论文阅读 - Optimizing Graph Algorithms on Pregel-like
博客专区 > wdfnst 的博客 > 博客详情
论文阅读 - Optimizing Graph Algorithms on Pregel-like
wdfnst 发表于2年前
论文阅读 - Optimizing Graph Algorithms on Pregel-like
  • 发表于 2年前
  • 阅读 2
  • 收藏 0
  • 点赞 2
  • 评论 0
摘要: SuperVertex
  • Highlight, page 1
    Standard graph algorithms in this setting can incur unnecessary in- efficiencies such as slow convergence or high communication or computation cost

  • Highlight, page 1
    Standard graph algorithms in this setting can incur unnecessary in- efficiencies such as slow convergence or high communication or computation cost

  • Highlight, page 1
    large diameters or skew in component sizes.

  • Highlight, page 1
    performing some serial computation on a tiny fraction of the in- put graph

  • Underline, page 1
    complementing

  • Highlight, page 1
    our open-source Pregel implementation

  • Highlight, page 2
    FCS monitors the size of the “active” graph on which the computation is executing. If the active graph becomes small enough, FCS sends it to the master, which performs the end of the computation serially inside master.compute(), then sends the results back to the workers.

  • Highlight, page 2
    sets of vertices (called subvertices) are merged to form supervertices.

  • Highlight, page 6
    FCS monitors the size the active- subgraph. Once the size of the active-subgraph is below a threshold (5M edges by default), it sends the active-subgraph to the mas- ter, which performs the rest of the computation serially, and sends the results back to the workers.

  • Highlight, page 8
    Our SEAS optimiza- tion instead stores the edges of a supervertex s in a distributed fash- ion among all of its subvertices.

  • Highlight, page 8
    In our implementation of SEAS, subvertices store a pointer to their latest supervertices.

共有 人打赏支持
粉丝 3
博文 18
码字总数 22859
作品 1
×
wdfnst
如果觉得我的文章对您有用,请随意打赏。您的支持将鼓励我继续创作!
* 金额(元)
¥1 ¥5 ¥10 ¥20 其他金额
打赏人
留言
* 支付类型
微信扫码支付
打赏金额:
已支付成功
打赏金额: