WEBVTT

00:00.000 --> 00:02.730
>> When it comes to providing
redundancy for hard drives,

00:02.730 --> 00:04.920
the first thing you
should think of is RAID,

00:04.920 --> 00:08.025
Redundant Array of
Independent Devices,

00:08.025 --> 00:10.245
or you could hear
independent disks.

00:10.245 --> 00:12.510
That's fine either way. The idea

00:12.510 --> 00:13.995
is there's redundant array.

00:13.995 --> 00:17.760
We have multiple disks acting
as a single logical disk,

00:17.760 --> 00:20.085
and it's usually for the
purpose of redundancy.

00:20.085 --> 00:22.740
That being said, the very
first RAID we look at,

00:22.740 --> 00:24.795
RAID 0 is not redundant.

00:24.795 --> 00:27.780
If I rule the world,
I would call it AID.

00:27.780 --> 00:30.095
It's an array of
independent disks,

00:30.095 --> 00:31.825
but it's not redundant.

00:31.825 --> 00:33.560
Without redundancy, what that

00:33.560 --> 00:35.660
means is if one of
your disks fails,

00:35.660 --> 00:38.385
you run the possibility
of losing all your data.

00:38.385 --> 00:40.580
What RAID 0 does is it takes

00:40.580 --> 00:42.110
two physical disks and

00:42.110 --> 00:44.260
uses a feature called
disk striping.

00:44.260 --> 00:46.340
If there's 24 kilobytes worth

00:46.340 --> 00:48.290
of data to be
written to the disk,

00:48.290 --> 00:50.250
12 kilobytes gets
written to disk 1,

00:50.250 --> 00:51.680
and simultaneously,

00:51.680 --> 00:54.430
the other 12 kilobytes
is written to disk 2.

00:54.430 --> 00:56.310
You get a simultaneous RAID

00:56.310 --> 00:57.854
>> instead of a sequential RAID.

00:57.854 --> 01:00.885
>> It saves time, it's
a speed improvement.

01:00.885 --> 01:02.930
Not only is it
faster for writing,

01:02.930 --> 01:05.435
but it's also faster for
reading because its data is

01:05.435 --> 01:08.800
being located on drive 1 or
is being read on drive 1,

01:08.800 --> 01:10.820
it's being located on drive 2.

01:10.820 --> 01:12.480
You get a performance boost,

01:12.480 --> 01:13.790
but because of the
fact that it's not

01:13.790 --> 01:15.350
redundant is not really a

01:15.350 --> 01:17.329
>> great choice
for most servers.

01:17.329 --> 01:19.115
>> This would be good, however,

01:19.115 --> 01:20.660
on media [inaudible] servers

01:20.660 --> 01:21.904
where we get that performance,

01:21.904 --> 01:24.169
but we don't really have the
need for fault tolerance.

01:24.169 --> 01:27.450
If high availability
is my need, RAID 1.

01:27.450 --> 01:28.995
What RAID 1 is,

01:28.995 --> 01:30.425
is it's disk mirroring.

01:30.425 --> 01:32.300
I have two disks, each one

01:32.300 --> 01:34.059
>> an exact replica
of the other.

01:34.059 --> 01:36.305
>> This provides really
high availability

01:36.305 --> 01:38.735
because if there is a
failure of one of the disks,

01:38.735 --> 01:40.280
it's very easy and very

01:40.280 --> 01:42.410
quick to transfer
over to the other.

01:42.410 --> 01:44.675
You hardly have any
downtime at all.

01:44.675 --> 01:46.190
Drawback there is, is I'll

01:46.190 --> 01:48.535
always lose half
of my disk space.

01:48.535 --> 01:50.310
By that, I mean if I go out and

01:50.310 --> 01:52.670
buy two five-terabyte drives,

01:52.670 --> 01:55.135
I have 10 terabytes of space.

01:55.135 --> 01:57.465
Five terabytes is
for redundancy,

01:57.465 --> 01:59.700
so I don't get to
use that per se.

01:59.700 --> 02:02.480
That's one of the drawbacks
to disk mirroring,

02:02.480 --> 02:04.420
but really good redundancy.

02:04.420 --> 02:08.015
RAID 5, disk striping
with parity.

02:08.015 --> 02:10.790
This takes a minimum of
three physical disks.

02:10.790 --> 02:12.335
I have disk striping,

02:12.335 --> 02:14.210
which gives me the
speed improvement.

02:14.210 --> 02:17.560
But also on each disk contains
parity for another disk.

02:17.560 --> 02:20.310
Ultimately, if disk 1 fails,

02:20.310 --> 02:21.830
parity on disk 2 can be

02:21.830 --> 02:24.190
used to rebuild the
disk that's failed.

02:24.190 --> 02:25.935
That being said, this is

02:25.935 --> 02:28.550
a dicey process and sometimes
rebuilding the disk

02:28.550 --> 02:30.470
can actually cause data loss or

02:30.470 --> 02:33.205
a disk failure. It's not ideal.

02:33.205 --> 02:36.390
RAID 6 is coming around
to replace RAID 5.

02:36.390 --> 02:39.125
Here, we need four
physical disks.

02:39.125 --> 02:41.270
By using four physical disks,

02:41.270 --> 02:43.415
two disks are for
fault tolerance,

02:43.415 --> 02:45.635
so not as many as
two drives can fail.

02:45.635 --> 02:48.070
That's a little bit
better of the redundancy.

02:48.070 --> 02:51.815
RAID 10 or sometimes
referred to as 1 plus 0,

02:51.815 --> 02:53.480
we said disk striping.

02:53.480 --> 02:55.885
The stripes isn't
fault-tolerant.

02:55.885 --> 02:57.720
What if I mirror that stripes

02:57.720 --> 02:59.419
>> at to another set of disks?

02:59.419 --> 03:02.210
>> That's what RAID
10 is. That requires

03:02.210 --> 03:04.195
four physical disks as well.

03:04.195 --> 03:06.000
Of course, half
of my space would

03:06.000 --> 03:07.915
be used for fault tolerance.

03:07.915 --> 03:10.640
RAID is always
discussed on the exam.

03:10.640 --> 03:12.395
Expect to get questions,

03:12.395 --> 03:15.395
expect to get a couple of
performance-based questions.

03:15.395 --> 03:17.425
Thou shalt know thy RAID.

03:17.425 --> 03:19.940
Another hot topic is
going to be backups

03:19.940 --> 03:22.160
because even though we
have RAID in place,

03:22.160 --> 03:23.780
file corruption malware

03:23.780 --> 03:25.764
>> can infect the
entire RAID array.

03:25.764 --> 03:27.770
>> RAID doesn't really
protect our data,

03:27.770 --> 03:28.880
that way we need it to be

03:28.880 --> 03:31.355
protected so we
back up our data.

03:31.355 --> 03:34.100
There are multiple types
of backups we can use.

03:34.100 --> 03:35.735
There is a full and incremental,

03:35.735 --> 03:38.035
a differential, and a copy.

03:38.035 --> 03:40.100
The full is the
easiest to understand

03:40.100 --> 03:42.515
because a full backup
backs everything.

03:42.515 --> 03:44.690
Incremental backup
backs up what's

03:44.690 --> 03:47.195
changed since the last
backup of any kind.

03:47.195 --> 03:49.265
We do a full backup
on Sunday night,

03:49.265 --> 03:51.125
and we do an
incremental on Monday

03:51.125 --> 03:53.285
that backs up what's
changed since Sunday.

03:53.285 --> 03:55.290
We do another
incremental on Tuesday,

03:55.290 --> 03:57.660
and backs up what's
changed since Monday.

03:57.660 --> 03:59.945
An incremental
Wednesday backs up,

03:59.945 --> 04:01.565
which changed since Tuesday.

04:01.565 --> 04:03.740
This takes less time in backup.

04:03.740 --> 04:06.005
However, in the event
that we need to restore,

04:06.005 --> 04:07.670
we have to restore
the full backup

04:07.670 --> 04:09.280
in each day's incremental.

04:09.280 --> 04:11.055
That can take a little longer.

04:11.055 --> 04:13.160
A differential backup backs up

04:13.160 --> 04:15.365
which changed since
the last full backup.

04:15.365 --> 04:17.465
We do a full backup on Sunday.

04:17.465 --> 04:18.920
Monday is incremental and

04:18.920 --> 04:20.905
backs up which changed
since Tuesday.

04:20.905 --> 04:23.760
Tuesday is incremental,
what's changed since Sunday.

04:23.760 --> 04:26.510
Wednesday is incremental,
which changed since Sunday.

04:26.510 --> 04:28.100
This is going to
give us a quicker

04:28.100 --> 04:29.330
restore because we have to

04:29.330 --> 04:30.710
restore the full backup

04:30.710 --> 04:33.079
>> and the most
recent differential.

04:33.079 --> 04:35.030
>> All this has to do in

04:35.030 --> 04:36.680
Windows environments
with a little bit

04:36.680 --> 04:38.200
called the archive bit.

04:38.200 --> 04:41.150
The archive bit is just a
flag that pops up and says,

04:41.150 --> 04:43.895
I've changed. I need
to be backed up.

04:43.895 --> 04:45.350
When you do a full backup,

04:45.350 --> 04:46.745
all the bits get reset.

04:46.745 --> 04:48.350
That's just a way
of acknowledging,

04:48.350 --> 04:50.165
I've backed up everything.

04:50.165 --> 04:53.240
Everything that needs to
be backed up is backed up.

04:53.240 --> 04:55.685
On Monday, as files
start to change,

04:55.685 --> 04:57.095
those flags pop up.

04:57.095 --> 04:58.955
We do an incremental
Monday night,

04:58.955 --> 05:00.770
now backs up everything
for the day and it

05:00.770 --> 05:03.475
clears the bits and the
signals it's taken care of.

05:03.475 --> 05:05.880
Tuesday's new files
are modified.

05:05.880 --> 05:07.755
The file pops back up.

05:07.755 --> 05:09.590
Tuesday's incremental
backs up which

05:09.590 --> 05:11.775
changed on Tuesday
and clears the Bits.

05:11.775 --> 05:13.100
The differential backup

05:13.100 --> 05:14.350
>> is different
from the other two,

05:14.350 --> 05:15.640
>> because the differential Bit

05:15.640 --> 05:17.290
backup does not clear the Bit,

05:17.290 --> 05:20.265
which is exactly why
Monday's files changed.

05:20.265 --> 05:22.310
It backs up everything
that's changed since

05:22.310 --> 05:24.440
Sunday, the Bit isn't cleared.

05:24.440 --> 05:26.090
As files pop up Tuesday,

05:26.090 --> 05:27.500
you have the things
for Monday with

05:27.500 --> 05:30.005
flag set and you have
the things from Tuesday.

05:30.005 --> 05:33.490
We back up everything since
the last full backup.

05:33.490 --> 05:36.650
There's also a copy.
With virtual machines,

05:36.650 --> 05:38.255
you can copy specific files

05:38.255 --> 05:39.925
or you could do a full backup.

05:39.925 --> 05:42.765
This just does not
reset the archive bit.

05:42.765 --> 05:44.930
Also with virtualization, we now

05:44.930 --> 05:47.240
think about just
reverting to Snapshots.

05:47.240 --> 05:48.920
Whatever our strategy is,

05:48.920 --> 05:50.150
we need to make
sure that we have

05:50.150 --> 05:51.500
fault tolerance for our data

05:51.500 --> 05:53.850
>> as well as our hard drives.