Cache and CPU Power
Unique Cache Architecture
Unlike any other storage system, Nextra's design embeds the read/write cache into the same hardware module as the disks. This has distinct advantages:
- Distributed cache. The cache is implemented as a distributed cache, so that all cache units can
concurrently serve host I/O and perform cache-to-disk
I/O. This ensures that the cache will never become a
bottle-neck.
- High cache-to-disk bandwidth. Aggressive pre-fetching is enabled by the fact that cache-to-disk bandwidth is
the internal bandwidth of a module. Implemented over two PCI-X buses, a module thus provides dozens of gigabytes per second for the whole system.
- Powerful cache management. The unique design
enables the Nextra system to read a large cache slot
per disk read, while managing least-recently-used
statistics in small cache slots. This dual capability is made possible by the combination of Nextra's huge processing
power and high cache-to-disk bandwidth.
Massive CPU Power
The fact that each data module has its own processing power provides the Nextra system with dozens of powerful CPUs. This
power is used for sophisticated caching algorithms,
support for small cache slots, snapshot performance, and so on.
The massive CPU power ensures high performance through higher cache hit rates and smaller snapshot overhead.
