Initial implementation of a new lockless hashtable mpmc

[danielealbano]

This PR implements a new lockless multi producer and multi consumer parallel hashtable which is capable of achieving amazing numbers (included at the end of the summary)!

This new hashtable works in a very different way from the previous one although the goal keeps being the same: spread the contention of the threads as much as possible and reduce the amount of times the data in memory are changed.

The old approach was relying on a combination of memory fencing and userspace spinlocks to perform lockless GET operations meanwhile using userspace spinlocks for the SET and DEL operations with a lock every 14 buckets, the results were great as an hashtable with 10000 was being controlled by about 714 locks, but the downside was a lot of memory changes, also impacting the cache of the cores, done by 1 single thread to write a value.

The new approach relies on a combination of approaches and components:

• an epoch operation queue, each time an operation is triggered with the hashtable an operation is pushed to the queue and marked as completed once done, guaranteeing that the data read by the thread after pushing the new operation to the queue will never be deleted until the operation itself is completed
• an epoch garbage collector, which operates in combination with the epoch operation queue, every time a bucket is deleted the associated key value data are not deleted immediately but insted staged for deletion, will only be deleted once all the operation started before the staging are marked as completed
• a 3 pass approach to insert new buckets which guarantees that at some point if multiple threads are fighting to insert the same key one will be successful
  • which becomes 4 pass in case an insertion is carried out during an upsize if the key is new
• the key value data are not stored inside the hashtable itself but as a pointer to an external structure, this makes a bucket much larger (16 bytes) but drammatically speed up the copy time and makes possible to use 128 bit atomic operations to update a bucket
• pointer tagging to store status in the pointer of the key value itself
• a transaction spinlock, as it's in use in the current hashtable, for the single/multi key transactions and the Read-Modify-Write operations

The hashtable will implement some extra optimizations to box the upper level data (the storagedb_entry_index) into the lower level data (the key value itself) using a combination of an union and the code in the header using some defines to create a typed version of the hashtable.

This new hashtable will provide much better performances when batching operations or operating in a cluster.

The PR includes new tests and new benchmarks, used to generate the numbers initially mentioned.

The PR is drammatically large:

• 74 commits
• 18 files changed
• 4465 new lines
• only 71 deletions

Talking about numbers, here some

Operation	Threads	V1 - Million Op/s	V2 - Million Op/s
INSERT	1	3.22191	5.02507
INSERT	2	6.26603	9.74552
INSERT	4	11.73655	19.20719
INSERT	8	24.03989	38.92227
INSERT	16	42.58111	72.36672
INSERT	32	69.45982	129.38364
INSERT	64	109.68172	197.67271
UPDATE	1	3.18695	5.55167
UPDATE	2	6.16648	11.06459
UPDATE	4	11.53689	21.79368
UPDATE	8	22.8819	43.98683
UPDATE	16	37.42774	83.61501
UPDATE	32	63.96726	143.30408
UPDATE	64	89.5212	236.87745

The V2 hashtable is between 2 and 2.5 times faster than the V1, the current implementation.

The hardware used for benchmarking was an EPYC 7502P (32 core, 64 hw thread, default bios settings) with 256GB RDIMM DDR4 3200mhz.

Closes #103

[codecov]

Codecov Report

Base: 82.34% // Head: 82.74% // Increases project coverage by +0.41% 🎉

Coverage data is based on head (a602326) compared to base (48ba7eb).
Patch coverage: 91.43% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #262      +/-   ##
==========================================
+ Coverage   82.34%   82.74%   +0.41%     
==========================================
  Files         157      158       +1     
  Lines        9795    10240     +445     
==========================================
+ Hits         8065     8473     +408     
- Misses       1730     1767      +37     

Flag	Coverage Δ
unittests	82.74% <91.43%> (+0.41%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
...rc/data_structures/hashtable/spsc/hashtable_spsc.h	100.00% <ø> (ø)
...rc/data_structures/hashtable_mpmc/hashtable_mpmc.c	91.03% <91.03%> (ø)
...rc/data_structures/hashtable/spsc/hashtable_spsc.c	94.51% <100.00%> (-0.06%)	⬇️
src/epoch_gc.c	99.03% <100.00%> (+0.05%)	⬆️
src/random.c	100.00% <100.00%> (ø)
src/xalloc.c	97.14% <0.00%> (+1.43%)	⬆️
src/spinlock.h	94.44% <0.00%> (+5.56%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.