Memory & Storage: Crash Course Computer Science #19

0:03 Hi, I'm Carrie Anne, and welcome to Crash Course Computer Science!

0:05 We’ve talked about computer memory several times in this series,

0:08 and we even designed

0:09 some in Episode 6.

0:11 In general, computer memory is non-permanent.

0:13 If your xbox accidently gets unplugged and turns off,

0:16 any data saved in memory is lost.

0:17 For this reason, it’s called volatile memory.

0:20 What we haven’t talked so much about this series is storage,

0:23 which is a tad different.

0:24 Any data written to storage, like your hard drive,

0:26 will stay there until it’s over-written

0:28 or deleted, even if the power goes out.

0:30 It’s non-volatile.

0:31 It used to be that volatile memory was fast and non-volatile storage was slow,

0:35 but as computing technologies have improved,

0:37 this distinction is becoming less true, and the

0:40 terms have started to blend together.

0:41 Nowadays, we take for granted technologies like this little USB stick,

0:45 which offers gigabytes of memory, reliable over long periods of time,

0:49 all at low cost, but this wasn’t always true.

0:52 INTRO The earliest computer storage was paper punch cards,

1:03 and its close cousin, punched paper tape.

1:05 By the 1940s,

1:06 punch cards had largely standardized into a grid of 80 columns and 12 rows,

1:10 allowing for a maximum of 960 bits of data to be stored on a single card.

1:15 The largest program ever punched onto cards,

1:17 that we know of, was the US Military’s Semi-Automatic

1:19 Ground Environment, or SAGE,

1:21 an Air Defense System that became operational in 1958.

1:24 The main program was stored on 62,500 punchcards,

1:28 roughly equivalent to 5 megabytes of data,

1:30 that’s the size of an average smartphone photo today.

1:34 Punch cards were a useful and popular form of storage for decades,

1:37 they didn’t need power, plus paper was cheap and reasonably durable.

1:40 However, punchcards were slow and write-once, you can’t easily un-punch a hole.

1:44 So they were a less useful form of memory,

1:46 where a value might only be needed for a fraction

1:48 of a second during a program's execution, and then discarded.

1:51 A faster, larger and more flexible form of computer memory was needed.

1:55 An early and practical approach was developed by J.

1:57 Presper Eckert, as he was finishing

1:59 work on ENIAC in 1944.

2:01 His invention was called Delay Line Memory, and it worked like this.

2:05 You take a tube and fill it with a liquid, like mercury.

2:07 Then, you put a speaker at one end and microphone at the other.

2:10 When you pulse the speaker, it creates a pressure wave.

2:13 This takes time to propagate to the other end of the tube,

2:16 where it hits the microphone,

2:17 converting it back into an electrical signal.

2:19 And we can use this propagation delay to store data!

2:22 Imagine that the presence of a pressure wave is

2:24 a 1 and the absence of a pressure wave

2:26 is a 0.

2:27 Our speaker can output a binary sequence like 1010 0111.

2:30 The corresponding waves will travel down the tube,

2:33 in order, and a little while later,

2:35 hit the microphone, which converts the signal back into 1’s and 0’s.

2:38 If we create a circuit that connects the microphone to the speaker,

2:41 plus a little amplifier to

2:43 compensate for any loss, we can create a loop that stores data.

2:46 The signal traveling along the wire is near instantaneous,

2:49 so there’s only ever one

2:50 bit of data showing at any moment in time.

2:53 But in the tube, you can store many bits!

2:55 After working on ENIAC, Eckert and his colleague John Mauchly,

2:58 set out to build a bigger and

2:59 better computer called EDVAC, incorporating Delay Line Memory.

3:03 In total, the computer had 128 Delay Lines, each capable of storing 352 bits.

3:09 That’s a grand total of 45 thousands bits of memory, not too shabby for 1949!

3:15 This allowed EDVAC to be one of the very earliest Stored-Program Computers,

3:18 which we talked about in Episode 10.

3:20 However, a big drawback with delay line memory is that

3:22 you could only read one bit of data

3:24 from a tube at any given instant.

3:26 If you wanted to access a specific bit,

3:27 like bit 112, you’d have to wait for it to come

3:29 around in the loop,

3:31 what’s called sequential or cyclic-access memory, whereas we really

3:34 want random access memory, where we can access any bit at any time.

3:38 It also proved challenging to increase the density of the memory,

3:41 packing waves closer together meant they were more easily mixed up.

3:44 In response, new forms of delay line memory were invented,

3:47 such as magnetostrictive delay

3:48 lines.

3:49 These delay lines use a metal wire that could be twisted,

3:51 creating little torsional waves

3:52 that represented data.

3:54 By forming the wire into a coil,

3:56 you could store around 1000 bits in a 1 foot by 1 foot square.

3:59 However, delay line memory was largely obsolete by the mid 1950s,

4:03 surpassed in performance, reliability and cost by a new kid on the block:

4:07 magnetic core memory which was constructed

4:09 out of little magnetic donuts, called cores.

4:11 If you loop a wire around this core….

4:13 and run an electrical current through the wire,

4:15 we can magnetize the core in a certain direction.

4:17 If we turn the current off, the core will stay magnetized.

4:20 If we pass current through the wire in the opposite direction,

4:24 the magnetization direction, called polarity, flips the other way.

4:27 In this way, we can store 1’s and 0’s!

4:29 1 bit of memory isn’t very useful,

4:31 so these little donuts were arranged into grids.

4:34 There were wires for selecting the right row and column,

4:36 and a wire that ran through every

4:37 core, which could be used to read or write a bit.

4:40 Here is an actual piece of core memory!

4:42 In each of these little yellow squares,

4:44 there are 32 rows and 32 columns of tiny cores,

4:47 each one holding 1 bit of data.

4:49 So, each of these yellow squares could hold 1024 bits.

4:53 In total, there are 9 of these,

4:55 so this memory board could hold a maximum of 9216 bits, which

4:59 is around 9 kilobytes.

5:00 The first big use of core memory was MIT’s Whirlwind 1 computer,

5:05 in 1953, which used

5:06 a 32 by 32 core arrangement.

5:08 And, instead of just a single plane of cores,

5:10 like this, it was 16 boards deep, providing

5:13 roughly 16 thousand bits of storage.

5:15 Importantly, unlike delay line memory, any bit could be accessed at any time.

5:19 This was a killer feature,

5:20 and magnetic core memory became the predominant Random Access

5:23 Memory technology for two decades,

5:24 beginning in the mid 1950s even though it was typically

5:28 woven by hand!

5:29 Although starting at roughly 1 dollar per bit,

5:31 the cost fell to around 1 cent per bit

5:34 by the 1970s.

5:35 Unfortunately, even 1 cent per bit isn’t cheap enough for storage.

5:39 As previously mentioned,

5:40 an average smartphone photo is around 5 megabytes in size, that’s

5:44 roughly 40 million bits.

5:45 Would you pay 4 hundred thousand dollars to store a photo on core memory?

5:50 If you have that kind of money to drop,

5:52 did you know that Crash Course is on Patreon?

5:54 Right?

5:54 Wink wink.

5:55 Anyway,

5:55 there was tremendous research into storage technologies happening at this time.

5:59 By 1951, Eckert and Mauchly had started their own company,

6:02 and designed a new computer called

6:04 UNIVAC, one of the earliest commercially sold computers.

6:07 It debuted with a new form of computer storage: magnetic tape.

6:10 This was a long, thin and flexible strip of magnetic material, stored in reels.

6:15 The tape could be moved forwards or backwards

6:17 inside of a machine called a tape drive.

6:19 Inside is a write head,

6:20 which passes current through a wound wire to generate a magnetic

6:23 field, causing a small section of the tape to become magnetized.

6:26 The direction of the current sets the polarity,

6:28 again, perfect for storing 1’s and 0’s.

6:30 There was also a separate read head could detect the polarity non-destructively.

6:34 The UNIVAC used half-inch-wide tape with 8 parallel data tracks,

6:38 each able to store 128

6:40 bits of data per inch.

6:41 With each reel containing 1200 feet of tape,

6:44 it meant you could store roughly 15 million

6:46 bits– that’s almost 2 megabytes!

6:48 Although tape drives were expensive,

6:50 the magnetic tape itself was cheap and compact, and for

6:53 this reason, they’re still used today for archiving data.

6:56 The main drawback is access speed.

6:58 Tape is inherently sequential,

7:00 you have to rewind or fast-forward to get to data you

7:03 want.

7:03 This might mean traversing hundreds of feet of tape to retrieve a single byte,

7:07 which is slow.

7:08 A related popular technology in the 1950s and 60s was Magnetic Drum Memory.

7:13 This was a metal cylinder– called a drum–

7:15 coated in a magnetic material for recording data.

7:18 The drum was rotated continuously,

7:19 and positioned along its length were dozens of read and write heads.

7:23 These would wait for the right spot to rotate underneath

7:25 them to read or write a bit of data.

7:27 To keep this delay as short as possible,

7:30 drums were rotated thousand of revolutions per minute!

7:32 By 1953, when the technology started to take off,

7:35 you could buy units able to record 80,000

7:37 bits of data– that’s 10 kilobytes,

7:40 but the manufacture of drums ceased in the 1970s.

7:43 However,

7:43 Magnetic Drums did directly lead to the development of Hard Disk Drives, which

7:47 are very similar, but use a different geometric configuration.

7:50 Instead of large cylinder, hard disks use, well… disks… that are hard.

7:54 Hence the name!

7:55 The storage principle is the same,

7:56 the surface of a disk is magnetic, allowing write and

7:59 read heads to store and retrieve 1’s and 0’s.

8:01 The great thing about disks is that they are thin,

8:04 so you can stack many of them together,

8:06 providing a lot of surface area for data storage.

8:08 That’s exactly what IBM did for the world's first computer with a disk drive:

8:12 the RAMAC 305.

8:13 Sweet name BTW.

8:14 It contained fifty, 24-inch diameter disks,

8:16 offering a total storage capacity of roughly

8:19 5 megabytes.Yess!!

8:20 We’ve finally gotten to a technology that can store a single smartphone photo!

8:25 The year was 1956.

8:27 To access any bit of data,

8:28 a read/write head would travel up or down the stack to the right

8:31 disk, and then slide in between them.

8:34 Like drum memory, the disks are spinning,

8:35 so the head has to wait for the right section

8:38 to come around.

8:39 The RAMAC 305 could access any block of data,

8:42 on average, in around 6/10ths of a second,

8:44 what’s called the seek time.

8:46 While great for storage, this was not nearly fast enough for memory,

8:49 so the RAMAC 305 also

8:50 had drum memory and magnetic core memory.

8:53 This is an example of a memory hierarchy,

8:55 where you have a little bit of fast memory,

8:57 which is expensive, slightly more medium-speed memory,

9:00 which is less expensive, and then

9:02 a lot of slowish memory, which is cheap.

9:04 This mixed approach strikes a balance between cost and speed.

9:07 Hard disk drives rapidly improved and became commonplace by the 1970s.

9:11 A hard disk like this can easily hold 1 terabyte of data

9:14 today– that’s a trillion bytes– or roughly 200,000 five megabyte photos!

9:19 And these types of drives can be bought online for as little as 40 US dollars.

9:23 That’s 0.0000000005 cents per bit.

9:29 A huge improvement over core memory’s 1 cent per bit!

9:32 Also, modern drives have an average seek time of under 1/100th of a second.

9:36 I should also briefly mention a close cousin of hard disks,

9:39 the floppy disk, which is basically

9:41 the same thing, but uses a magnetic medium that’s, floppy.

9:44 You might recognise it as the save icon on some of your applications,

9:47 but it was once

9:48 a real physical object!

9:50 It was most commonly used for portable storage,

9:52 and became near ubiquitous from the mid 1970s

9:54 up to the mid 90s.

9:56 And today it makes a pretty good coaster.

9:58 Higher density floppy disks, like Zip Disks,

10:00 became popular in the mid 1990s, but fell

10:03 out of favor within a decade.

10:04 Optical storage came onto the scene in 1972,

10:07 in the form of a 12-inch “laser disc.”

10:09 However, you are probably more familiar with its later,

10:12 smaller, are more popular cousin,

10:13 the Compact Disk, or CD, as well as the DVD which took off in the 90s.

10:17 Functionally,

10:18 these technologies are pretty similar to hard disks and floppy disks, but

10:21 instead of storing data magnetically,

10:23 optical disks have little physical divots in their

10:25 surface that cause light to be reflected differently,

10:27 which is captured by an optical sensor, and

10:30 decoded into 1’s and 0’s.

10:31 However, today, things are moving to solid state technologies,

10:34 with no moving parts,

10:35 like this hard drive and also this USB stick.

10:38 Inside are Integrated Circuits, which we talked about in Episode 15.

10:41 The first RAM integrated circuits became available in 1972 at 1 cent per bit,

10:46 quickly making magnetic core memory obsolete.

10:49 Today, costs have fallen so far,

10:51 that hard disk drives are being replaced with non-volatile,

10:54 Solid State Drives, or SSDs, as the cool kids say.

10:57 Because they contain no moving parts,

10:58 they don’t really have to seek anywhere, so SSD

11:01 access times are typically under 1/1000th of a second.

11:04 That’s fast!

11:05 But it’s still many times slower than your computer’s RAM.

11:08 For this reason, computers today still use memory hierarchies.

11:11 So, we’ve come along way since the 1940s.

11:14 Much like transistor count and Moore’s law,

11:16 which we talked about in Episode 14, memory

11:18 and storage technologies have followed a similar exponential trend.

11:21 From early core memory costing millions of dollars per megabyte,

11:24 we’re steadily fallen,

11:25 to mere cents by 2000, and only fractions of a cent today.

11:29 Plus, there’s WAY less punch cards to keep track of.

11:31 Seriously,

11:32 can you imagine if there was a slight breeze in that room containing the

11:34 SAGE program?

11:36 62,500 punch cards.

11:38 I don’t even want to think about it.

11:39 I'll see you next week.