Protecting RAID systems with BBU or SuperCapacitor
This is the best way to protect your drive managers in everyday storage operations.
Internal backup for RAID controller
RAID controllers form a subsystem within a computer that is exclusively responsible for managing the connected drives – including its own processor and even its own memory. Depending on the RAID mode used, the mini computers must process the data to be stored (e.g. using XOR operations for parities under RAID5 and higher) before it is written to the data carriers.
Write jam in the controller
Ideally, the writing should happen in real time, but this is not always entirely successful. Reasons could be:
- Commercially available drives such as hard disks and SSDs do not store their data bit by bit, but in blocks. So the controller must successively collect the data before writing.
- Furthermore, the cooperation between the operating system and the RAID controller plays a role in the cache. Synchronous writes are already acknowledged by the RAID controller to the operating system as written, although they are actually still in the cache at this point.
- In addition, under heavy load, reading and writing from multiple clients is not always instantaneous. Then the controller must work through the tasks in order. For this purpose, the data is in turn stored in the memory or cache of the controller, sorted, processed and finally written.
Murphy’s law comes true
And this is where a problem area opens up: What happens to the data in the write caches, which are often several gigabytes in size, if the power fails? Will the data be lost if not written to the attached drives?
Answer: Yes
If you do not secure the memory, you will lose the data that was waiting to be written. In the worst case – for example with large databases – this results in an inconsistency between the logged writes and the contents of the active database.
Solution
BBU or SuperCapacitor
- A BBU has a docked battery that powers the volatile cache memory for up to 72 hours. Like all Li-ion batteries, they will age and need to be replaced in a maintenance slot after about three to five years.
- A SuperCapacitor works differently, but also provides higher security: With the energy stored in the capacitor, the data is quickly shifted into a non-volatile memory and is thus ready for the next start.
When the server/storage system is restarted, the RAID controller completes its unfinished write operations via the cache saved in this way or informs the operating system that they must be repeated.
Brand names of our BBUs or SuperCaps!
Areca: BBM/FBM (Battery Back-up Modules/Flash-based Back-up Modules)
ATTO: CacheAssure™ Supercap Technology
Broadcom/LSI: CacheVault® Flash Cache Protection
Note
If the backup via BBU or SuperCap is switched on, the drive cache should be switched off. Otherwise, in the event of a power failure, the data in the drive cache may be lost even though the controller cache has been backed up.
Only SSDs with Power-Loss Protection can leave the drive cache enabled.
All-round protection via UPS
A UPS, on the other hand, protects not only the RAID controller cache but also all connected servers and storage systems from power failure. Built-in batteries ensure that the IT systems continue to operate for a certain period of time. Depending on the capacity of the batteries, this can be a few minutes or several hours.
The size depends on your needs: Do you want the server to only complete its writes and shut down securely, or do you want it to continue functioning for as long as possible. By the way, the UPS permanently signals the server whether there is still power on the input side. Thus, the server, for its part, immediately initiates the shutdown procedure if the power suddenly runs out.
