IS 420 Final

storage hierarchy

1. primary

2. secondary

3. tertiary

primary storage

fastest media but volatile

cache, main memory

secondary storage (on-line storage)

nonvolatile, moderately fast access time

flash memory, magnetic disks

tertiary storage (off-line storage)

nonvolatile, slow access time

magnetic tape, optical storage

structure of a magnetic hard disk mechanism

1. track

2. sector

3. cylinder

4. spindle

5. platter

6. arm

7. read-write head

track

circular path on the surface of the platter where data is stored magnetically

16k per platter

sector

smallest unit of data that can be read or written

512 bytes

cylinder

a cylindrical intersection through the stack of platters in a disk, centered around the disk's spindle

access time

the time it takes from start of a read or write request to when data transfer begins.

what does the access time consist of

1. seek time

2. rotational latency

seek time

time to position the arm over the correct track

rotational latency

time it takes for the sector to be accessed to appear under the head

solid state drives (SSD)

based on flash memory

contains a controller as a bridge between the OS and the NAND block storage

SSD architecture

1. interface

2. controller

3. buffer/cache

4. flash memory components

SSD interface

Electrical IF between processor and SSD (commonly SATA, IDE)

SSD controller

used for error correction, wear leveling, bad block mapping, and encryption

SSD buffer/cache

high speed RAM

SSD flash memory components

NAND or NOR technology (2GB-64GB)

SSD vs magnetic disk

• SSD has no moving parts (arm of magnetic disk)

• SSD has a faster boot and access time (flash memory vs rotation)

• SSD has faster data transfer rate

• SSD is more costly

• SSD not as reliable when faced with power failure

how many primary indexes can a table have

1

how many secondary indexes can a table have

many

when is it suggested to create an index

after adding data

is an index a table

no, it references data in a table

database index

brings relevant records

1. first query brings all records from disk

2. second query doesn’t need to bring all

composite index

multiple sorted columns for queries containing multiple search conditions

triggers

specification of database event associated with a database table

what can triggers NOT accept

input parameters, commit or rollback statements

trigger timing

before/after event

trigger levels

scope affected by the trigger

for each row/statement

trigger event

statement that causes trigger to execute (fire)

UPDATE, INSERT, DELETE

BEFORE trigger

action first, event second

AFTER trigger

event first, action second

NEW trigger

rowtype variable that stores values after the modification

OLD trigger

rowtype variable that stores values before the modification

motivation for NoSQL

• scalability

• flexibility

• availability

• cost control

how to maintain data availability in NoSQL environments

use primary and backup servers

2-phase commit for updates

eventual consistency

2 DB servers temporarily do not have the same value (inconsistent) but eventually become consistent

what does CAP in CAP theorem stand for

• consistency

• availability

• partition tolerance

CAP theorem

distributed databases can’t have CAP all at the same time

no more than 2 at a time

consistency

consistent copies of data on different servers

availability

response to any query

partition tolerance

system can still operate correctly even if there is an inability to send messages between servers

4 types of NoSQL DBs

1. key value

2. document

3. column

4. graph

relational vs column DBs

• column DB don’t have tables; object info in one long row

• column DB rows have no fixed structure

• there are no joins in column DB

which was the first column family DB

Google BigTable (2006)

gossip protocol

allows for quick exchange of information

information entropy

the actual amount of information stored in this variable

sum of information of all particular states

anti entropy

process to determine differences in replicas

detect and resolve inconsistencies w a min amount of data exhange

what is indexing used for

allow for a rapid lookup of data in
table

namespace

they help avoid naming conflicts and make it easier to manage objects

quorum

the number of servers that must respond to a read or write operation, for the operation to be considered complete