## GoogleCloudPlatform/PerfKitBenchmarker — v1.13.0…v1.14.0

_1460+ commits._

### Features
- **Add netperf UDP stream test (#2075)** (07441b9)
- **Update Azure Virtual Network to support location level Virtual Networks. With this update, One Azure Virtual Network will be created per Azure Location. Each Azure Availability Zone will result in a Azure subnet that is a part of the respective virtual network.** (7171ace)
- **Add --nccl_net_plugin flag** (d804dda)
- **Add ability to load datasets into BigQuery from csv files on GCS.** (bb0464f)
- **Added Low Priority support for Azure VMs** (c6e2171)
- **Add --tcpdump to PKB to collect tcpdumps during runs.** (41dbbc9)
- **Add support for generating CLI commands to object_storage service.** (aa1f667)
- **Added placement group flag support for Azure.** (85cbe0a)
- **add custom subnet support (#1986)** (84f4e47)
- **Support for un-managed data processing yarn cluster benchmarking.** (547b7f6)
- **Refactor OpenFOAM flags and add validation for dimensions.** (eaeea22)
- **Adds support for Google Cloud Firestore YCSB bencharks.** (981c8f1)
- **Add stress-ng support to run on different stress-ng versions** (6a91630)
- **Add a flag that replaces the YCSB jar url** (a2e48f4)
- **Add an AWS provider to provision a DAX cluster** (09213b1)
- **Add mpirun mapping flag to OpenFOAM.** (1c717aa)
- **Add ability to extract a dataset from BigQuery to GCS.** (e2ae314)
- **Add flag --aws_ena to use AWS Elastic Fiber Adapter.** (9df9b17)
- **Implement num_vms_per_host support for AWS.** (750bc39)
- **Add support to perform speccpu2017 on multiple vms. This is necessary to test** (1d3b38d)
- **Add generic dimension resizing logic to OpenFOAM.** (ae15a7f)
- **Add AWS T3 cpu credits argument to sample metadata.** (1878270)
- **Implement num_vms_per_host support for Azure.** (561ebbc)
- **Add Amd Blis support for HPCC.** (0551d96)
- **update README to new azure cli instructions** (c8ce91b)
- **Add support for running OpenFOAM in parallel on multiple VMs.** (228c3c8)
- **Add ability to parse and suppress IssueCommand failures.** (5b22e65)
- **Add support for running OpenFOAM in parallel on multiple VMs.** (4e5f33e)
- **Update AliCloud support and documentation** (5a16bc8)
- **Add NVIDIA's certificate before installing cuDNN package** (15b3a24)
- **Add flag --cluster_boot_test_rdp_port_listening to measure time until RDP port accepts a connection in Windows.** (9d35976)
- **Add support for unmanaged NFS.** (d2c25b2)
- **Add support for reporting raw latency with wrk2.** (443ea17)

### Fixes
- **Pulling future changes forward to fix failing tests.** (fc29548)
- **Fix OpenFOAM required_flags warnings.** (3181e71)
- **Linter fix.** (0a5fda6)
- **Fix lint issue.** (7b2843c)
- **Fix comment spacing linter issue.** (d8e5d74)
- **Fix openstack subnet creation issue** (5affa74)
- **Fix wrong gsutil path in mlperf benchmark.** (4e6440c)
- **Fix linter issue.** (c3c6a34)
- **Fix OpenFOAM worker node SSH authentication.** (2c8bea0)
- **Add `-m PEM` to `ssh-keygen` to fix AWS Windows VMs.** (a31bd50)
- **Tox spacing fix** (9d7a8c9)
- **Fix Beam integration test broken by** (f02d630)
- **Fix minor bug in Athena edw creation. Bucket name needs to be guaranteed unique, which is best done with a prefix like 'pkb'.** (86e8fa4)
- **Fix AWS user name resolution.** (64ab9c7)
- **Fix leaking username in 3rd party code.** (eac69c2)
- **Fix two broken Azure Windows Server Core Image URNs. In the case of Windows Server 2012 Server Core there did not seem to be an available image, so it was left the same as Server Base.** (bb00158)
- **Fix LMBench installation.** (e507644)
- **Restore formatting after "auto-fix"** (dcf06dd)
- **linter fix for vm_util** (5088be8)
- **Cassandra branch name fix** (b834f4a)
- **Fix linter.** (953844a)
- **Fix linter issue.** (8ff72cb)
- **Fix check for availability zone before determining whether or not to create an** (e2615ec)

### Backend
- **Default Docker image for cluster_boot on Kubernetes (#2061)** (a0c9c82)
- **Enable GCE Disk NFS type.** (deb94e3)
- **Delete existing libnccl before installing NCCL** (a82d419)
- **Fixes to MySQL server and client settings to ensure enough data and temp directory space, and to prevent connection timeouts. This way, sysbench will have enough space to create the large tables and won't lose connection doing so.** (67fbd3e)
- **Update NCCL Version to 2.5.6-2** (a5a9baf)
- **Update benchmark_spec.py** (2611b3d)
- **Update benchmark_spec.py** (a8b408f)
- **Update benchmark_spec.py** (ecfc416)
- **More merge conflicts.** (04c3eb0)
- **Fixes for unit testing and python linter.** (e6fb80f)
- **Update MLPerf GPU training test to run MLPerf 0.6** (91c86be)
- **Adds NCCL test to guitar.** (bce1a43)
- **Usability improvement for bq commands.** (92b2b1d)
- **Placement groups are used by Aws and Azure. Changed the AWS specific flag to a** (d95dc77)
- **Remove tables from TPC table list that were in there by mistake.** (7c0194d)
- **Moe linter fixes** (bedc9a5)
- **Misc fixes to NCCL benchmark** (42524e3)
- **Use yaml.safe_load().** (aa45e60)
- **Use PyYAML 5.1** (d5e0487)
- **Update pyyaml version.** (6556b66)
- **Enable OpenFOAM to run a list of dimensions as input.** (19bcced)
- **Setting internal ip to pod ip. (#2045)** (6b7e42e)
- **Fixes the dangling -x mpirun option when env is empty** (c1a1d8f)
- **Surface Azure OverconstrainedZonalAllocationRequest error as Capacity Error.** (6e6334d)
- **Name Azure network based on Availability Zone.** (b44b7a3)
- **Azure skip availability set creation (#2054)** (1d70db9)
- **-------------** (04f00fd)
- **Merge pull request #2060 from GoogleCloudPlatform/moe_writing_branch_from_1576265226** (958480a)
- **Update tensorflow_client to be compatible with tensorflow 2.0.** (b4c8faf)
- **Include any timestamp where YCSB reported results in an aggregated YCSB timeseries (instead of dropping them as was previously done)** (42f2a22)
- **Only edit keys in one key file if they appear in multiple in AWS dynamo db test** (5079258)
- **Updated NCCL package installer to take in CUDA related parameters and to work** (124167c)
- **Authenticate OpenFOAM vms in parallel.** (58b89c5)
- **Updated AWS Virtual Machine to handle shutting_down case when creating an** (7d54b5e)
- **Change Azure VM creation timeout from 10 min to 20 min.** (2ca19f6)
- **Move some enhancement release notes that I accidentally listed as bugs.** (565b671)
- **Adds NCCL benchmark for GPU networking.** (9fd5c18)
- **-------------** (ee4a864)
- **Provision Athena databases with parquet data instead of csv, and empty the database before running.** (5fea2cc)
- **Increase timeout for _DeleteResource to 1 hour.** (c26bbac)
- **Reorder the scripts run for provisioning the bucketed tables in Athena. The temp table needs to be created before the bucketed one.** (1be71d0)
- **Set the number of VMs to use in the OpenFOAM benchmark via the --num_vms flag.** (c8562cb)
- **Update OpenFOAM metadata labels.** (bded197)
- **Updating the list of Athena tables to provision.** (d157baf)
- **Adding a functionality to distcp cross cloud from Dataproc.** (bfb5b51)
- **Update mlperf to v0.6, using NVIDIA's submission on GPUs and Google's submission on TPUs;** (602de81)
- **Merge pull request #2046 from GoogleCloudPlatform/moe_writing_branch_from_2500081906769990576** (57a463e)
- **Fixes a "E701 multiple statements on one line (colon)" linter error.** (0a70c0b)
- **Merge pull request #2043 from GoogleCloudPlatform/moe_writing_branch_from_1442447461804000145** (2784642)
- **Refactor dpb_service.SubmitJob to not require arguments we routinely set to None.** (f8d770b)
- **Adds the ability to extract the credentials for the locally configured AWS account.** (c3f9c01)
- **Adding more ycsb workloads** (1aa598a)
- **Updated ycsb.py to handle the cases of read only and update only, where there is no result for certain hdr histogram group.** (ee831dc)
- **Raise error if host capacity is reached with user specified --num_vms_per_host** (9e40459)
- **Merge pull request #2040 from GoogleCloudPlatform/moe_ruwa** (6c9050b)
- **Remove RDP setup from Windows Startup script as it is not needed by** (12e4cd5)
- **AWS Athena provisioning functionality.** (e9701b5)
- **Moving the shared FLAGs to the cloud specific flag module.** (7952a0c)
- **Updated to surface any 'quota limit' errors in Azure as Quota Exceeded.** (987115b)
- **Properly handle when IssueCommand failures are not suppressed.** (c327d97)
- **-------------** (5228f25)
- **update to more generic azure cli command** (4a0c61c)
- **Set preemptied vm's failed substatus as insufficient capacity.** (91bf41c)
- **Be consistent with Windows Server versions.** (d110031)
- **changes order** (631ea48)
- **Update Alicloud provider, cassandra version to 2.1** (3d7b28b)
- **update resnet to work on cpu and gpus using the offical model, and tpu using the tpu model** (58031b5)
- **Updated CHANGES.next.md** (81bbf7b)
- **Make AliCloud startup script template more readable** (a3bd106)
- **Use six to make sure b64encode result is a string** (dce23be)
- **Remove paramiko import** (93ab924)
- **Update CHANGES.next.md with Paramiko** (009a8b1)
- **Remove Paramiko dependency and rely on ssh key pairs** (b84a33b)
- **Keep python3 compatibility** (5ebd4fb)
- **Move AliCloud to Breaking Changes** (b55b9a8)
- **Added --cluster_boot_test_port_listening for cluster_boot benchmark. When** (13d0dbd)
- **Extending command timeout for create and delete on cloud redis.** (01800c8)
- **Change travis to py37 and PyYAML to 3.13** (2c574ed)
- **Remove the remote file before pushing a local file to the remote place. This can avoid permission issue as the pushed file only allows user to read or execute.** (a64a81b)
- **Merge pull request #2026 from GoogleCloudPlatform/moe_writing_branch_from_6802028852407616880** (5764272)
- **Automated g4 rollback of changelist 277771496.** (6a43ef7)
- **Added --cluster_boot_test_port_listening for cluster_boot benchmark. When** (75df9cb)
- **Use py37 as the tox testing environment.** (7edd7aa)
- **Merge pull request #2024 from GoogleCloudPlatform/moe_writing_branch_ruwa** (2df6a1c)
- **Make num_vms_per_host a shared flag across clouds** (7840706)
- **Creates samples of /proc/cpuinfo** (e294026)
- **Remove the remote file before pushing a local file to the remote place. This can avoid permission issue as the pushed file only allows user "per...** (7629dd9)
- **Unifying metadata creation on cluster boot.** (85f150a)
- **Log the http server requests and responses to stderr to debug 'backend error'** (39eee7a)
- **Remove the remote file before pushing a local file to the remote place. This can avoid permission issue as the pushed file only allows user "perfkit" to do read or execute.** (750ccca)
- **Using "floor division" to compute the value of terasort_dataset_size_in_GB, necessary due to Python 3 migration.** (784aae6)
- **Added testing for fio offset_increment as part of fio_parameters flag.** (64f45ec)
- **Increase timeout for GkeCluster creation.** (a445428)
- **Have PkbCommonTestCase extend absltest.TestCase for py3 test backports.** (5b05b46)

_Recap by [Repo Wrapped](https://repowrapped.com/gh/GoogleCloudPlatform/PerfKitBenchmarker?utm_source=github-action)._