Estimates how parallel batches count affects load time of same amount of persistent data.
The lesser score (average time load), the better.

Results in short: looks like 1 batch per thread is enough. But we can keep x2 for network issues.

servers - server nodes
maxDsOps - streamer's batches per node
sendMsgDelay - simulated network delay or write responses.

==================================================================================================================================================================
Benchmark                                                    (cacheWriteMode)  (maxDsOps)  (sendMsgDelay)  (servers)  (walMode)  Mode  Cnt   Score    Error  Units

SERVERS: 2
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated      PRIMARY_SYNC           4               3          2   LOG_ONLY  avgt    5  11,825 ± 13,959   s/op
>>> [Load time is same since `CPUs * 1` / 16 batches]
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated      PRIMARY_SYNC           8               3          2   LOG_ONLY  avgt    5   9,138 ±  5,238   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated      PRIMARY_SYNC          16               3          2   LOG_ONLY  avgt    5   7,334 ±  5,799   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated      PRIMARY_SYNC          32               3          2   LOG_ONLY  avgt    5   7,247 ±  5,518   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated      PRIMARY_SYNC          64               3          2   LOG_ONLY  avgt    5   8,529 ±  1,681   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated      PRIMARY_SYNC         128               3          2   LOG_ONLY  avgt    5   9,001 ± 2,424   s/op
=== FULL_SYNC:
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated         FULL_SYNC           4               3          2   LOG_ONLY  avgt    5  13,336 ±  4,264   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated         FULL_SYNC           8               3          2   LOG_ONLY  avgt    5  10,416 ±  3,984   s/op
>>> [Load time is same  `CPUs * 1`]
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated         FULL_SYNC          16               3          2   LOG_ONLY  avgt    5   8,394 ±  5,186   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated         FULL_SYNC          32               3          2   LOG_ONLY  avgt    5   9,273 ±  8,043   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated         FULL_SYNC          64               3          2   LOG_ONLY  avgt    5   8,088 ±  6,422   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated         FULL_SYNC         128               3          2   LOG_ONLY  avgt    5   8,809 ± 5,519   s/op

SERVERS: 3
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated      PRIMARY_SYNC           4               3          3   LOG_ONLY  avgt    5  14,594 ± 10,666   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated      PRIMARY_SYNC           8               3          3   LOG_ONLY  avgt    5  15,176 ±  7,848   s/op
>>> [Load time is same  `CPUs * 1`]
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated      PRIMARY_SYNC          16               3          3   LOG_ONLY  avgt    5   9,578 ±  4,345   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated      PRIMARY_SYNC          32               3          3   LOG_ONLY  avgt    5  10,280 ±  4,121   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated      PRIMARY_SYNC          64               3          3   LOG_ONLY  avgt    5   8,976 ±  3,753   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated      PRIMARY_SYNC         128               3          3   LOG_ONLY  avgt    5   9,295 ± 4,269   s/op
=== FULL_SYNC:
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated         FULL_SYNC           4               3          3   LOG_ONLY  avgt    5  21,961 ± 12,751   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated         FULL_SYNC           8               3          3   LOG_ONLY  avgt    5  16,261 ± 12,808   s/op
>>> [Load time is same  `CPUs * 1`]
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated         FULL_SYNC          16               3          3   LOG_ONLY  avgt    5  10,426 ±  9,156   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated         FULL_SYNC          32               3          3   LOG_ONLY  avgt    5  10,237 ±  2,163   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated         FULL_SYNC          64               3          3   LOG_ONLY  avgt    5  10,551 ±  3,310   s/op
JmhPersistentStreamerReceiverBenchmark.benchDefaultIsolated         FULL_SYNC         128               3          3   LOG_ONLY  avgt    5   9,811 ± 2,689   s/op
==================================================================================================================================================================

OS: Linux void 5.18.0-4-amd64 #1 SMP PREEMPT_DYNAMIC Debian 5.18.16-1 (2022-08-10) x86_64 GNU/Linux

JMH 1.13
VM version: JDK 11.0.2, VM 11.0.2+9
VM options: -Xms2g -Xmx2g -server -XX:+AlwaysPreTouch

CPU:
vendor_id       : AuthenticAMD
cpu family      : 25
model           : 80
model name      : AMD Ryzen 7 5800U with Radeon Graphics
cpu MHz         : 1600.000
cache size      : 512 KB
siblings        : 16
cpu cores       : 8
cpuid level     : 16

Disk:
`dd if=/dev/zero of=/tmp/test1.img bs=1G count=1 oflag=dsync`: 1073741824 bytes (1,1 GB, 1,0 GiB) copied, 1,25661 s, 854 MB/s
`dd if=/dev/zero of=/tmp/test2.img bs=512 count=1000 oflag=dsync`: 512000 bytes (512 kB, 500 KiB) copied, 2,80265 s, 183 kB/s