*Rani of Hearts-Reena Kapoor Paristan#55*The Cave Woman - Page 7

Created

Last reply

Replies

1199

Views

39906

Users

5

Likes

80

Frequent Posters

MasoomaBukhari thumbnail
Anniversary 15 Thumbnail Group Promotion 8 Thumbnail + 4
Posted: 4 years ago

so multiprocessing is aka multicore .. hmm didnt know that from attending lec.. probably slept πŸ€ͺ

MasoomaBukhari thumbnail
Anniversary 15 Thumbnail Group Promotion 8 Thumbnail + 4
Posted: 4 years ago

ok i wanna write down stuff which is not in slides but might appear... MIGHT

MasoomaBukhari thumbnail
Anniversary 15 Thumbnail Group Promotion 8 Thumbnail + 4
Posted: 4 years ago

SMP - symmetric multiprocessing architecture

multiple processors connected to memory with single bus

all processors treated equally

no processor reserved for specific purpose

advantages:

has cost effective ways to increase throughput

disadvantages:

limits on scalability due to cache coherence and shared objects

MasoomaBukhari thumbnail
Anniversary 15 Thumbnail Group Promotion 8 Thumbnail + 4
Posted: 4 years ago

Image

  • lacks cache affinity (i dont know why.. yaad ni aarha mam ne kya btaya tha )
  • does not scale well
  • sync problem
Edited by MasoomaBukhari - 4 years ago
MasoomaBukhari thumbnail
Anniversary 15 Thumbnail Group Promotion 8 Thumbnail + 4
Posted: 4 years ago

Image

multilevel queue  ^

adv: 

no sync overhead

provides cache affinity

disadv:

load imbalancing

MasoomaBukhari thumbnail
Anniversary 15 Thumbnail Group Promotion 8 Thumbnail + 4
Posted: 4 years ago

for load imabalanced issue

we can migrate processes to other queues

-when is it ok to migrate?

-scalability problems?

-how often to migrate?

MasoomaBukhari thumbnail
Anniversary 15 Thumbnail Group Promotion 8 Thumbnail + 4
Posted: 4 years ago

group scheduling

Image

so basically related threads different CPUs pe hote he lakin ek hi time pe start aur end hote he taake sync kr ske πŸ‘πŸΌπŸ˜ƒ

itni si baat thi.. lol

MasoomaBukhari thumbnail
Anniversary 15 Thumbnail Group Promotion 8 Thumbnail + 4
Posted: 4 years ago

challenges of modern computer hardware

NUMA - non uniform memory access latencies

synchronization costs

gap between cpu and memory latencies

cache coherence cost

MasoomaBukhari thumbnail
Anniversary 15 Thumbnail Group Promotion 8 Thumbnail + 4
Posted: 4 years ago

load balancer should work on:

not number of threads

not weight

but on

load = weight + average CPU utilization

-cgroup -> load further divided by total number of threads of cgroup

Edited by MasoomaBukhari - 4 years ago
MasoomaBukhari thumbnail
Anniversary 15 Thumbnail Group Promotion 8 Thumbnail + 4
Posted: 4 years ago

cache locality:

temporal locality:

a piece of data can be accessed again in near future

spatial locality:

if data is accessed at address x then data near that address can also be accessed in near future