The best optimizations I can offer is:
[parity]
light = false
[network]
warp = true
[footprint]
tracing = "off"
fat_db = "off"
pruning = "fast"
pruning_history = 8
pruning_memory = 32
db_compaction = "ssd"
[snapshots]
disable_periodic = true
This disables light mode and enables warp-sync. Warp-sync does only verify PoW of ancient blocks and does not fully compute all historic states while fetching them. However, it stores all blocks on disk.
Transaction tracing and fat DB should be turned off, as this bloats the database. DB compaction should be set to ssd, you are running on SSD, aren't you? Try hdd otherwise.
Pruning should be set to fast to discard historic states, this reduces the database size significantly by around -90%. The minimum possible pruning history is 8. Setting lower would make no sense as this increases the danger of getting caught in chain reorganizations. With 15 seconds block times, reorgs of 8 and less blocks are not uncommon. Pruning memory can be further reduced if you like, but it will always keep at least 8 states afaik.
There is one more option which significantly reduces storage footprint as it does not keep all block history. I'll hide this option behind a spoiler quote to make sure everyone hovering over this below is aware that this is dangerous and should be in no way used in production.
--no-ancient-blocks
Running clients in this mode significantly reduces your footprint by several GB but will allow malicious actors in the network to trick your client into syncing a tampered chain. Do not use unless you understand the implications. Running --light
mode is more secure than this. :)