![](bg6.png)
ables significantly higher performance of file access
than using the block device framework by removing
the overheads of the framework.
5 RELATED WORK
eNVy (Wu and Zwaenepoel, 1994) proposes NVM
storage based main memory. It assumes the utiliza-
tion of NOR flash memory, which is byte address-
able and of which read access latency is compara-
ble to DRAM. Moneta (Caulfield et al., 2010) is a
storage array architecture designed for NVM. While
its evaluation revealed the necessity of reducing the
software costs to deal with block devices, it does not
consider the removal of the block device interface.
(Yang et al., 2012) also investigated the software costs
to deal with block devices on the premise that PCIe
gen3 based flash storage devices will become even
faster. While it proposed the synchronous interface
to block devices, it is still based on the block device
interface. (Tanakamaru et al., 2013) proposed a stor-
age state storage device that is a hybrid of flash mem-
ory and ReRAM. Their hybrid architecture is simi-
lar to the NVM storage based main memory, they do
not discuss the interaction with a file system. (Meza
et al., 2013) describes the idea to coordinate the man-
agement of memory and storage under a single hard-
ware unit in a single address space. They focused
energy efficiency, and did not propose any software
architecture. (Condit et al., 2009) and SCMFS (Wu
and Reddy, 2011) proposed the file systems that were
designed to be constructed directly on NVM. They,
however, have no consideration of hybrid storage ar-
chitecture.
6 SUMMARY AND FUTURE
WORK
Non-volatile memory (NVM) storage is becoming
more popular as its performance and cost efficiency
improve. High performance NVM storage can expand
to become NVM storage based main memory since
polling based access is faster and its hardware can be
simplified. This paper proposed constructing a file
system directly on NVM storage based main memory
in order to overcome its drawbacks when used simply
as main memory. The performance projection of the
proposed architecture is that accessing files on such
a file system can reduce the overhead introduced by
handling block devices.
We are currently developing a simulation environ-
ment of NVM storage based main memory based on
a virtual machine. By employing a virtual machine
environment, the simulation of NVM storage and its
interface with the DRAM buffer becomes possible.
It also enables more precise performance evaluation
with realistic benchmark and workload.
REFERENCES
Caulfield, A. M., De, A., Coburn, J., Mollow, T. I.,
Gupta, R. K., and Swanson, S. (2010). Moneta: A
high-performance storage array architecture for next-
generation, non-volatile memories. In Proceedings
of the 2010 43rd Annual IEEE/ACM International
Symposium on Microarchitecture, MICRO ’43, pages
385–395, Washington, DC, USA. IEEE Computer So-
ciety.
Condit, J., Nightingale, E. B., Frost, C., Ipek, E., Lee, B.,
Burger, D., and Coetzee, D. (2009). Better i/o through
byte-addressable, persistent memory. In Proceedings
of the ACM SIGOPS 22nd Symposium on Operating
Systems Principles, SOSP ’09, pages 133–146, New
York, NY, USA. ACM.
Josephson, W. K., Bongo, L. A., Li, K., and Flynn, D.
(2010). Dfs: A file system for virtualized flash stor-
age. Trans. Storage, 6(3):14:1–14:25.
Meza, J., Luo, Y., Khan, S., Zhao, J., Xie, Y., and Mutlu, O.
(2013). A case for efficient hardware-software coop-
erative management of storage and memory. In Pro-
ceedings of the 5th Workshop on Energy-Efficient De-
sign (WEED), pages 1–7.
Tanakamaru, S., Doi, M., and Takeuchi, K. (2013). Unified
solid-state-storage architecture with nand flash mem-
ory and reram that tolerates 32x higher ber for big-
data applications. In Solid-State Circuits Conference
Digest of Technical Papers (ISSCC), 2013 IEEE Inter-
national, pages 226–227.
Wu, M. and Zwaenepoel, W. (1994). envy: a non-volatile,
main memory storage system. In Proceedings of the
6th International Conference on Architectural Sup-
port for Programming Languages and Operating Sys-
tems, ASPLOS VI, pages 86–97, New York, NY,
USA. ACM.
Wu, X. and Reddy, A. L. N. (2011). Scmfs: a file system for
storage class memory. In Proceedings of 2011 Inter-
national Conference for High Performance Comput-
ing, Networking, Storage and Analysis, SC ’11, pages
39:1–39:11, New York, NY, USA. ACM.
Yang, J., Minturn, D. B., and Hady, F. (2012). When
poll is better than interrupt. In Proceedings of the
10th USENIX Conference on File and Storage Tech-
nologies, FAST’12, pages 1–7, Berkeley, CA, USA.
USENIX Association.
TowardsNewInterfaceforNon-volatileMemoryStorage
179