<s>
GPFS	B-Application
(	O
General	B-Application
Parallel	I-Application
File	I-Application
System	I-Application
,	O
brand	O
name	O
IBM	O
Storage	O
Scale	O
and	O
previously	O
IBM	B-Application
Spectrum	I-Application
Scale	I-Application
)	O
is	O
high-performance	O
clustered	B-Application
file	I-Application
system	I-Application
software	O
developed	O
by	O
IBM	O
.	O
</s>
<s>
It	O
can	O
be	O
deployed	O
in	O
shared-disk	B-General_Concept
or	O
shared-nothing	B-Operating_System
distributed	O
parallel	O
modes	O
,	O
or	O
a	O
combination	O
of	O
these	O
.	O
</s>
<s>
It	O
is	O
used	O
by	O
many	O
of	O
the	O
world	O
's	O
largest	O
commercial	O
companies	O
,	O
as	O
well	O
as	O
some	O
of	O
the	O
supercomputers	B-Architecture
on	O
the	O
Top	B-Operating_System
500	I-Operating_System
List	I-Operating_System
.	O
</s>
<s>
at	O
Oak	O
Ridge	O
National	O
Laboratory	O
which	O
was	O
the	O
#1	O
fastest	B-Operating_System
supercomputer	I-Operating_System
in	O
the	O
world	O
in	O
the	O
November	O
2019	O
TOP500	B-Operating_System
list	B-Operating_System
of	I-Operating_System
supercomputers	I-Operating_System
.	O
</s>
<s>
Summit	B-Device
is	O
a	O
200	O
Petaflops	O
system	O
composed	O
of	O
more	O
than	O
9,000	O
POWER9	B-Device
processors	O
and	O
27,000	O
NVIDIA	B-General_Concept
Volta	I-General_Concept
GPUs	B-Architecture
.	O
</s>
<s>
The	O
storage	O
filesystem	O
called	O
Alpine	O
has	O
250	O
PB	O
of	O
storage	O
using	O
Spectrum	B-Application
Scale	I-Application
on	O
IBM	O
ESS	O
storage	O
hardware	O
,	O
capable	O
of	O
approximately	O
2.5TB/s	O
of	O
sequential	O
I/O	O
and	O
2.2TB/s	O
of	O
random	O
I/O	O
.	O
</s>
<s>
Like	O
typical	O
cluster	O
filesystems	O
,	O
GPFS	B-Application
provides	O
concurrent	B-Operating_System
high-speed	O
file	O
access	O
to	O
applications	O
executing	O
on	O
multiple	O
nodes	O
of	O
clusters	O
.	O
</s>
<s>
It	O
can	O
be	O
used	O
with	O
AIX	B-Application
clusters	O
,	O
Linux	B-Application
clusters	O
,	O
on	O
Microsoft	O
Windows	B-Device
Server	I-Device
,	O
or	O
a	O
heterogeneous	O
cluster	O
of	O
AIX	B-Application
,	O
Linux	B-Application
and	O
Windows	O
nodes	O
running	O
on	O
x86	B-Operating_System
,	O
Power	B-Architecture
or	O
IBM	B-Device
Z	I-Device
processor	O
architectures	O
.	O
</s>
<s>
In	O
addition	O
to	O
providing	O
filesystem	O
storage	O
capabilities	O
,	O
it	O
provides	O
tools	O
for	O
management	O
and	O
administration	O
of	O
the	O
GPFS	B-Application
cluster	O
and	O
allows	O
for	O
shared	B-General_Concept
access	I-General_Concept
to	O
file	O
systems	O
from	O
remote	O
clusters	O
.	O
</s>
<s>
GPFS	B-Application
began	O
as	O
the	O
Tiger	O
Shark	O
file	O
system	O
,	O
a	O
research	O
project	O
at	O
IBM	O
's	O
Almaden	O
Research	O
Center	O
as	O
early	O
as	O
1993	O
.	O
</s>
<s>
Vesta	O
introduced	O
the	O
concept	O
of	O
file	O
partitioning	O
to	O
accommodate	O
the	O
needs	O
of	O
parallel	O
applications	O
that	O
run	O
on	O
high-performance	O
multicomputers	B-Operating_System
with	O
parallel	B-Operating_System
I/O	I-Operating_System
subsystems	O
.	O
</s>
<s>
Vesta	O
was	O
commercialized	O
as	O
the	O
PIOFS	O
filesystem	O
around	O
1994	O
,	O
and	O
was	O
succeeded	O
by	O
GPFS	B-Application
around	O
1998	O
.	O
</s>
<s>
The	O
main	O
difference	O
between	O
the	O
older	O
and	O
newer	O
filesystems	O
was	O
that	O
GPFS	B-Application
replaced	O
the	O
specialized	O
interface	O
offered	O
by	O
Vesta/PIOFS	O
with	O
the	O
standard	O
Unix	B-Application
API	B-General_Concept
:	O
all	O
the	O
features	O
to	O
support	O
high	O
performance	O
parallel	B-Operating_System
I/O	I-Operating_System
were	O
hidden	O
from	O
users	O
and	O
implemented	O
under	O
the	O
hood	O
.	O
</s>
<s>
GPFS	B-Application
has	O
been	O
available	O
on	O
IBM	O
's	O
AIX	B-Application
since	O
1998	O
,	O
on	O
Linux	B-Application
since	O
2001	O
,	O
and	O
on	O
Windows	B-Device
Server	I-Device
since	O
2008	O
.	O
</s>
<s>
Today	O
it	O
is	O
used	O
by	O
many	O
of	O
the	O
top	B-Operating_System
500	I-Operating_System
supercomputers	I-Operating_System
listed	O
on	O
the	O
Top	B-Operating_System
500	I-Operating_System
Supercomputing	B-Architecture
List	O
.	O
</s>
<s>
In	O
2010	O
,	O
IBM	O
previewed	O
a	O
version	O
of	O
GPFS	B-Application
that	O
included	O
a	O
capability	O
known	O
as	O
GPFS-SNC	O
,	O
where	O
SNC	O
stands	O
for	O
Shared	B-Operating_System
Nothing	I-Operating_System
Cluster	I-Operating_System
.	O
</s>
<s>
This	O
allows	O
it	O
to	O
use	O
locally	O
attached	O
disks	O
on	O
a	O
cluster	O
of	O
network	O
connected	O
servers	O
rather	O
than	O
requiring	O
dedicated	O
servers	O
with	O
shared	B-General_Concept
disks	I-General_Concept
(	O
e.g.	O
</s>
<s>
FPO	O
is	O
suitable	O
for	O
workloads	O
with	O
high	O
data	O
locality	O
such	O
as	O
shared	B-Operating_System
nothing	I-Operating_System
database	O
clusters	O
such	O
as	O
SAP	O
HANA	O
and	O
DB2	O
DPF	O
,	O
and	O
can	O
be	O
used	O
as	O
a	O
HDFS-compatible	O
filesystem	O
.	O
</s>
<s>
It	O
is	O
a	O
clustered	B-Application
file	I-Application
system	I-Application
.	O
</s>
<s>
The	O
system	O
stores	O
data	O
on	O
standard	O
block	O
storage	O
volumes	O
,	O
but	O
includes	O
an	O
internal	O
RAID	B-Architecture
layer	O
that	O
can	O
virtualize	O
those	O
volumes	O
for	O
redundancy	O
and	O
parallel	O
access	O
much	O
like	O
a	O
RAID	B-Architecture
block	O
storage	O
system	O
.	O
</s>
<s>
This	O
ensures	O
the	O
filesystem	O
is	O
available	O
more	O
often	O
,	O
so	O
keeps	O
the	O
supercomputer	B-Architecture
cluster	O
itself	O
available	O
for	O
longer	O
.	O
</s>
<s>
Other	O
features	O
include	O
high	O
availability	O
,	O
ability	O
to	O
be	O
used	O
in	O
a	O
heterogeneous	O
cluster	O
,	O
disaster	O
recovery	O
,	O
security	O
,	O
DMAPI	B-General_Concept
,	O
HSM	B-Application
and	O
ILM	B-Protocol
.	O
</s>
<s>
Hadoop	B-Application
's	O
HDFS	O
filesystem	O
,	O
is	O
designed	O
to	O
store	O
similar	O
or	O
greater	O
quantities	O
of	O
data	O
on	O
commodity	O
hardware	O
—	O
that	O
is	O
,	O
datacenters	O
without	O
RAID	B-Architecture
disks	O
and	O
a	O
storage	O
area	O
network	O
(	O
SAN	O
)	O
.	O
</s>
<s>
GPFS	B-Application
has	O
full	O
Posix	O
filesystem	O
semantics	O
.	O
</s>
<s>
GPFS	B-Application
distributes	O
its	O
directory	O
indices	O
and	O
other	O
metadata	O
across	O
the	O
filesystem	O
.	O
</s>
<s>
Hadoop	B-Application
,	O
in	O
contrast	O
,	O
keeps	O
this	O
on	O
the	O
Primary	O
and	O
Secondary	O
Namenodes	O
,	O
large	O
servers	O
which	O
must	O
store	O
all	O
index	O
information	O
in-RAM	O
.	O
</s>
<s>
GPFS	B-Application
breaks	O
files	O
up	O
into	O
small	O
blocks	O
.	O
</s>
<s>
Hadoop	B-Application
HDFS	O
likes	O
blocks	O
of	O
or	O
more	O
,	O
as	O
this	O
reduces	O
the	O
storage	O
requirements	O
of	O
the	O
Namenode	O
.	O
</s>
<s>
For	O
example	O
,	O
one	O
pool	O
could	O
be	O
high-performance	O
Fibre	B-Architecture
Channel	I-Architecture
disks	O
and	O
another	O
more	O
economical	O
SATA	O
storage	O
.	O
</s>
