hyrise: tpc-h compression seems not work well

I want to test the memory usage for tpc-h. use hyriseConsole to run the tpch.

(release)> generate_tpch 5
Generating all TPCH tables (this might take a while) ...
- Loading/Generating tables
- Loading/Generating tables done (54 s 805 ms)
- Encoding tables (if necessary) and generating pruning statistics
-  Encoding 'nation' - encoding applied (375 µs 162 ns)
-  Encoding 'region' - encoding applied (375 µs 872 ns)
-  Encoding 'supplier' - encoding applied (689 ms 556 µs)
-  Encoding 'customer' - encoding applied (1 s 644 ms)
-  Encoding 'partsupp' - encoding applied (2 s 104 ms)
-  Encoding 'part' - encoding applied (2 s 669 ms)
-  Encoding 'orders' - encoding applied (12 s 411 ms)
-  Encoding 'lineitem' - encoding applied (1 min 6 s)
- Encoding tables and generating pruning statistic done (1 min 6 s)
- Adding tables to StorageManager and generating table statistics
-  Added 'nation' (952 µs 576 ns)
-  Added 'region' (1 ms 118 µs)
-  Added 'supplier' (132 ms 589 µs)
-  Added 'customer' (2 s 34 ms)
-  Added 'part' (2 s 910 ms)
-  Added 'partsupp' (10 s 134 ms)
-  Added 'orders' (21 s 328 ms)
-  Added 'lineitem' (1 min 9 s)
- Adding tables to StorageManager and generating table statistics done (1 min 9 s)
- No indexes created as --indexes was not specified or set to false

The memory used :

   PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND
 25958 root      20   0  152.1g  12.6g  18432 S   0.0  2.5  29:00.69 hyriseConsole

So the data is about 5G, but the memory usage is 12.6G.

Then I modified the default encoding type from dictionary to lz4 for generate_and_store function.

--- a/src/benchmarklib/abstract_table_generator.cpp
+++ b/src/benchmarklib/abstract_table_generator.cpp
@@ -207,11 +207,12 @@ void AbstractTableGenerator::generate_and_store() {
     for (auto& table_info_by_name_pair : table_info_by_name) {
       const auto& table_name = table_info_by_name_pair.first;
       auto& table_info = table_info_by_name_pair.second;
+      auto encoding_config = EncodingConfig{SegmentEncodingSpec{EncodingType::LZ4}};

       const auto encode_table = [&]() {
         Timer per_table_timer;
         table_info.re_encoded =
-            BenchmarkTableEncoder::encode(table_name, table_info.table, _benchmark_config->en
+            BenchmarkTableEncoder::encode(table_name, table_info.table, encoding_config);
         auto output = std::stringstream{};
         output << "-  Encoding '" + table_name << "' - "
                << (table_info.re_encoded ? "encoding applied" : "no encoding necessary") <<

The memory used :

   PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND
 57645 root      20   0  153.4g   4.3g  19200 S   0.0  0.8  94:18.67 hyriseConsole

I want to reduce the memory usage as much as possible. AFAIK many column store database can reach a high compression ratio. So Is there any best practice for memory tuning?