Pure Mystery Series – PART 1: Pure Storage Mysteries…Demystified…NVRAM
Hello Everyone!!! Starting another series…..Pure Storage Mysteries…Demystified!!
Each day we have the honor of meeting with customers to discuss Pure Storage awesomeness. The questions we receive during technical deep dive sessions are amazing! The interest and excitement from potential customers really makes me happy to be a Puritan. That said, I wanted to start a new series to demystify the top questions we have hear from our customers. This will be a weekly session in which a new question will be reviewed. Feel free to write me if you would like me to cover a specific question in future blog posts. I can be reached at firstname.lastname@example.org.
Let’s dive right in……..
Question 1: “What is the story with your NVRAM outside of the controllers? Is there a battery backup or standby power supply hidden somewhere?”
While most arrays today require battery backups for NVRAM or standby power supplies for controllers to de-stage data to disk during a power failure, our system is different. We are super unique in that we persist all data before acknowledging the write back to the host. That said you can pull the power on Pure and not have to worry about anything.
So how does it work with Pure Storage? Great question……As an IO traverses the fabric into any port of the active/active Pure Storage controllers it is then buffered into DRAM and then mirrored to two NVRAM devices in the shelves. Once completed, a write acknowledgement is sent back to the host to validate data is safe, as the data is now fully persisted. There is no need for any battery backup or standby power supply. Awesome!
Why is the approach better? Well, since most solutions do not persist their data before acknowledging the write back to the host, battery backup and standby power supplies are a requirement. Unfortunately, battery backups fail, and typically at the worst possible time. Standby power supplies take a ton of rack space and need to be monitored to ensure integrity, and even so will most likely fail over time. At the end of the day, you have enough stuff to manage, and we believe that these types of complexities are really a thing of the past anyways…..
So, putting the NVRAM in the shelves also added a ton of benefits for us:
- Completely Stateless Controllers: All configuration information and data in transit is stored and mirrored on the NVRAM devices. That said, you can lose a controller, even lose both controllers, and you can replace them and be back up and running in minutes. Any state held in DRAM within the controller can be replayed from the data within NVRAM when the system comes back up. Simple, easy, and nothing for you to manage or worry about.
- Staging For Efficiency Technology: NVRAM allows us to complete dedupe, compression, pattern removal, and thin provisioning before the data is written to the cMLC SSDs.
- Buffers Writes Away from the SSDs: Instead of using expensive eMLC SSDs we build our system from the ground up on cMLC (consumer based SSDs), which are really the same SSD as eMLC, minus the high cost, the massive over-provisioning (flash you pay for but cannot use), and larger CPUs. With NVRAM, we have the ability to provide the SSDs data in the format that does not require the additional over provisioned capacity, massive CPU complex, etc……this would be a waste of your money. Overall, the less expensive flash we successfully deploy in our solution, decreases the cost of our solution.
- No Need for Battery Backup or Standby Power Supply: Batteries fail and go bad…..With our Pure Storage data is safe and persisted as soon as it written and mirrored to NVRAM. #StorageSweetness
With Pure Storage, simplicity is built into all layers of our technology. Forget about all the challenges you deal with today in your traditional disk architecture. You can get ALL FLASH FOR THE PRICE OF DISK……today!!! Pretty sweet solution!!
See you next mystery! Stay flashy my friends!