Shock horror, someone spills the beans, the same day they spill the beans....
Not sure this is giving the game away considering its being announced today. In around 1 hour.
Nimble Storage is going to do a Tegile and have all-flash shelves in a new high-end product that will also undercut its main rival, fellow all-flash array startup Pure Storage, or so it claims. Tegile and Nimble both supply networked hybrid flash/disk arrays using operating software expressly designed to use flash optimally, …
The dirty little secret is that all this flash is still connected to a single, legacy disk drive controller. Remember, Nimble is active-passive, so there is only one LSI disk controller feeding all this flash. My guess is that the IOPS specs are generated out of NVRAM. Ask them for large sequential IO data for different sizes of flash and you will see it doesn't scale performance as you add flash. It can't go any faster than the one PCIe slot the controller is plugged into, and IO contention through a single, shared disk controller must be horrible. Bottom line: if flash devices are not on their own PCIe bus then the architecture is bottlenecked and will become obsolete very soon. If you don't believe me stay tuned.
John
Yes, FUD alert, good job, and I'm not trying to hide who I am. An SVP in title, but an engineer at heart. Do the math - a shelf of SSDs behind a 6Gbps SAS drive cable and disk controller, the same controller that's feeding all the disks in the head unit as well. These legacy SSD bolt-ons to disk architectures will go away, it's just a matter of time, and who knows, Nimble may move away from it as well some day.
With regards to the question on why active-active matters, why don't you ask Dell HP, EMC, NetApp and just about everyone else that makes an enterprise class storage solution why it matters, because they are all active-active. Why have half of your resources sitting idle?
John
Good question, Joe. You're right. If an Active/Active array runs at 100% with two controllers online, it’ll run at 50% with one down. With Fusion’s QoS feature you can prioritize performance to workloads based on business priorities (i.e. mission-critical or non-critical). With this, If one controller fails, performance is shifted from workloads where performance isn’t critical at that moment, to the workloads that are mission-critical. Meanwhile, when both controllers are healthy, they’re both chugging away with CPU, RAM and PCIe flash.