Hello,

If you ask me a question like:

Amine, is there any new parallel Skiplist algorithm that is more efficient 
than 
your  conccurrent algorithm that uses a distributed read-writer mutex  ?

I have thought more, and i have come with a new Parallel Skiplist algorithm
that is scalable on a read mostly workload, this algorithm uses only 
an mfence on the reader side, that's really cool ! and it elevates the 
weakness 
of the distributed reader-writer mutex that becomes slower and slower on 
the writer side
with more an more threads, in my new concurrent Skiplist algorithm i am 
using one  SeqlockX 
between the search() and the insert() methods , you will find my SeqlockX 
algorithm that avoids 
livelock when there is many writers here:

https://sites.google.com/site/aminer68/scalable-seqlockx

and i am also using one distributed reader-writer mutex between the 
delete() and the 
search() methods... 

This new algorithm is faster than my previous concurrent SkipList algorthm, 
but 
i have made a scalability prediction for my new algorithm and i have 
noticed that since the serial part of the Amdahl's law of the writer side 
is a little bit expensive since the average search in a Skiplist can become 
a little 
bit expensive, so my calculation have proved that you will not get much 
improvement with 
this new algorithm compared to my previous algorithm, you will get around 
3x improvement max,
so i have finally decided to not implement my new concurrent SkipList 
algorithm, so 
i will stay with my current concurrent SkipList algorithm that gives a 
decent scalability.



Thank you,
Amine Moulay Ramdane.





-- 

--- 
You received this message because you are subscribed to the Google Groups 
"Scalable Synchronization Algorithms" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/lock-free/f808bfa3-27cf-4d1e-acaf-f7f422412b8a%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to