterprises to buy a separate AFA to support smaller apps and workloads that require high performance and low latency. These apps and workloads can be consolidated into the InfiniBox.
The SSA Express software is part of the InfuzeOS v7.3 release and comes with no additional charge. It provides up to 320TB of usable all-flash capacity and 95 percent of all existing InfiniBox systems are supported. Additional flash capacity can be purchased if needed, and installed non-disruptively. Note that data efficiency services are not supported with the SSA Express software.
The entry level to the InfiniBox SSA has been lowered with 60 and 80 percent capacity populated models. As storage needs progress, these models can scale up to 100 percent capacity – the 60 percent model scaling up to 80 and then 100 percent.
Scott Sinclair, ESG practice director for Cloud, Infrastructure and DevOps, was impressed, saying: “It’s a brilliant move that opens up the InfiniBox hybrid to a wider range of enterprise applications and workloads.”
Larger SSA II array
The larger capacity 4316T SSA II array accompanies the prior 4308T and the capacity options are tabulated in an Infinidat blog by Tim Dales:
Capacity upgrades are non-disruptive. The larger capacity comes with no increase in electrical power requirements, meaning up to 50 percent less power is needed per TB or floor space. Dales says Infinidat SSAs are more environmentally friendly (use less electricity) than competing arrays and has a calculator available here for you to test how non-Infinidat arrays compare.
SSA Express software will be available in Q4 2023. The larger InfiniBox SSA II all-flash array is now shipping.
The transformation of a hybrid array’s flash cache into an embedded or virtual all-flash-array is one of those moves that, with hindsight, seem obvious. In fact it has been done before. Hyperconverged system supplier Nutanix has been effectively doing this for years with a pinning feature or flash mode. Using “flash mode for VM allows admins to set the storage tier preference to SSD for a virtual machine or volume group … By default, you can use up to 25 percent of the cluster-wide SSD tier as flash mode space for VMs or VGs.”
HPE-acquired Nimble Storage provided a pinning feature. An HPE Infosight webpage states: “Volume pinning allows you to keep active blocks of a volume in the cache, as well as writing them to disk. This provides a 100 percent cache hit rate for specific volumes (for example, volumes dedicated to critical applications), and delivers the response times of an all-flash storage system.”
An Infindat spokesperson said: “While Infinidat is leveraging pinning capability, the unique difference from past approaches is Infinidat’s patented Neural Cache software that loads read data into memory data for the lowest possible system and application latency.”