hadoop：四核机器上映射/减少作业的最佳数量
我试图通过谷歌找到答案，但没有找到好的参考资料，
- I have a Quad-core Ubuntu box running a map-reduce job.
- running default 2 maps and taking lot of time
- what be a good number of map/reduce job for a machine of such config?
请提出建议
TIA
I tried to find out over the google but no good reference that I found
- I have a Quad-core Ubuntu box running a map-reduce job.
- running default 2 maps and taking lot of time
- what be a good number of map/reduce job for a machine of such config?
Please advice
TIA
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
看起来像是一个经验问题。从 1 开始，然后逐步提高一段时间。绘制一条曲线，看看它的结果如何。我希望您可能会在可用核心数量或可用硬件线程数量（您的机器上可能相同）中找到最佳点。然而，进行实际实验是找出答案的最佳方法。
Seems like an empirical problem. Start at 1 and work your way up for awhile. Plot a curve, see how it ends up. I'd expect you'll probably find the sweet spot at either the number of cores available or the number of hardware threads available (may be the same on your box). However, doing the actual experiment is the best way to find out.