Support Questions
Find answers, ask questions, and share your expertise
Announcements
Alert: Welcome to the Unified Cloudera Community. Former HCC members be sure to read and learn how to activate your account here.

HDP on Isilon - CPU Usage

Solved Go to solution
Highlighted

HDP on Isilon - CPU Usage

Rising Star

We need to set up a HDP cluster based on Isilon storage and customer is asking how much impact would it have on the CPU usage on Isilon nodes ? Currently the Isilon cluster is shared with other work loads as well. What are our experiences around this ? Would the namenode operations in Isilon cause lot of CPU spikes which results in degradation in performance with other workloads on Isilon ?

1 ACCEPTED SOLUTION

Accepted Solutions

Re: HDP on Isilon - CPU Usage

@hrongali

Isilon is designed to handle mult protocol workload. The namenode is installed by default. You would have to asses how much of CPU is being used today and what workload you are bringing on.

Make sure it is X410 model of the Isilon that they are using. You can use other models, but it is for experimental purpose only. At the worst case they may need to add more Isilon nodes to the architecture incase there is more workload. Get an Isilon SE involved in the Isilon sizing as they have incentive to sell more hardware. A good use case would be to run some TPC DS work load with Hive Tez and see how it behaves

Also make sure that the compute nodes are co-located with the Isilon nodes, as the compute and storage will be pretty chatty.

3 REPLIES 3

Re: HDP on Isilon - CPU Usage

@hrongali

Isilon is designed to handle mult protocol workload. The namenode is installed by default. You would have to asses how much of CPU is being used today and what workload you are bringing on.

Make sure it is X410 model of the Isilon that they are using. You can use other models, but it is for experimental purpose only. At the worst case they may need to add more Isilon nodes to the architecture incase there is more workload. Get an Isilon SE involved in the Isilon sizing as they have incentive to sell more hardware. A good use case would be to run some TPC DS work load with Hive Tez and see how it behaves

Also make sure that the compute nodes are co-located with the Isilon nodes, as the compute and storage will be pretty chatty.

Re: HDP on Isilon - CPU Usage

New Contributor

Isilon is storage only, and since it uses it's own system for handling the files (OneFS) it's likely to have very little impact on other workloads on the Isilon cluster. This of course, is assuming that you have already accounted for the additional files that you are storing on your Isilon cluster and are not attempting to add files that would max out the storage on the cluster as well.

Re: HDP on Isilon - CPU Usage

New Contributor

@hrongali

As a rule of thumb CPU usage on Isilon is dictated by the write workload. This is mainly to do the FEC and striping calculations. There are of course exceptions to this rule but it is a good rule to use in these environments. The namenode service is relatively light weight and never gets in the way. The majority of the network traffic is datanode traffic between the compute nodes and Isilon.