absolutely
so basically dragonfly does not even bother using its CPU
Also here’s most of the spec sheet from the supplier on that machine (at least the important parts):
1x Dell PowerEdge R660
1x Intel Xeon Platinum 8468 2.1G 48c/96t
2x 960GB SSD vSAS Read Intensive 12Gbps (Raid 1)
16x 32GB DIMM DDR5 4800MT/s Dual Rank
1x Broadcom 10/25GbE SFP NIC
can you please paste 'redis-cli INFO ALL` command?
it does not contain any data there
yep no worries
# Server
redis_version:6.2.11
dragonfly_version:df-v1.12.1
redis_mode:standalone
arch_bits:64
multiplexing_api:iouring
tcp_port:6379
thread_count:96
uptime_in_seconds:581841
uptime_in_days:6
# Clients
connected_clients:905
client_read_buffer_bytes:234240
blocked_clients:0
dispatch_queue_entries:0
# Memory
used_memory:122685944
used_memory_human:117.00MiB
used_memory_peak:886319928
used_memory_rss:1427558400
used_memory_rss_human:1.33GiB
used_memory_peak_rss:2169511936
comitted_memory:1773666304
maxmemory:427967859916
maxmemory_human:398.58GiB
object_used_memory:18446744072240473152
table_used_memory:116698312
num_buckets:1401120
num_entries:13154
inline_keys:9997
strval_bytes:1637924352
updateval_amount:-791269584
listpack_blobs:0
listpack_bytes:0
small_string_bytes:422192
pipeline_cache_bytes:0
dispatch_queue_bytes:0
dispatch_queue_peak_bytes:0
client_read_buffer_peak_bytes:409600
cache_mode:store
maxmemory_policy:noeviction
# Stats
total_connections_received:1805562
total_commands_processed:25326333240
instantaneous_ops_per_sec:66828
total_pipelined_commands:0
total_net_input_bytes:1304468534320
total_net_output_bytes:232612687595
instantaneous_input_kbps:-1
instantaneous_output_kbps:-1
rejected_connections:-1
expired_keys:1879868192
evicted_keys:0
hard_evictions:0
garbage_checked:1496957
garbage_collected:59133
bump_ups:0
stash_unloaded:91332
oom_rejections:0
traverse_ttl_sec:12
delete_ttl_sec:10
keyspace_hits:2285631900
keyspace_misses:19639946432
total_reads_processed:25258620568
total_writes_processed:25224819403
defrag_attempt_total:0
defrag_realloc_total:0
defrag_task_invocation_total:0
eval_io_coordination_total:0
eval_shardlocal_coordination_total:35617462
eval_squashed_flushes:0
tx_schedule_cancel_total:0
# Tiered
tiered_entries:0
tiered_bytes:0
tiered_reads:0
tiered_writes:0
tiered_reserved:0
tiered_capacity:0
tiered_aborted_write_total:0
tiered_flush_skip_total:0
# Persistence
last_save:1701127319
last_save_duration_sec:0
last_save_file:
loading:0
rdb_changes_since_last_save:3386506308
# Replication
role:master
connected_slaves:0
master_replid:1ebafee12327bb0c44c62a5c8e179c1a2e8aebbd
# Commandstats
cmdstat_CLIENT:calls=40,usec=127,usec_per_call=3.175
cmdstat_COMMAND:calls=9,usec=21307,usec_per_call=2367.44
cmdstat_EVAL:calls=35617462,usec=1078148396,usec_per_call=30.2702
cmdstat_EXPIRE:calls=30289476,usec=1421746,usec_per_call=0.0469386
cmdstat_GET:calls=5962461194,usec=123420008405,usec_per_call=20.6995
cmdstat_HELLO:calls=20,usec=668,usec_per_call=33.4
cmdstat_INCR:calls=35617462,usec=6421336,usec_per_call=0.180286
cmdstat_INFO:calls=28,usec=28611,usec_per_call=1021.82
cmdstat_LATENCY:calls=12,usec=628,usec_per_call=52.3333
cmdstat_LPOP:calls=15897210221,usec=283684521484,usec_per_call=17.8449
cmdstat_PING:calls=35736612,usec=114230685,usec_per_call=3.19646
cmdstat_RPUSH:calls=35617962,usec=787849303,usec_per_call=22.1194
cmdstat_SET:calls=3293782736,usec=69795708123,usec_per_call=21.1901
unknown_LATEHCY:1
unknown_VERSION:1
# Search
search_memory:0
search_num_indices:0
search_num_entries:0
# Errorstats
syntax_error:17
unknown_cmd:2
# Keyspace
db0:keys=13154,expires=3157,avg_ttl=-1
# Cpu
used_cpu_sys:392505.145294
used_cpu_user:222523.791315
used_cpu_sys_children:0.0
used_cpu_user_children:0.0
used_cpu_sys_main_thread:6178.161887
used_cpu_user_main_thread:3333.97309
# Cluster
cluster_enabled:0
ok, according to these stats, Dragonfly behaves perfectly well, it does not even feel any load.
I suspect the latency is either from networking or on the client side
yeah totally, that’s something I’ve observed from the outside, just cant verify it from the stats…
I feel as if the latency is just wrong in our case, I can’t imagine 1s latency for anything we run within our datacenter (redis/memcached/kafka) most are very low 1-5ms on avg
I mean 1ms latency is also high 
of course it’s not 1s
I saw your memtier run with average latency of 1ms
i think it’s too high
okay
is there common tuning steps? or do you think it’s a networking issue?
i mean, maybe it’s ok - I am not familiar with on-prem networking standards so much
the test I ran was from a VM in our VMWare cluster which is in the same, or a few racks down from this new machine