Skip to content

Latest commit

 

History

History
55 lines (42 loc) · 2.52 KB

README.md

File metadata and controls

55 lines (42 loc) · 2.52 KB

Aspen

Aspen Dependencies

GNU C++ compiler version 7 g++-7 should be the default compiler for ubuntu18
Installing gcc-7 is simple on ubuntu20: sudo apt install g++-7
With ubuntu22, the process is a little more involved:

echo "deb [arch=amd64] http://archive.ubuntu.com/ubuntu focal main universe" | sudo tee -a /etc/apt/sources.list
sudo apt update
sudo apt install g++-7

jemalloc malloc implementation

sudo apt install libjemalloc-dev

You are done setting up Aspen, return to GraphZeppelin Instructions. Supplementary information is below.

Aspen Experiments

Ingestion Test

The command line arguments to the ingestion_test executable are as follows:

  1. Path to the (binary) stream file
  2. Update batch size. Unlike our system, Aspen is not designed to process updates at the granularity of a single edge. Instead, Aspen only accepts updates in batches of a single type (i.e. either all edge insertions or all edge deletions). This parameter controls how large this batch should be. We used a batch size of 1,000,000 for the results we reported in the paper.
  3. Buffer size. This parameter controls how many edges we should retrieve from external memory at once. We used a buffer size of 1,000,000 for the results we reported in the paper.
  4. Path to file to write timing data.

If you'd like to take a closer look at the actual source code of the speed and memory consumtion test, you may find it in code/tools/ingestion_test.cpp.

Continous Query Test

The command line arguments to the continuous_query_test executable are as follows:

  1. Path to the (binary) stream file
  2. Update batch size. Unlike our system, Aspen is not designed to process updates at the granularity of a single edge. Instead, Aspen only accepts updates in batches of a single type (i.e. either all edge insertions or all edge deletions). This parameter controls how large this batch should be. We used a batch size of 1,000,000 for the results we reported in the paper.
  3. Buffer size. This parameter controls how many edges we should retrieve from external memory at once. We used a buffer size of 1,000,000 for the results we reported in the paper.
  4. Path to file where we record the results of continuous query testing.
  5. Number of CC queries. If this parameter is X, then we perform X connected components queries every 1/X of the way through the stream.

If you'd like to take a closer look at the actual source code of the continuous query test, you may find it in code/tools/continuous_query_test.cpp