<s>
High	B-Application
availability	I-Application
software	I-Application
is	O
software	O
used	O
to	O
ensure	O
that	O
systems	O
are	O
running	O
and	O
available	O
most	O
of	O
the	O
time	O
.	O
</s>
<s>
This	O
characteristic	O
is	O
weaker	O
than	O
fault	B-General_Concept
tolerance	I-General_Concept
,	O
which	O
typically	O
seeks	O
to	O
provide	O
100%	O
availability	O
,	O
albeit	O
with	O
significant	O
price	O
and	O
performance	O
penalties	O
.	O
</s>
<s>
High	B-Application
availability	I-Application
software	I-Application
is	O
measured	O
by	O
its	O
performance	O
when	O
a	O
subsystem	O
fails	O
,	O
its	O
ability	O
to	O
resume	O
service	O
in	O
a	O
state	O
close	O
to	O
the	O
state	O
of	O
the	O
system	O
at	O
the	O
time	O
of	O
the	O
original	O
failure	O
,	O
and	O
its	O
ability	O
to	O
perform	O
other	O
service-affecting	O
tasks	O
(	O
such	O
as	O
software	O
upgrade	O
or	O
configuration	O
changes	O
)	O
in	O
a	O
manner	O
that	O
eliminates	O
or	O
minimizes	O
down	O
time	O
.	O
</s>
<s>
All	O
faults	O
that	O
affect	O
availability	O
–	O
hardware	O
,	O
software	O
,	O
and	O
configuration	O
need	O
to	O
be	O
addressed	O
by	O
High	B-Application
Availability	I-Application
Software	I-Application
to	O
maximize	O
availability	O
.	O
</s>
<s>
Typical	O
high	B-Application
availability	I-Application
software	I-Application
provides	O
features	O
that	O
:	O
</s>
<s>
The	O
“	O
scale-out	O
”	O
property	O
of	O
a	O
system	O
refers	O
to	O
the	O
ability	O
to	O
create	O
multiple	O
copies	O
of	O
a	O
subsystem	O
to	O
address	O
increasing	O
demand	O
,	O
and	O
to	O
efficiently	O
distribute	O
incoming	O
work	O
to	O
these	O
copies	O
(	O
Load	B-Application
balancing	I-Application
(	O
computing	O
)	O
)	O
preferably	O
without	O
shutting	O
down	O
the	O
system	O
.	O
</s>
<s>
High	B-Application
availability	I-Application
software	I-Application
should	O
enable	O
scale-out	O
without	O
interrupting	O
service	O
.	O
</s>
<s>
High	B-Application
Availability	I-Application
Software	I-Application
can	O
provide	O
communications	O
abstractions	O
like	O
redundant	O
message	O
and	O
event	O
queues	O
to	O
help	O
active	O
subsystems	O
in	O
this	O
task	O
.	O
</s>
<s>
This	O
idea	O
is	O
commonly	O
abstracted	O
as	O
a	O
distributed	B-Operating_System
hash	I-Operating_System
table	I-Operating_System
–	O
the	O
active	O
writes	O
key/value	O
records	O
into	O
the	O
table	O
and	O
both	O
the	O
active	O
and	O
standby	O
subsystems	O
read	O
from	O
it	O
.	O
</s>
<s>
Unlike	O
a	O
“	O
cloud	B-Architecture
”	O
distributed	B-Operating_System
hash	I-Operating_System
table	I-Operating_System
(	O
Chord	B-Application
(	O
peer-to-peer	O
)	O
,	O
Kademlia	B-Operating_System
,	O
etc	O
.	O
)	O
</s>
<s>
For	O
example	O
,	O
many	O
cloud	B-Architecture
solutions	O
will	O
restart	O
a	O
virtual	O
machine	O
on	O
another	O
physical	O
machine	O
if	O
the	O
underlying	O
physical	O
machine	O
fails	O
.	O
</s>
<s>
High	B-Application
availability	I-Application
software	I-Application
can	O
help	O
engineers	O
create	O
complex	O
system	O
architectures	O
that	O
are	O
designed	O
to	O
minimize	O
the	O
scope	O
of	O
failures	O
and	O
to	O
handle	O
specific	O
failure	O
modes	O
.	O
</s>
<s>
Sophisticated	O
policies	O
can	O
be	O
specified	O
by	O
high	B-Application
availability	I-Application
software	I-Application
to	O
differentiate	O
software	O
from	O
hardware	O
faults	O
,	O
and	O
to	O
attempt	O
time-delayed	O
restarts	O
of	O
individual	O
software	O
processes	O
,	O
entire	O
software	O
stacks	O
,	O
or	O
entire	O
systems	O
.	O
</s>
<s>
Telecommunications	O
:	O
High	B-Application
Availability	I-Application
Software	I-Application
is	O
an	O
essential	O
component	O
of	O
telecommunications	O
equipment	O
since	O
a	O
network	O
outage	O
can	O
result	O
in	O
significant	O
loss	O
in	O
revenue	O
for	O
telecom	O
providers	O
and	O
telephone	O
access	O
to	O
emergency	O
services	O
is	O
an	O
important	O
public	O
safety	O
issue	O
.	O
</s>
<s>
Space	O
:	O
High	B-Application
Availability	I-Application
Software	I-Application
is	O
proposed	O
for	O
use	O
of	O
non-radiation	O
hardened	O
equipment	O
in	O
space	O
environments	O
.	O
</s>
<s>
But	O
High	B-Application
Availability	I-Application
Software	I-Application
running	O
on	O
a	O
single	O
or	O
pair	O
of	O
rad-hardened	O
controllers	O
can	O
manage	O
many	O
redundant	O
high	O
performance	O
non-rad-hard	O
computers	O
,	O
potentially	O
failing	O
over	O
and	O
resetting	O
them	O
in	O
the	O
event	O
of	O
a	O
fault	O
.	O
</s>
<s>
Typical	O
cloud	B-Architecture
services	I-Architecture
provide	O
a	O
set	O
of	O
networked	O
computers	O
(	O
typical	O
a	O
virtual	O
machine	O
)	O
running	O
a	O
standard	O
server	O
OS	O
like	O
Linux	O
.	O
</s>
<s>
The	O
cloud	B-Architecture
infrastructure	O
may	O
provide	O
simple	O
fault	O
detection	O
and	O
restart	O
at	O
the	O
virtual	O
machine	O
level	O
.	O
</s>
<s>
Additionally	O
,	O
cloud	B-Architecture
services	I-Architecture
cannot	O
detect	O
software	O
failures	O
within	O
the	O
virtual	O
machines	O
.	O
</s>
<s>
High	B-Application
Availability	I-Application
Software	I-Application
running	O
inside	O
the	O
cloud	B-Architecture
virtual	O
machines	O
can	O
detect	O
software	O
(	O
and	O
virtual	O
machine	O
)	O
failures	O
in	O
seconds	O
and	O
can	O
use	O
checkpointing	O
to	O
ensure	O
that	O
standby	O
virtual	O
machines	O
are	O
ready	O
to	O
take	O
over	O
service	O
.	O
</s>
