<s>
Actian	O
Vector	O
(	O
formerly	O
known	O
as	O
VectorWise	B-Application
)	O
is	O
an	O
SQL	B-Language
relational	B-Application
database	I-Application
management	I-Application
system	I-Application
designed	O
for	O
high	O
performance	O
in	O
analytical	O
database	O
applications	O
.	O
</s>
<s>
It	O
published	O
record	O
breaking	O
results	O
on	O
the	O
Transaction	B-General_Concept
Processing	I-General_Concept
Performance	I-General_Concept
Council	I-General_Concept
's	O
TPC-H	O
benchmark	O
for	O
database	O
sizes	O
of	O
100	O
GB	O
,	O
300	O
GB	O
,	O
1	O
TB	O
and	O
3	O
TB	O
on	O
non-clustered	O
hardware	O
.	O
</s>
<s>
Vectorwise	B-Application
originated	O
from	O
the	O
X100	O
research	O
project	O
carried	O
out	O
within	O
the	O
Centrum	O
Wiskunde	O
&	O
Informatica	O
(	O
CWI	O
,	O
the	O
Dutch	O
National	O
Research	O
Institute	O
for	O
Mathematics	O
and	O
Computer	O
Science	O
)	O
between	O
2003	O
and	O
2008	O
.	O
</s>
<s>
It	O
was	O
spun	O
off	O
as	O
a	O
start-up	O
company	O
in	O
2008	O
,	O
and	O
acquired	O
by	O
Ingres	B-Application
Corporation	O
in	O
2011	O
.	O
</s>
<s>
It	O
was	O
released	O
as	O
a	O
commercial	O
product	O
in	O
June	O
,	O
2010	O
,	O
initially	O
for	O
64-bit	O
Linux	B-Application
platform	O
,	O
and	O
later	O
also	O
for	O
Windows	O
.	O
</s>
<s>
In	O
June	O
2014	O
,	O
Actian	O
Vortex	O
was	O
announced	O
as	O
a	O
clustered	O
massive	B-Operating_System
parallel	I-Operating_System
processing	I-Operating_System
version	O
of	O
Vector	O
,	O
in	O
Hadoop	B-Application
with	O
storage	O
in	O
HDFS	O
.	O
</s>
<s>
Actian	O
Vortex	O
was	O
later	O
renamed	O
to	O
Actian	O
Vector	O
in	O
Hadoop	B-Application
.	O
</s>
<s>
The	O
basic	O
architecture	O
and	O
design	O
principles	O
of	O
the	O
X100	O
engine	O
of	O
the	O
VectorWise	B-Application
database	O
were	O
well	O
described	O
in	O
two	O
Phd	O
theses	O
of	O
VectorWise	B-Application
founders	O
Marcin	O
Żukowski	O
:	O
"	O
Balancing	O
Vectorized	O
Query	O
Execution	O
with	O
Bandwidth-Optimized	O
Storage	O
"	O
and	O
Sandor	O
Héman	O
:	O
"	O
Updating	O
Compressed	O
Column	B-General_Concept
Stores	I-General_Concept
"	O
,	O
under	O
supervision	O
of	O
another	O
founder	O
,	O
professor	O
Peter	O
Boncz	O
.	O
</s>
<s>
The	O
X100	O
engine	O
was	O
integrated	O
with	O
Ingres	B-Application
SQL	B-Language
front-end	O
,	O
allowing	O
the	O
database	O
to	O
use	O
the	O
Ingres	B-Application
SQL	B-Language
syntax	O
,	O
and	O
Ingres	B-Application
set	O
of	O
client	O
and	O
database	B-General_Concept
administration	I-General_Concept
tools	O
.	O
</s>
<s>
The	O
query	O
execution	O
architecture	O
makes	O
use	O
of	O
"	O
Vectorized	O
Query	O
Execution	O
"	O
processing	O
in	O
chunks	O
of	O
cache-fitting	O
vectors	O
of	O
data	O
.	O
</s>
<s>
This	O
allows	O
to	O
involve	O
the	O
principles	O
of	O
vector	B-Operating_System
processing	I-Operating_System
and	O
single	B-Device
instruction	I-Device
,	I-Device
multiple	I-Device
data	I-Device
(	O
SIMD	B-Device
)	O
to	O
perform	O
the	O
same	O
operation	O
on	O
multiple	O
data	O
simultaneously	O
and	O
exploit	O
data	O
level	O
parallelism	O
on	O
modern	O
hardware	O
.	O
</s>
<s>
The	O
database	O
storage	O
is	O
in	O
a	O
compressed	O
column-oriented	B-General_Concept
format	O
,	O
with	O
scan-optimised	O
buffer	O
manager	O
.	O
</s>
<s>
Loading	O
big	O
amounts	O
of	O
data	O
is	O
supported	O
through	O
direct	O
appends	O
to	O
stable	O
storage	O
,	O
while	O
small	O
transactional	O
updates	O
are	O
supported	O
through	O
patent-pending	O
Positional	O
Delta	O
Trees	O
(	O
PDTs	O
)	O
specialized	O
B-tree-like	O
structures	O
of	O
indexed	O
differences	O
on	O
top	O
of	O
stable	O
storage	O
,	O
which	O
are	O
seamlessly	O
patched	O
during	O
scans	O
,	O
and	O
which	O
are	O
transparently	O
propagated	O
to	O
stable	O
storage	O
in	O
a	O
background	O
process	O
.	O
</s>
<s>
A	O
comparative	O
Transaction	B-General_Concept
Processing	I-General_Concept
Performance	I-General_Concept
Council	I-General_Concept
TPC-H	O
performance	O
test	O
of	O
MonetDB	B-Application
carried	O
out	O
by	O
its	O
original	O
creator	O
at	O
Centrum	O
Wiskunde	O
&	O
Informatica	O
(	O
CWI	O
)	O
in	O
2003	O
showed	O
room	O
for	O
improvement	O
in	O
its	O
performance	O
as	O
an	O
analytical	O
database	O
.	O
</s>
<s>
As	O
a	O
result	O
,	O
CWI	O
researchers	O
proposed	O
a	O
new	O
architecture	O
using	O
pipelined	O
query	B-Application
processing	I-Application
(	O
"	O
vectorised	O
processing	O
"	O
)	O
to	O
improve	O
the	O
performance	O
of	O
analytical	O
queries	O
.	O
</s>
<s>
This	O
led	O
to	O
the	O
creation	O
of	O
the	O
"	O
X100	O
"	O
project	O
,	O
with	O
the	O
intention	O
of	O
designing	O
a	O
new	O
kernel	O
for	O
MonetDB	B-Application
,	O
to	O
be	O
called	O
"	O
MonetDB/X100	O
"	O
.	O
</s>
<s>
The	O
X100	O
project	O
team	O
won	O
the	O
2007	O
DaMoN	O
Best	O
Paper	O
Award	O
for	O
the	O
paper	O
"	O
Vectorized	O
Data	O
Processing	O
on	O
the	O
Cell	O
Broadband	O
Engine	O
"	O
as	O
well	O
as	O
the	O
2008	O
DaMoN	O
Best	O
Paper	O
Award	O
for	O
the	O
paper	O
"	O
DSM	O
vs.	O
NSM	O
:	O
CPU	B-General_Concept
Performance	O
Tradeoffs	O
in	O
Block-Oriented	O
Query	B-Application
Processing	I-Application
"	O
.	O
</s>
<s>
In	O
August	O
2009	O
the	O
originators	O
for	O
the	O
X100	O
project	O
won	O
the	O
"	O
Ten	O
Year	O
Best	O
Paper	O
Award	O
"	O
at	O
the	O
35th	O
International	O
Conference	O
on	O
Very	O
Large	O
Data	B-Application
Bases	I-Application
(	O
VLDB	O
)	O
for	O
their	O
1999	O
paper	O
"	O
Database	O
architecture	O
Optimized	O
for	O
the	O
new	O
bottleneck	O
:	O
Memory	O
access	O
"	O
.	O
</s>
<s>
The	O
central	O
premise	O
of	O
the	O
paper	O
is	O
that	O
traditional	O
relational	B-Application
database	I-Application
systems	O
were	O
designed	O
in	O
the	O
late	O
1970s	O
and	O
early	O
1980s	O
during	O
a	O
time	O
when	O
database	O
performance	O
was	O
dictated	O
by	O
the	O
time	O
required	O
to	O
read	O
from	O
and	O
write	O
data	O
to	O
hard	O
disk	O
.	O
</s>
<s>
At	O
that	O
time	O
available	O
CPU	B-General_Concept
was	O
relatively	O
slow	O
and	O
main	O
memory	O
was	O
relatively	O
small	O
,	O
so	O
that	O
very	O
little	O
data	O
could	O
be	O
loaded	O
into	O
memory	O
at	O
a	O
time	O
.	O
</s>
<s>
Over	O
time	O
hardware	O
improved	O
,	O
with	O
CPU	B-General_Concept
speed	O
and	O
memory	O
size	O
doubling	O
roughly	O
every	O
two	O
years	O
in	O
accordance	O
with	O
Moore	O
’s	O
law	O
,	O
but	O
that	O
the	O
design	O
of	O
traditional	O
relational	B-Application
database	I-Application
systems	O
had	O
not	O
adapted	O
.	O
</s>
<s>
In	O
2008	O
the	O
X100	O
project	O
was	O
spun	O
off	O
from	O
MonetDB	B-Application
as	O
a	O
separate	O
project	O
,	O
with	O
its	O
own	O
company	O
,	O
and	O
renamed	O
"	O
VectorWise	B-Application
"	O
.	O
</s>
<s>
In	O
June	O
2010	O
,	O
the	O
VectorWise	B-Application
technology	O
was	O
officially	O
announced	O
by	O
Ingres	B-Application
Corporation	O
,	O
with	O
the	O
release	O
of	O
Ingres	B-Application
VectorWise	B-Application
1.0	O
.	O
</s>
<s>
In	O
March	O
2011	O
,	O
VectorWise	B-Application
1.5	O
was	O
released	O
,	O
publishing	O
a	O
record	O
breaking	O
result	O
on	O
TPC-H	O
100	O
GB	O
benchmark	O
.	O
</s>
<s>
New	O
features	O
included	O
parallel	O
query	O
execution	O
(	O
single	O
query	O
executed	O
on	O
multiple	O
CPU	B-General_Concept
cores	O
)	O
,	O
improved	O
bulk	O
loading	O
and	O
enhanced	O
SQL	B-Language
support	O
.	O
</s>
<s>
In	O
June	O
2011	O
,	O
VectorWise	B-Application
1.6	O
was	O
released	O
,	O
publishing	O
record	O
breaking	O
results	O
on	O
TPC-H	O
100	O
GB	O
,	O
300	O
GB	O
and	O
1	O
TB	O
non-clustered	O
benchmark	O
.	O
</s>
<s>
In	O
December	O
2011	O
,	O
VectorWise	B-Application
2.0	O
was	O
released	O
with	O
new	O
SQL	B-Language
support	O
for	O
analytical	O
functions	O
such	O
as	O
rank	O
and	O
percentile	O
and	O
enhanced	O
date	O
,	O
time	O
and	O
timestamp	O
datatypes	O
,	O
and	O
support	O
for	O
disk	O
spilling	O
in	O
hash	O
joins	O
and	O
aggregation	O
.	O
</s>
<s>
In	O
June	O
2012	O
,	O
VectorWise	B-Application
2.5	O
was	O
released	O
.	O
</s>
<s>
In	O
March	O
2013	O
,	O
VectorWise	B-Application
3.0	O
was	O
released	O
.	O
</s>
<s>
New	O
features	O
included	O
more	O
efficient	O
storage	O
engine	O
,	O
support	O
for	O
more	O
data	O
types	O
and	O
analytical	O
SQL	B-Language
functions	I-Language
,	O
enhanced	O
DDL	O
features	O
,	O
and	O
improved	O
monitoring	O
and	O
profiling	O
accessibility	O
.	O
</s>
<s>
New	O
features	O
included	O
support	O
for	O
partitioned	O
tables	O
,	O
improved	O
disk	O
spilling	O
,	O
online	O
backup	O
capabilities	O
and	O
improved	O
SQL	B-Language
support	O
-	O
e.g.	O
</s>
<s>
In	O
June	O
2014	O
,	O
at	O
Hadoop	B-Application
Summit	O
2014	O
in	O
San	O
Jose	O
,	O
Actian	O
announced	O
Actian	O
Vortex	O
clustered	O
MPP	O
version	O
of	O
Vector	O
,	O
with	O
same	O
level	O
of	O
SQL	B-Language
support	O
working	O
in	O
Hadoop	B-Application
with	O
storage	O
directly	O
in	O
HDFS	O
.	O
</s>
<s>
Actian	O
Vortex	O
was	O
later	O
renamed	O
to	O
Actian	O
Vector	O
in	O
Hadoop	B-Application
,	O
and	O
non-clustered	O
Actian	O
Vector	O
releases	O
are	O
also	O
updated	O
to	O
match	O
.	O
</s>
<s>
In	O
March	O
2015	O
Actian	O
Vector	O
4	O
was	O
released	O
,	O
and	O
Actian	O
Vector	O
in	O
Hadoop	B-Application
4	O
was	O
released	O
in	O
December	O
2015	O
.	O
</s>
