caching Flashcards
cold storage
data not accessed frequently
what is an example of something that would be stored on cold storage
backup data
hard disk
rotating and magnetic
usually used for hard storage
solid state drive
flash memory with fast access times
what type of memory are disks
non volatile
network file system
remote server/cloud based storage
what type of memory is ram
volatile
how is ram packaged
as a chip
what are the 2 types of ram
static and dynamic
sram dram
sram
shorter access time
has 6x more transistors to store bits than dram making it more expensive
dram
usually used as memory on personal devices as its cheaper than sram
registers
v small on chip memory
what is the issue with the cpu memory gap
cpu speed gets bottlenecked by memory speed
what is the cpu memory gap
cpu speed is increasing faster than memory access time
cache
a smaller faster storage device that acts as a staging area for a subset of data in a larger slower device
what is the goal of cache
achieve access speeds of the fastest memory whilst paying the cost of the cheapest
locality principle
accessing a small localised subset of memory
temporal locality
recently accessed data likely to be accessed again in the near future
spacial locality
data near recently accessed addresses likely to be accessed
how do we exploit temporal memory
keep the recently accessed data in the cache
how do we exploit spacial memory
prefetch data before its requested
how do we calculate the cache hit rate
number of hits/ number of lookups
row major order
elements in the same row of an array are stored next to eachother in memory
stride k reference pattern
k = the unit of jumps between memory addresses being accessed
how does the cache work
read the cache to look for the block
if its already there then you get a cache hit
if not then you get a cache miss
if its a miss then you fetch it from slower memory
if the cache isnt full then add the block to it
if its full then you need to evict an existing block based on the cache eviction policy
what are 3 examples of a cache eviction policy
belady’s algorithm
least recently used
first in first out
belady’s algorithm cache eviction policy
evicts the block that wont be used for the longest time
what is the issue with belady’s algorithm
you will need to know future lookups therefore its not always practical
least recently used cache eviction policy
evicts the block that hasnt been used in the longest time
how can we implement the least recently used cache eviction policy
a doubly linked list maintains the block from most to least recently used
a hash table implemented in a dictionary is used for lookups
fifo cache eviction policy
the cache implements a queue
what is the issue with fifo cache eviction policy
belady’s anomaly; increasing the cache size leads to a higher number of misses
you may be removing the entry that is most frequently used
server side caching
stores data temporarily on the server to speed up future access to that data
usually as a db query result
what are some benefits of server side caching
allows near real time communication by providing low latency
manages popular shared content
massive scalability
allows redundancy and replication ensuring data availability and fault tolerance
how do content distribution networks work
a client requests a url for a video
the client resolves the domain name from the local dns
the dns returns the domain name to the client who then passes it to the cnd to get the content
the edge serve has cached copies of static content originally hosted on the origin server
when it is first requested it is fetched from the og server and is then cached for resuse
what does an edge/ cdn server do
replicated the content providers data
