[ 
https://issues.apache.org/jira/browse/YARN-3856?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14607945#comment-14607945
 ] 

jaehoon ko commented on YARN-3856:
----------------------------------

Hi, Ravi. Thanks for your comment.

I wasn't aware of YARN-18 and related discussion at the time of submission. As 
far as I understand, it is about adding another layer of locality - node group 
- between host and rack for clustesr deployed on VMs or containers. It tries to 
make arbitrary network topology pluggable for such needs. 
My patch doesn't consider an arbitrary network topology. However, according to 
the last comment of YARN-18, YARN community thinks that a hierarchical network 
is enough. In that case, I think my patch provides just-enough functionality; 
choosing the closest host based on network hierarchy.

> YARN shoud allocate container that is closest to the data
> ---------------------------------------------------------
>
>                 Key: YARN-3856
>                 URL: https://issues.apache.org/jira/browse/YARN-3856
>             Project: Hadoop YARN
>          Issue Type: Improvement
>          Components: resourcemanager
>    Affects Versions: 2.7.0
>         Environment: Hadoop cluster with multi-level network hierarchy
>            Reporter: jaehoon ko
>         Attachments: YARN-3856.001.patch, YARN-3856.002.patch
>
>
> Currently, given a Container request for a host, ResourceManager allocates a 
> Container with following priorities (RMContainerAllocator.java):
>  - Requested host
>  - a host in the same rack as the requested host
>  - any host
> This can lead to a sub-optimal allocation if Hadoop cluster is deployed on 
> multi-level networked hosts (which is typical). For example, let's suppose a 
> network architecture with one core switches, two aggregate switches, four ToR 
> switches, and 8 hosts. Each switch has two downlinks. Rack IDs of hosts are 
> as follows:
> h1, h2: /c/a1/t1
> h3, h4: /c/a1/t2
> h5, h6: /c/a2/t3
> h7, h8: /c/a2/t4
> To allocate a container for data in h1, Hadoop first tries h1 itself, then 
> h2, then any of h3 ~ h8. Clearly, h3 or h4 are better than h5~h8 in terms of 
> network distance and bandwidth. However, current implementation choose one 
> from h3~h8 with equal probabilities.
> This limitation is more obvious when considering hadoop clusters deployed on 
> VM or containers. In this case, only the VMs or containers running in the 
> same physical host are considered rack local, and actual rack-local hosts are 
> chosen with same probabilities as far hosts.
> The root cause of this limitation is that RMContainerAllocator.java performs 
> exact matching on rack id to find a rack local host. Alternatively, we can 
> perform longest-prefix matching to find a closest host. Using the same 
> network architecture as above, with longest-prefix matching, hosts are 
> selected with the following priorities:
>  h1
>  h2
>  h3 or h4
>  h5 or h6 or h7 or h8



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to