With the rise of PCI Express (PCIe) on servers, many IT organizations have started to plug Flash memory into servers to replace hard disk drives as either forms of tier zero or tier one primary storage. Now A3Cube wants to make it possible to treat all that storage as a shared cluster across multiple servers.
With the introduction of Fortissimo Foundation today, A3Cube is making available software needed to create clusters using storage devices that are plugged directly into a PCIe bus on multiple servers.https://o1.qnsr.com/log/p.gif?;n=203;c=204663295;s=11915;x=7936;f=201904081034270;u=j;z=TIMESTAMP;a=20410779;e=iA3Cube CTO Emilio Billi says, Fortissimo Foundation provides the software needed to create a scale-out architecture that eliminates any dependencies on a single node to act as a metadata server for the cluster. Fortissimo Foundation is designed to run on top of the RONNIEE Express data path accelerator and RONNIEE NIC from A3Cube that provide the physical bridge between all the memory and storage devices connected to those servers.
In combination with one another, Billi says IT organizations can set up terabytes of clusters that run in memory at price points that start around $1,000 per PCIe card. Included in that capability, says Billi, is support for in-line data deduplication, compression and data replication that all run in memory.
Billi says A3Cube is essentially providing a low-cost way to create a storage fabric capable of supporting multiple types of devices without requiring IT organizations to invest in separate storage arrays that are not only significantly slower, but usually rely on some combination of external Ethernet and internal Infiniband fabrics. In the case of A3Cube, Billi says a massive global naming space allows IT organizations to rely on single fabric to eliminate all the overhead associated with indexes stored in metadata servers that other approaches need to keep track of which files are being accessed.
With the advent of multiple forms of memory as primary storage, continued reliance on storage architectures that were originally designed for magnetic storage doesn’t make much sense. It’s too early to say what storage architecture will emerge as a replacement. But the one thing that is for certain is that with the rise of in-memory databases, for example, the time to think about primary storage differently is now at hand.