Parallel algorithms for minimum spanning trees

In graph theory a minimum spanning tree (MST) of a graph with and is a tree subgraph of that contains all of its vertices and is of minimum weight.

MSTs are useful and versatile tools utilised in a wide variety of practical and theoretical fields. For example, a company looking to supply multiple stores with a certain product from a single warehouse might use an MST originating at the warehouse to calculate the shortest paths to each company store. In this case the stores and the warehouse are represented as vertices and the road connections between them - as edges. Each edge is labelled with the length of the corresponding road connection.

If is edge-unweighted every spanning tree possesses the same number of edges and thus the same weight. In the edge-weighted case, the spanning tree, the sum of the weights of the edges of which is lowest among all spanning trees of , is called a minimum spanning tree (MST). It is not necessarily unique. More generally, graphs that are not necessarily connected have minimum spanning forests, which consist of a union of MSTs for each connected component.

As finding MSTs is a widespread problem in graph theory, there exist many sequential algorithms for solving it. Among them are Prim's, Kruskal's and Borůvka's algorithms, each utilising different properties of MSTs. They all operate in a similar fashion - a subset of is iteratively grown until a valid MST has been discovered. However, as practical problems are often quite large (road networks sometimes have billions of edges), performance is a key factor. One option of improving it is by parallelising known MST algorithms.[1]

Prim's algorithm edit

This algorithm utilises the cut-property of MSTs. A simple high-level pseudocode implementation is provided below:

 
  where   is a random vertex in  
repeat   times
    find lightest edge   s.t.   but  
     
     
return T

Each edge is observed exactly twice - namely when examining each of its endpoints. Each vertex is examined exactly once for a total of   operations aside from the selection of the lightest edge at each loop iteration. This selection is often performed using a priority queue (PQ). For each edge at most one decreaseKey operation (amortised in  ) is performed and each loop iteration performs one deleteMin operation ( ). Thus using Fibonacci heaps the total runtime of Prim's algorithm is asymptotically in  .

It is important to note that the loop is inherently sequential and can not be properly parallelised. This is the case, since the lightest edge with one endpoint in   and on in   might change with the addition of edges to  . Thus no two selections of a lightest edge can be performed at the same time. However, there do exist some attempts at parallelisation.

One possible idea is to use   processors to support PQ access in   on an EREW-PRAM machine,[2] thus lowering the total runtime to  .

Kruskal's algorithm edit

Kruskal's MST algorithm utilises the cycle property of MSTs. A high-level pseudocode representation is provided below.

  forest with every vertex in its own subtree
foreach   in ascending order of weight
    if   and   in different subtrees of  
         
return T

The subtrees of   are stored in union-find data structures, which is why checking whether or not two vertices are in the same subtree is possible in amortised   where   is the inverse Ackermann function. Thus the total runtime of the algorithm is in  . Here   denotes the single-valued inverse Ackermann function, for which any realistic input yields an integer less than five.

Approach 1: Parallelising the sorting step edit

Similarly to Prim's algorithm there are components in Kruskal's approach that can not be parallelised in its classical variant. For example, determining whether or not two vertices are in the same subtree is difficult to parallelise, as two union operations might attempt to join the same subtrees at the same time. Really the only opportunity for parallelisation lies in the sorting step. As sorting is linear in the optimal case on   processors, the total runtime can be reduced to  .

Approach 2: Filter-Kruskal edit

Another approach would be to modify the original algorithm by growing   more aggressively. This idea was presented by Osipov et al.[3][4] The basic idea behind Filter-Kruskal is to partition the edges in a similar way to quicksort and filter out edges that connect vertices that belong to the same tree in order to reduce the cost of sorting. A high-level pseudocode representation is provided below.

filterKruskal( ):
if   KruskalThreshold:
    return kruskal( )
pivot = chooseRandom( )
 ,  partition( , pivot)
  filterKruskal( )
  filter( )
    filterKruskal( )
return  

partition( , pivot):
  
 
foreach  :
    if weight( )   pivot:
          
    else
         
return ( ,  )

filter( ):
 
foreach  :
    if find-set(u)   find-set(v):
         
return  

Filter-Kruskal is better suited for parallelisation, since sorting, partitioning and filtering have intuitively easy parallelisations where the edges are simply divided between the cores.

Borůvka's algorithm edit

The main idea behind Borůvka's algorithm is edge contraction. An edge   is contracted by first removing   from the graph and then redirecting every edge   to  . These new edges retain their old edge weights. If the goal is not just to determine the weight of an MST but also which edges it comprises, it must be noted between which pairs of vertices an edge was contracted. A high-level pseudocode representation is presented below.

 
while  
      
    for  
            lightest  
    for  
        contract  
     
return T

It is possible that contractions lead to multiple edges between a pair of vertices. The intuitive way of choosing the lightest of them is not possible in  . However, if all contractions that share a vertex are performed in parallel this is doable. The recursion stops when there is only a single vertex remaining, which means the algorithm needs at most   iterations, leading to a total runtime in  .

Parallelisation edit

One possible parallelisation of this algorithm[5][6][7] yields a polylogarithmic time complexity, i.e.   and there exists a constant   so that  . Here   denotes the runtime for a graph with   edges,   vertices on a machine with   processors. The basic idea is the following:

while  
    find lightest incident edges //  
    assign the corresponding subgraph to each vertex //  
    contract each subgraph //  

The MST then consists of all the found lightest edges.

This parallelisation utilises the adjacency array graph representation for  . This consists of three arrays -   of length   for the vertices,   of length   for the endpoints of each of the   edges and   of length   for the edges' weights. Now for vertex   the other end of each edge incident to   can be found in the entries between   and  . The weight of the  -th edge in   can be found in  . Then the  -th edge in   is between vertices   and   if and only if   and  .

Finding the lightest incident edge edit

First the edges are distributed between each of the   processors. The  -th processor receives the edges stored between   and  . Furthermore, each processor needs to know to which vertex these edges belong (since   only stores one of the edge's endpoints) and stores this in the array  . Obtaining this information is possible in   using   binary searches or in   using a linear search. In practice the latter approach is sometimes quicker, even though it is asymptotically worse.

Now each processor determines the lightest edge incident to each of its vertices.

  find( ,  )
for  
    if  
         
    if 
         

Here the issue arises some vertices are handled by more than one processor. A possible solution to this is that every processor has its own   array which is later combined with those of the others using a reduction. Each processor has at most two vertices that are also handled by other processors and each reduction is in  . Thus the total runtime of this step is in  .

Assigning subgraphs to vertices edit

Observe the graph that consists solely of edges collected in the previous step. These edges are directed away from the vertex to which they are the lightest incident edge. The resulting graph decomposes into multiple weakly connected components. The goal of this step is to assign to each vertex the component of which it is a part. Note that every vertex has exactly one outgoing edge and therefore each component is a pseudotree - a tree with a single extra edge that runs in parallel to the lightest edge in the component but in the opposite direction. The following code mutates this extra edge into a loop:

parallel forAll   
     
    if   
         

Now every weakly connected component is a directed tree where the root has a loop. This root is chosen as the representative of each component. The following code uses doubling to assign each vertex its representative:

while  
    forAll   
         

Now every subgraph is a star. With some advanced techniques this step needs   time.

Contracting the subgraphs edit

In this step each subgraph is contracted to a single vertex.

  number of subgraphs
 
find a bijective function   star root   
 

Finding the bijective function is possible in   using a prefix sum. As we now have a new set of vertices and edges the adjacency array must be rebuilt, which can be done using Integersort on   in   time.

Complexity edit

Each iteration now needs   time and just like in the sequential case there are   iterations, resulting in a total runtime of  . If   the efficiency of the algorithm is in   and it is relatively efficient. If   then it is absolutely efficient.

Further algorithms edit

There are multiple other parallel algorithms that deal the issue of finding an MST. With a linear number of processors it is possible to achieve this in  .[8][9] Bader and Cong presented an MST-algorithm, that was five times quicker on eight cores than an optimal sequential algorithm.[10]

Another challenge is the External Memory model - there is a proposed algorithm due to Dementiev et al. that is claimed to be only two to five times slower than an algorithm that only makes use of internal memory[11]

References edit

  1. ^ Sanders; Dietzfelbinger; Martin; Mehlhorn; Kurt; Peter (2014-06-10). Algorithmen und Datenstrukturen Die Grundwerkzeuge. Springer Vieweg. ISBN 978-3-642-05472-3.
  2. ^ Brodal, Gerth Stølting; Träff, Jesper Larsson; Zaroliagis, Christos D. (1998), "A Parallel Priority Queue with Constant Time Operations", Journal of Parallel and Distributed Computing, 49 (1): 4–21, CiteSeerX 10.1.1.48.3272, doi:10.1006/jpdc.1998.1425
  3. ^ Osipov, Vitaly; Sanders, Peter; Singler, Johannes (2009), "The filter-kruskal minimum spanning tree algorithm", Proceedings of the Eleventh Workshop on Algorithm Engineering and Experiments (ALENEX). Society for Industrial and Applied Mathematics: 52–61, CiteSeerX 10.1.1.218.2574
  4. ^ Sanders, Peter. "Algorithm Engineering script" (PDF). Algorithm Engineering KIT Homepage. Retrieved 25 February 2019.
  5. ^ Sanders, Peter. "Parallel Algorithms script" (PDF). Parallel Algorithms KIT Homepage. Retrieved 25 February 2019.
  6. ^ Zadeh, Reza. "Distributed Algorithms and Optimization" (PDF). Distributed Algorithms and Optimization Stanford University Homepage. Retrieved 25 February 2019.
  7. ^ Chun, Sun; Condon, Anne (1996). "Parallel implementation of Bouvka's minimum spanning tree algorithm". Proceedings of International Conference on Parallel Processing. pp. 302–308. doi:10.1109/IPPS.1996.508073. ISBN 0-8186-7255-2. S2CID 12710022.
  8. ^ Chong, Ka Wong; Han, Yijie; Lam, Tak Wah (2001), "Concurrent threads and optimal parallel minimum spanning trees algorithm", Journal of the Association for Computing Machinery, 48 (2): 297–323, CiteSeerX 10.1.1.32.1554, doi:10.1145/375827.375847, MR 1868718, S2CID 1778676
  9. ^ Pettie, Seth; Ramachandran, Vijaya (2002), "A randomized time-work optimal parallel algorithm for finding a minimum spanning forest" (PDF), SIAM Journal on Computing, 31 (6): 1879–1895, doi:10.1137/S0097539700371065, MR 1954882
  10. ^ Bader, David A.; Cong, Guojing (2006), "Fast shared-memory algorithms for computing the minimum spanning forest of sparse graphs", Journal of Parallel and Distributed Computing, 66 (11): 1366–1378, doi:10.1016/j.jpdc.2006.06.001
  11. ^ Dementiev, Roman; Sanders, Peter; Schultes, Dominik; Sibeyn, Jop F. (2004), "Engineering an external memory minimum spanning tree algorithm", Proc. IFIP 18th World Computer Congress, TC1 3rd International Conference on Theoretical Computer Science (TCS2004) (PDF), pp. 195–208.