How to config or modify the code and shell if test NovaLSM on local machine?

[CharlesJhonson]

hello，I noticed that experiment "We conduct the experiments mostly on a platform consisting of two servers ..." in the paper.

What I want to ask is how to to config or modify the code and script if test NovaLSM with two local machine ? thank you very much.

[ruihong123]

(1) You need to enable tmpfs and set db directory onto it to simulate the speed of the disaggregated memory.

(2) Besides, you need to use my forked repo "https://github.com/ruihong123/NovaLSM", or use "db_bench" branch on the original repo. I made some changes to make the code physically disaggregated.

(3) On the memory server, you need to run "nova_server_main" to enable the disaggregated server.

[CharlesJhonson]

Thank you so much! I previously pulled the original master branch of NovaLSM and extracted the startup parameters of "nova_server_main" based on "nova_lsm_tutorial_exp.sh", but it has not been able to run.

I will try it now following your instructions.

In addition, should I use "nova_lsm_tutorial_exp.sh" or "nova_lsm_tutorial_backup.sh"? They look very similar.

[CharlesJhonson]

I used your forked repo "https://github.com/ruihong123/NovaLSM", and coredumped when running the memory serve "nova_server_main". The startup parameters are configured as follows:

“/home/caoq/novaDB” is a tmpfs dir. `#!/bin/bash

home_dir="/home/caoq/NovaTest/NovaLSM-master" config_dir="$home_dir/config"

recordcount="100000" value_size="1024"

ltc_migration_policy="immediate" enable_range_index="true" num_migration_threads="1" num_sstable_replicas="1" level="2" l0_start_compaction_mb=$((4*1024)) subrange_no_flush_num_keys="100" enable_detailed_db_stats="false"

major_compaction_type="sc" major_compaction_max_parallism="1" major_compaction_max_tables_in_a_set="15"

enable_flush_multiple_memtables="false" recover_dbs="false" num_recovery_threads="32" zipfian_dist_file_path="/tmp/zipfian" dist="uniform" enable_subrange="false" num_log_replicas="1" log_record_mode="none" scatter_policy="power_of_two" number_of_ltcs="2" enable_lookup_index="true" l0_stop_write_mb=$((10*1024)) num_memtable_partitions="64" num_memtables="256" num_rdma_bg_workers="2"

db_path="/home/caoq/novaDB/db/nova-db-$recordcount-$value_size"

num_storage_workers="8"

cc_stoc_files_path="/home/caoq/novaDB/db/stoc_files"

max_stoc_file_size_mb="4" sstable_size_mb="2" ltc_num_stocs_scatter_data_blocks="1"

port=$((10000+RANDOM%1000)) nova_servers="192.168.10.205:$port" echo "server: $nova_servers"

server_id=0

mem_pool_size_gb="30"

ltc_config_path="" ltc_config_path="$config_dir/nova-tutorial-config"

cc_nconn_workers="1" num_rdma_fg_workers="1" num_compaction_workers="1" block_cache_mb="1" row_cache_mb="4096" memtable_size_mb="4" cc_log_buf_size="1024"

rdma_max_msg_size=$((256*1024)) rdma_port=$((20000+RANDOM%1000)) rdma_max_num_sends="32" rdma_doorbell_batch_size="8" enable_rdma="true" enable_load_data="true" use_local_disk="false"

ltc_num_stocs_scatter_data_blocks="1"

cmd="stdbuf --output=0 --error=0 ./nova_server_main --ltc_migration_policy=$ltc_migration_policy --enable_range_index=$enable_range_index --num_migration_threads=$num_migration_threads --num_sstable_replicas=$num_sstable_replicas --level=$level --l0_start_compaction_mb=$l0_start_compaction_mb --subrange_no_flush_num_keys=$subrange_no_flush_num_keys --enable_detailed_db_stats=$enable_detailed_db_stats --major_compaction_type=$major_compaction_type --major_compaction_max_parallism=$major_compaction_max_parallism --major_compaction_max_tables_in_a_set=$major_compaction_max_tables_in_a_set --enable_flush_multiple_memtables=$enable_flush_multiple_memtables --recover_dbs=$recover_dbs --num_recovery_threads=$num_recovery_threads --sampling_ratio=1 --zipfian_dist_ref_counts=$zipfian_dist_file_path --client_access_pattern=$dist --memtable_type=static_partition --enable_subrange=$enable_subrange --num_log_replicas=$num_log_replicas --log_record_mode=$log_record_mode --scatter_policy=$scatter_policy --number_of_ltcs=$number_of_ltcs --enable_lookup_index=$enable_lookup_index --l0_stop_write_mb=$l0_stop_write_mb --num_memtable_partitions=$num_memtable_partitions --num_memtables=$num_memtables --num_rdma_bg_workers=$num_rdma_bg_workers --db_path=$db_path --num_storage_workers=$num_storage_workers --stoc_files_path=$cc_stoc_files_path --max_stoc_file_size_mb=$max_stoc_file_size_mb --sstable_size_mb=$sstable_size_mb --ltc_num_stocs_scatter_data_blocks=$ltc_num_stocs_scatter_data_blocks --all_servers=$nova_servers --server_id=$server_id --mem_pool_size_gb=$mem_pool_size_gb --use_fixed_value_size=$value_size --ltc_config_path=$ltc_config_path --ltc_num_client_workers=$cc_nconn_workers --num_rdma_fg_workers=$num_rdma_fg_workers --num_compaction_workers=$num_compaction_workers --block_cache_mb=$block_cache_mb --row_cache_mb=$row_cache_mb --memtable_size_mb=$memtable_size_mb --cc_log_buf_size=$cc_log_buf_size --rdma_port=$rdma_port --rdma_max_msg_size=$rdma_max_msg_size --rdma_max_num_sends=$rdma_max_num_sends --rdma_doorbell_batch_size=$rdma_doorbell_batch_size --enable_rdma=$enable_rdma --enable_load_data=$enable_load_data --use_local_disk=$use_local_disk" echo "$cmd" mkdir -p $cc_stoc_files_path && mkdir -p $db_path && $cmd `

run the bash,and coredump info as follow: