Selective Parallel Loading of Large-Scale Compressed Graphs with ParaGrapher – arXiv Version

PDF version
DOI: 10.48550/arXiv.2404.19735

Comprehensive evaluation is one of the basis of experimental science. In High-Performance Graph Processing, a thorough evaluation of contributions becomes more achievable by supporting common input formats over different frameworks. However, each framework creates its specific format, which may not support reading large-scale real-world graph datasets. This shows a demand for high-performance libraries capable of loading graphs to (i) accelerate designing new graph algorithms, (ii) to evaluate the contributions on a wide range of graph algorithms, and (iii) to facilitate easy and fast comparison over different graph frameworks.

To that end, we present ParaGrapher, a high-performance API and library for loading large-scale and compressed graphs. ParaGrapher supports different types of requests for accessing graphs in shared- and distributed-memory and out-of-core graph processing. We explain the design of ParaGrapher and present a performance model of graph decompression, which is used for evaluation of ParaGrapher over three storage types.

Our evaluation shows that by decompressing compressed graphs in WebGraph format, ParaGrapher delivers up to 3.2 times speedup in loading and up to 5.2 times speedup in end-to-end execution in comparison to the binary and textual formats.

ParaGrapher is available online on


  title = { Selective Parallel Loading of Large-Scale 
            Compressed Graphs with ParaGrapher}, 
  author = { {Mohsen} {Koohi Esfahani} and Marco D'Antonio and 
             Syed Ibtisam Tauhidi and Thai Son Mai and 
             Hans Vandierendonck},
  year = {2024},
  eprint = {2404.19735},
  archivePrefix = {arXiv},
  primaryClass = {cs.AR},
  doi = {10.48550/arXiv.2404.19735}

Related Posts & Source Code

ParaGrapher Web Page

An Evaluation of Bandwidth of Different Storage Types (HDD vs. SSD vs. LustreFS) for Different Block Sizes and Different Read Methods (mmap vs pread vs read)

Short URL of this post:

We evaluate read bandwidth of three storage types:

  • HDD: A 6TB Hitachi HUS726060AL 7200RPM SATA v3.1
  • SSD: A 4TB Samsung MZQL23T8HCLS-00A07 PCIe4 NVMe v1.4
  • LustreFS: A parallel file system with total 2PB with a SSD pool

and for three read methods:

and for two block sizes:

  • 4 KB blocks
  • 4 MB blocks

The source code is available on ParaGrapher repository:

The OS cache of storage contents have been dropped after each evaluation
(sudo sh -c 'echo 3 >/proc/sys/vm/drop_caches').
The flushcache.c file ( can be used with the same functionality for users without sudo access, however, it usually takes more time to be finished.

For LustreFS, we have repeated the evaluation of read and pread using O_DIRECT flag as this flag prevents client-side caching.

For HDD and SSD experiments, we have used a machine with Intel W-2295 3.00GHz CPU, 18 cores, 36 hyper-threads, 24MB L3 cache, 256 GB DDR4 2933Mhz memory, running Debian 12 Linux 6.1. For LustreFS, we have used a machine with 2TB 3.2GHz DDR4 memory, 2 AMD 7702 CPUs, in total, 128 cores, 256 threads.

The results of the evaluation using read_bandwidth.c are in the following table. The values are Bandwidth in MB/s. Also, 1-2 digits close to each number with a white background are are percentage of load imbalance between parallel threads.

Please click on the image to expand.


MS-BioGraphs on IEEE DataPort

MS-BioGraph sequence similarity graph datasets are now publicly available on IEEE DataPort:

To access the files, you need to register/login to IEEE DataPort and then visit the MS-BioGraphs page. By saving the page as an HTML file such as dp.html, you may download the datasets (as an example MS1) using the following script:


urls=`cat $html_file  | sed  -e 's/\&/\&/g'  | grep -Eo "(http|https)://[a-zA-Z0-9./?&=_%:-]*" | grep amazonaws  | sort | uniq | grep -E "$dsname[-_\.]"`

for u in $urls; do
    wget $u
    if [ $? != 0 ]; then break; fi

# removing query strings
for f in $(find $1 -type f); do
    if [ $f = ${f%%\?*} ]; then continue; fi
    mv "${f}" "${f%%\?*}"

# liking offsets.bin to be found by ParaGrapher
ln -s ${dsname}_offsets.bin ${dsname}-underlying_offsets.bin

Instead of wget you may use axel -n 10 to use multiple connections (here, 10) for downloading each file (


Related Posts

ParaGrapher Integrated to LaganLighter

Poplar source code has been integrated to LaganLighter and access to different WebGraph formats are available in LaganLighter:


For further details, please refer to
– LaganLighter source coder Repository:, particularly, the graph.c file.
– ParaGrapher source code repository: particularly, the src/webgraph.c and src/WG*.java files.

Read more about ParaGrapher and LaganLighter.

Related Posts

ParaGrapher Source Code For WebGraph Types

ParaGrapher source code for accessing WebGraphs have been published. The supported graph types are:

ParaGrapher uses its asynchronous and parallel API to implement these graph types. The user needs to implement a callback function that is called by the API upon completion of reading a block of edges. Poplar uses a shared memory for interaction between its C library and the Java library that deploys the WebGraph framework.

For further details, please refer to Poplar source code repository:, particularly, src/webgraph.c and src/WG*.java files.


Related Posts

MS-BioGraphs Validation



We provide a Shell script,, and a Java program,, to verify the the correctness of the graphs. Each graph has a .ojson file whose shasum is verified by the value retreived from our server. Files such as offsets.bin, wcc.bin, n2o.bin, trans_offsets.bin, and edges_shas.txt have shasum records in the ojson file which is used for validation of these files.

The graph in WebGraph format has been compressed in MS??-underlying.* and MS??-weights.* files. In order to validate the compressed graph, the is used. It is a parallel Java code that uses the WebGraph library to traverse the graph and calculate the shasum of blocks of edges (endpoints and weights). Then, the calculated results are matched with the edges_shas.txt file of the graph.

It is also possible to validate some particular blocks by matching the calculated shasum with the relevant row in the edges_shas.txt file. This file has a format such as the following. Each block contains 64 Million consecutive edges. The start of each block is identified by a vertex ID and its edge index. The Column endpoint_sha is the shasum of the 64 Million endpoints when stored as an array of 4-Bytes elements in the binary format and in the little endian order. Similarly, Column weights_sha shows the shasum of weights (labels). We have separated weights from endpoints as in some applications weights are not needed and therefore it is not necessary to read and validate them.

64MB blk#;     vertex; edge index;                             endpoint_sha;                              weights_sha;
         0;          0;          0; 509784b158cb9404241afb21d0ceaf590b88d2f2; 57da4ad7bb89c5922e436b0535d791fa8f40dffd;
         1;    2315113;        705; fafc118563c1d7b5fbff64af56edd6a56524f479; 13b7a9ca60bfb0715d563218d0a1cd787b00a07c;
         2;    4521625;        597; 4ed65aa07c8062a151166ef2e9bdb93e41d19357; 8158276bec426ee46eca9912759eb9bd57fcc957;
         3;    6347361;        112; d02e8913c807c3f4ecde9c638e0ded5ab80ba819; 26bc3296de65cba6ac539cd96b79ae6f7a4d37be;
         4;    8447869;         15; 61513c84db40124496cdf769516118b63598914f; 781b9f4372ac614e94d097017c756d015234deb6; 


  • JDK with version > 15
  • jq
  • wget

WebGraph Framework

Please visit .

ParaGrapher Graph Loading API and Library

The WebGraph formats can also be read using the ParaGrapher library:


Licensed under the GNU v3 General Public License, as published by the Free Software Foundation. You must not use this Software except in compliance with the terms of the License. Unless required by applicable law or agreed upon in writing, this Software is distributed on an “as is” basis, without any warranty; without even the implied warranty of merchantability or fitness for a particular purpose, neither express nor implied.

Copyright 2022-2023 The Queen’s University of Belfast, Northern Ireland, UK


Related Posts

LaganLighter Source Code


Algorithms in This Repo


git clone --recursive

Graph Types

LaganLighter supports the following graph formats:

  • CSR/CSC graph in text format, for testing. This format has 4 lines: (i) number of vertices (|V|), (ii) number of edges (|E|), (iii) |V| space-separated numbers showing offsets of the vertices, and (iv) |E| space-separated numbers indicating edges.
  • CSR/CSC WebGraph format: supported by the Poplar Graph Loading Library
    external git repository


In addition to execution time, we use the PAPI library to measure hardware counters such as L3 cache misses, hardware instructions, DTLB misses, and load and store memory instructions. ( papi_(init/start/reset/stop) and (print/reset)_hw_events functions defined in omp.c).

To measure load balance, we measure the total time of executing a loop and the time each thread spends in this loop (mt and ttimes in the following sample code). Using these values, PTIP macro (defined in omp.c) calculates the percentage of average idle time (as an indicator of load imbalance) and prints it with the total time (mt).

mt = - get_nano_time()
#pragma omp parallel  
   unsigned tid = omp_get_thread_num();
   ttimes[tid] = - get_nano_time();
   #pragma omp for nowait
   for(unsigned int v = 0; v < g->vertices_count; v++)
      // .....
   ttimes[tid] += get_nano_time();
mt += get_nano_time();
PTIP("Step ... ");

As an example, the following execution of Thrifty, shows that the “Zero Planting” step has been performed in 8.98 milliseconds and with a 8.22% load imbalance, while processors have been idle for 72.22% of the execution time, on average, in the “Initial Push” step.

NUMA-Aware and Locality-Preserving Partitioning and Scheduling

In order to assign consecutive partitions (vertices and/or their edges) to each parallel processor, we initially divide partitions and assign a number of consecutive partitions to each thread. Then, we specify the order of victim threads in the work-stealing process. During the initialization of LaganLighter parallel processing environment (in initialize_omp_par_env() function defined in file omp.c), for each thread, we create a list of threads as consequent victims of stealing.

A thread, first, steals jobs (i.e., partitions) from consequent threads in the same NUMA node and then from the threads in consequent NUMA nodes. As an example, the following image shows the stealing order of a 24-core machine with 2 NUMA nodes. This shows that thread 1 steals from threads 2, 3, …,11, and ,0 running on the same NUMA socket and then from threads 13, 14, …, 23, and 12 running on the next NUMA socket.

We use dynamic_partitioning_...() functions (in file partitioning.c) to process partitions by threads in the specified order. A sample code is in the following:

struct dynamic_partitioning* dp = dynamic_partitioning_initialize(pe, partitions_count);

#pragma omp parallel  
   unsigned int tid = omp_get_thread_num();
   unsigned int partition = -1U;		

      partition = dynamic_partitioning_get_next_partition(dp, tid, partition);
      if(partition == -1U)

      for(v = start_vertex[partition]; v < start_vertex[partition + 1]; v++)
	// ....


Bugs & Support

As “we write bugs that in particular cases have been tested to work correctly”, we try to evaluate and validate the algorithms and their implementations. If you receive wrong results or you are suspicious about parts of the code, please contact us or submit an issue.


Licensed under the GNU v3 General Public License, as published by the Free Software Foundation. You must not use this Software except in compliance with the terms of the License. Unless required by applicable law or agreed upon in writing, this Software is distributed on an “as is” basis, without any warranty; without even the implied warranty of merchantability or fitness for a particular purpose, neither express nor implied. For details see terms of the License.

Copyright 2022 The Queen’s University of Belfast, Northern Ireland, UK


Related Posts

Graptor Sources Published

Finally got around to this: publishing the Graptor source code. With time passing, the code has changed quite a bit compared to that used in the paper: Graptor: efficient pull and push style vectorized graph processing. The evolution of the code has advantages: it’s faster. There are also disadvantages: not all versions and variations of the code that were experimented with can still be compiled.

The source code can be found here:

There will likely be issues (errors, lack of documentation, …) as this is experimental research code. Drop me a line if you need a hand h {a dot} vandierendonck {an at} qub {another dot} ac {the last dot} uk .