[GitHub] [couchdb] nickva commented on issue #4308: Benchmark Replicator Optimizations in CouchDB 3.3

[GitBox <gi...@apache.org>]

nickva commented on issue #4308:
URL: https://github.com/apache/couchdb/issues/4308#issuecomment-1353820482

   #### Benchmark Setup:
   
   Couchdyno: https://github.com/cloudant-labs/couchdyno
   Document count: **1000000**
   Document size: **10Kb**
   Document data: https://gist.github.com/nickva/9c0223bedd6f225efd81df51819ae76e
   
   ```python
   > fh = open("generated10k.json")
   > jsonstr = fh.read()
   > json10kb = json.loads(jsonstr)
   > len(jsonstr)
   11487
   ```
   
   **Source** (UK): 8 vCPUs, Q=8
   **Target** (US East): 80 vCPUs, 64GBs memory (replication jobs run here), Q=8
   
   ```
   couchdyno config:
    >    -  connection_timeout = 30000
    >    -  create_target = False
    >    -  cycle_timeout = 28800
    >    -  delete_before_updating = False
    >    -  http_connections = 20
    >    -  num_branches = 1
    >    -  num_docs = 100000
    >    -  num_revs = 1
    >    -  reset_source = True
    >    -  reset_target = True
    >    -  skip_rev_check = True
    >    -  worker_processes = 4
   ```
   
   Using these settings:
   ```ini
   [chttpd]
   bulk_get_use_batches = true
   
   [replicator]
   use_bulk_get = true
   
   [ioq]
   concurrency = 20
   
   [ioq.bypass]
   interactive = true
   ```
   
   #### Before (Cloudant release 8310)
   
   ```
   r.replicate_1_to_n_and_compare(1, num=1000000, revs=1, branches=1, normal=True, reset_target=True, reset_source=True, src_params=json10kb)
   > replicated in 4578 sec
   > replicated in 4583 sec
   > replicated in 4553 sec
   ```
   
   #### After (Cloudant release 8341)
   
   ```
   r.replicate_1_to_n_and_compare(1, num=1000000, revs=1, branches=1, normal=True, reset_target=True, reset_source=True, src_params=json10kb)
   > replicated in 1518 sec
   > replicated in 1548 sec
   > replicated in 1518 sec
   ```
   
   
   ### Results:
   
   ```
   In [1]: with_optimizations = (1518 + 1548 + 1518) / 3
   In [2]: before_optimizations = (4578 + 4553 + 4553) / 3
   In [3]: round(before_optimizations / with_optimizations, 1)
   Out[3]: 3.0
   ```
   
   We got a **3x** speed-up for 1 million 10kb docs replicated across the Atlantic


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: notifications-unsubscribe@couchdb.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org