The Need for Efficiency
Enterprise-class storage platforms have become overly complex:
Having worked with storage for many years both as designers and users we realized that the complexity and cost curves of storage could not be sustained. Much of the problem was the result of the patch-work of band-aid solutions applied over the years by storage vendors. There certainly was no purely hardware solution. An entirely new approach was required that allowed the hardware and the software to complement each other not just co-exist.
We love storage and enjoy a challenge so we set to work on a new approach that would maximize scalability, performance, simplicity and be cost-effective. We saw three areas that would produce the greatest benefits and made them our initial targets. We are just getting started and have many more ideas in the pipeline.
- Complexity:
From the moment your new storage platform arrives its complexity makes you realize that you are not in control.
-
- Complexity in installation:
-
When was the last time your were able to install your enterprise-class storage system yourself? And once it was installed with the vendor's help, were you prepared to install the next one yourself? Your goal was to get the system up and running ASAP but first microcode updates had to be applied then apply software patches had to be applied and hours were spent formatting the disk drives for RAID.
- Complexity in configuration:
-
Configuring your storage platform has also become rocket science. You are exposed to all the technical compromises that your vendor made over the evolution of its product line. Your are required to make decisions on arcane technical parameters whose repercussions are only understood by the designers of the platform.
- Resiliency:
Keeping your data safe and accessible is paramount. Numerous options for resiliency are offered: all have drawbacks and few are inexpensive.
-
- RAID:
-
RAID comes in many levels but only guards against disk drive failures. At its most expensive RAID offers replication of your disk drives. Most of the RAID levels can only survive a single failure, though others can survive two failures. You're aware that disk drives are not the only hardware component that can fail, but are you informed of how long it takes to rebuild a RAID array once the failed drive is replaced? Are you aware that for that entire period of time (hours, maybe days) the system is highly vulnerable and (except for RAID 6) will not survive another failure?
- Cross-strapping, active-active, active-passive:
-
These fault-tolerance solutions are highly expensive yet they are only as strong as their weakest link. A faulty controller in one system at the same time as a fault in the CPU or RAM or NIC of the redundant system will still take down the storage system.
- Scalability:
When you ordered your storage system you knew it had limits to its scalability.
-
- Scalable capacity:
-
There are only so many disk drives per shelf and the system can only accommodate so many shelves. After this saturation point you must look for a new system. The saturation point looms over the horizon faster every year. Even worse the process of migrating your data to a new storage platform is usually difficult, disruptive, and time consuming.
- Scalable performance:
-
Disk drive density grows every year but I/O performance does not keep pace. Most high capacity storage systems with a large count of disk drives are unable to move data as fast as the drives could sustain due to limitations of the bus, NICs, etc. The concept of a fat node as a standalone repository does not lend itself to high performance or high resiliency.