SDSTOR-21407 Make per-service dev_type and size_pct configurable in i…#411
SDSTOR-21407 Make per-service dev_type and size_pct configurable in i…#411xiaoxichen wants to merge 2 commits intoeBay:stable/v4.xfrom
Conversation
…nit_homestore
Add HSDevTypeOverride enum and per-service {meta,log,index,replication}_{dev_type,size_pct}
fields to HSBackendSettings. Defaults are AUTO/-1.0 so existing hybrid/single mode logic
is fully preserved. When a config value is set it overrides the built-in default via
resolve_dev_type() and resolve_size_pct() helpers.
Also tune standalone (single-device) defaults: meta/index 5%- >1%, replication 79%->87%.
Signed-off-by: Xiaoxi Chen <xiaoxchen@ebay.com>
|
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## stable/v4.x #411 +/- ##
==============================================
Coverage ? 53.01%
==============================================
Files ? 36
Lines ? 5302
Branches ? 657
==============================================
Hits ? 2811
Misses ? 2199
Partials ? 292 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
| }); | ||
| } else { | ||
| auto run_on_type = has_fast_dev ? homestore::HSDevType::Fast : homestore::HSDevType::Data; | ||
| LOGD("Running with Single mode, all service on {}", run_on_type); |
There was a problem hiding this comment.
nit: if we have different final HSDevType for different service, it is better to print them respectively.
| - **Allocator overhead:** ~0.2% of S | ||
| - Uses `FixedBlkAllocator` (4KB fixed block size) | ||
| - Maintains all free blocks in memory via `folly::MPMCQueue` | ||
| - Each slot: 12 bytes (4B blk_num_t + 8B atomic sequence) |
There was a problem hiding this comment.
sequence is atomic < uint32 >, so it is 4B, and each slot is 8B?
There was a problem hiding this comment.
Thanks, nice catch, I copied this from the GDB result from Mehdi but actually he was in older version where BlkId (6B) was used for MPMCQ , 6+4=10 then padded to 12.
In 7.x we are using blk_num_t instead of blk_idwhich is 4B , then no padding is needed
| ``` | ||
| Worst case assuming blob size is 8KB, the Index size should be 0.61% of DataSize | ||
|
|
||
| (DataSize/8K)*(4KB/167)*1.01 ==> DataSize *1.01/167 = DataSize * 0.61%. |
There was a problem hiding this comment.
8K/ 4K = 2？
it should be ( (DataSize/8K) / 167 ) *4k * 1.01 = DataSize * 0.3%.
| Now we use 45% of META drive (200GB*45% =90GB) on HDD SKUs. The number is generally correct as worst case we can consume up to 60GB. | ||
| Shrinking it down from 90GB to 60GB saves 60MB memory which is negligible. | ||
| #### [QLC SKU] | ||
| 5% of QLC size makes it into 5721.96 GB, as a result it consumes 18GB memory for allocator. |
There was a problem hiding this comment.
can you also put the capacity of a QLS SSD here to make it more clear
| 3. Hard memory = Index disk / 4096 × 12 bytes | ||
| (Allocator) (Always resident, non-negotiable) | ||
|
|
||
| 4. Soft memory = min(Index disk × 10%, io_mem_size × 65%) |
There was a problem hiding this comment.
what does Index disk × 10% mean?
There was a problem hiding this comment.
trying to say keeping 10% of index in the memory for hot data (if any)
Signed-off-by: Xiaoxi Chen <xiaoxchen@ebay.com>
3 participants
…nit_homestore
Add HSDevTypeOverride enum and per-service {meta,log,index,replication}_{dev_type,size_pct} fields to HSBackendSettings. Defaults are AUTO/-1.0 so existing hybrid/single mode logic is fully preserved. When a config value is set it overrides the built-in default via resolve_dev_type() and resolve_size_pct() helpers.
Also tune standalone (single-device) defaults: meta/index 5%- >1%, replication 79%->87%.