According to scientists at MIT, the blink of an eye can be less than 100 microseconds. With the latest release of Pavilion Data’s NVMe-oF Storage Platform, you could blink twice in the time it takes for a 128K block of data to write from a host, across an NVMe-over-Fabrics network, through our array and to a RAID-6 volume of OpenChoice NVMe SSDs.
This extreme low-latency, combined with write bandwidth of 90GB/sec. redefines composable disaggregated storage and brings into question why anyone would use Direct-Attached Storage for modern scale-out applications, ever again.
Of course, an NVMe SSD inside a server specs out at 20 microseconds of latency. But that same SSD can deliver only 2GB/sec write throughput. Another reason to disaggregate.
Now, consider the hidden gem DAS Aggravation or Disaggregation that our VP of Product, Jeff Sosa, posted on this same blog site, April 16th. Jeff thoughtfully describes a huge issue for DAS deployments – node recovery.
As drive-makers inevitably increase the capacity of NVMe SSDs, the time to recover a failed drive/node increases proportionally with drive size. At best, it takes a single server 25 minutes to recover a 2TB drive. Add network traffic for sharded nodes to push data to the recovery node (slowing application performance) and you have the next big issue for DAS! Our RAID-6 with SWARM Recovery is a game changer. By harnessing the power of up to 20 storage controllers, we can recover that same 2TB drive in less than 5 minutes per TB with zero network impact!
Pavilion Data is moving fast. Today we announced new product enhancements and blazing write throughput of 90GB/sec. There’s more coming, so check back frequently. Don’t blink, or you might miss the news!