D3 Center, The University of Osaka » Blog Archive » SQUID: Cloud Bursting Using a Public Cloud (Microsoft Azure) (From January 16)

SQUID: Cloud Bursting Using a Public Cloud (Microsoft Azure) (From January 16)

As the usage of SQUID general-purpose CPU nodes has recently increased, job waiting times have become longer. To address this situation, starting on January 16, 2026, we will provide a cloud bursting feature using Microsoft Azure for jobs that use one node and request a maximum execution time of one hour or less.
With this feature, when SQUID compute nodes are congested, eligible jobs may be automatically transferred to and executed on Microsoft Azure compute nodes, which may reduce job waiting times. Please review the notes below and make use of the cloud bursting feature accordingly.

Service Start Date

January 16, 2026, 10:00 a.m. (JST)

Eligibility

Jobs submitted to the SQUID general-purpose CPU nodes that use "one node" and have a requested runtime of "one hour or less" may be executed on Microsoft Azure compute nodes.

Point Consumption

Even when a job is executed on Microsoft Azure compute nodes, the points consumed are equivalent to those for the SQUID general-purpose CPU nodes. No additional charges will be incurred, so you can use this feature with confidence. Further details are provided below.

How to Use Cloud Bursting

To use Microsoft Azure compute nodes, add the required options to your job submission shell script (job script file) and submit it to the scheduler. An example job script is shown below.

#!/bin/bash

#------- qsub option -----------

#PBS -q SQUID

# Cloud bursting is available only for the SQUID queue.

#PBS --group=G01234 # Your group name

#PBS -l cpunum_job=76 # Requested number of CPU cores

#PBS -l elapstim_req=01:00:00 # Requested maximum elapsed time (example: 1 hour)

#PBS --enable-cloud-bursting=yes # Allow cloud bursting

#PBS -U cloud_wait_limit=04:00:00 # If the waiting time exceeds this limit, the job may be executed on the cloud (example: 4 hours)

#------- Program execution -----------

module load BaseCPU # Load the base environment

cd $PBS_O_WORKDIR # Move to the directory where qsub was executed

./a.out > result.txt # Execute the program

Notes

1. The hardware configuration of Microsoft Azure compute nodes differs from that of SQUID general-purpose CPU nodes, which may result in different computation times and results.

・SQUID general-purpose CPU node:
Processor: Intel Xeon Platinum 8368 (Ice Lake / 2.40 GHz, 38 cores) × 2
Memory: 256 GB

・Microsoft Azure compute node:
Processor: Intel Xeon Platinum 8280 (Cascade Lake / 2.70 GHz, 24 cores) × 4
Memory: 2048 GB
Note: The actual available resources are limited to 76 CPU cores and 248 GB of memory (equivalent to SQUID general-purpose CPU nodes).

Because the NUMA configuration of Microsoft Azure compute nodes differs from that of SQUID general-purpose CPU nodes, specifying core allocation in a job script based on an on-premises environment may cause execution errors or performance degradation when bursting.

2. Even when cloud bursting is enabled and Microsoft Azure compute resources are used, point consumption is equivalent to that of SQUID general-purpose CPU nodes.

That is, points are calculated as follows:
Node execution time × General-purpose CPU node consumption coefficient (0.2998) × General-purpose CPU node fuel coefficient × General-purpose CPU node seasonal coefficient

3. If cloud bursting is enabled, you are deemed to have understood and accepted the following:

　・Your job may be executed on Microsoft Azure compute nodes.
　・Depending on resource availability, Microsoft Azure resources may not be used and your job may run only on SQUID general-purpose CPU nodes.

4. Microsoft Azure compute nodes access the SQUID file system over the network. I/O performance is lower than that of SQUID general-purpose CPU nodes.

Therefore, programs with heavy I/O operations may experience longer execution times.

We hope you will make use of the cloud bursting feature after reviewing the information above.

Posted : January 15,2026