Cloud Center Release Notes

December 2018

• Functionality being removed or changed

  Functionality	Result	Use Instead	Compatibility Considerations
  Cloud Center support for the Amazon EC2® Classic network type has been removed.	Errors	Set up a new cluster using the Amazon EC2 Virtual Private Cloud (VPC) network type.	Amazon® Web Services withdrew support for the EC2 Classic network for new accounts in 2013. To continue using MATLAB® with cloud resources, see Configure AWS VPC for Cloud Center.
  Cloud Center support for running MATLAB version R2016b will be removed in a future release. For more information on the migration policy, see Requirements for Using Cloud Center.	Warns	MATLAB versions R2018b, R2018a, R2017b, or R2017a.	As newer versions of MATLAB are supported, the support for older versions will be removed in future releases.

October 2018

• Functionality being removed or changed

  Functionality	Result	Use Instead	Compatibility Considerations
  Cloud Center support for the Amazon EC2 Classic network type will be removed in a future release.	Warns	Set up a new cluster using the Amazon EC2 Virtual Private Cloud (VPC) network type.	Amazon Web Services withdrew support for the EC2 Classic network for new accounts in 2013. To continue using MATLAB with cloud resources, see Configure AWS VPC for Cloud Center.
  Cloud Center support for running MATLAB version R2016b will be removed in a future release. For more information on the migration policy, see Requirements for Using Cloud Center.	Warns	MATLAB versions R2018b, R2018a, R2017b, or R2017a.	As newer versions of MATLAB are supported, the support for older versions will be removed in future releases.

September 2018

• Support for MATLAB Release R2018b including pretrained convolutional neural networks (CNN) for deep learning

  • You can run MATLAB Release R2018b on Cloud Center instances, including support for pretrained convolutional neural network (CNN) models. These networks are now available on all instances for MATLAB Release R2018b: AlexNet, DenseNet-201, GoogLeNet, Inception-ResNet-v2, Inception-v3, ResNet-18, ResNet-50, Resnet-101, SqueezeNet, VGG-16, and VGG-19. You can access the models using the functions alexnet, densenet201, googlenet, inceptionresnetv2, inceptionv3, resnet18, resnet50, resnet101, squeezenet, vgg16, and vgg19. You can use these pretrained models for classification and transfer learning. For details, see Deep Learning in Parallel and in the Cloud (Deep Learning Toolbox), and Pretrained Convolutional Neural Networks (Deep Learning Toolbox).

• Functionality being removed or changed

  Functionality	Result	Use Instead	Compatibility Considerations
  Cloud Center support for running MATLAB version R2016b will be removed in a future release. For more information on the migration policy, see Requirements for Using Cloud Center.	Warns	MATLAB versions R2018b, R2018a, R2017b, or R2017a.	As newer versions of MATLAB are supported, the support for older versions will be removed in future releases.

August 2018

• Functionality being removed or changed

  Functionality	Result	Use Instead	Compatibility Considerations
  Cloud Center support for running MATLAB version R2016a has been removed. For more information on the migration policy, see Requirements for Using Cloud Center.	Errors	MATLAB versions R2018a, R2017b, R2017a, or R2016b.	As newer versions of MATLAB are supported, the support for older versions will be removed in future releases.

April 2018

• Create clusters in dedicated headnode mode

  • Starting this release, you can create clusters in dedicated headnode mode. When enabled, the headnode instance exclusively runs management services, such as the job manager, and does not host any MATLAB workers. This cluster architecture improves performance. For details, see Use a Dedicated Headnode Instance for Management Services.

• Increased maximum number of workers to 1024

  • You can now create up to 1024 worker machines in VPC networks for MATLAB R2018a onwards.

• Functionality being removed or changed

  Functionality	Result	Use Instead	Compatibility Considerations
  Cloud Center support for running MATLAB version R2016a will be removed in a future release. For more information on the migration policy, see Requirements for Using Cloud Center.	Warns	MATLAB versions R2018a, R2017b, R2017a, or R2016b.	As newer versions of MATLAB are supported, the support for older versions will be removed in future releases.

March 2018

• Support for MATLAB Release R2018a including pretrained convolutional neural networks (CNN) for deep learning

  • You can run MATLAB Release R2018a on Cloud Center instances, including support for pretrained convolutional neural network (CNN) models. These networks are now available on all instances for MATLAB Release R2018a: AlexNet, VGG-16, VGG-19, GoogLeNet, ResNet-50, and ResNet-101. You can access the models using the functions alexnet, vgg16, vgg19, googlenet, resnet50, and resnet101. You can use these pretrained models for classification and transfer learning. For details, see Deep Learning in Parallel and in the Cloud (Deep Learning Toolbox), and Pretrained Convolutional Neural Networks (Deep Learning Toolbox).

• Functionality being removed or changed

  Functionality	Result	Use Instead	Compatibility Considerations
  Cloud Center support for running MATLAB version R2016a will be removed in a future release. For more information on the migration policy, see Requirements for Using Cloud Center.	Warns	MATLAB versions R2018a, R2017b, R2017a, or R2016b.	As newer versions of MATLAB are supported, the support for older versions will be removed in future releases.

February 2018

• Functionality being removed or changed

  Functionality	Result	Use Instead	Compatibility Considerations
  Cloud Center support for running MATLAB version R2016a will be removed in a future release. For more information on the migration policy, see Requirements for Using Cloud Center.	Warns	Migrate to MATLAB versions R2017b, R2017a, or R2016b.	As newer versions of MATLAB are supported, the support for older versions will be removed in future releases.

November 2017

• New pretrained convolutional neural networks (CNN) for deep learning

  • You can run MATLAB Release R2017b on Cloud Center instances, including support for pretrained convolutional neural network (CNN) models. These networks are now available on all instances for MATLAB R2017b: GoogLeNet, ResNet-50, AlexNet, VGG-16, and VGG-19. You can use these pretrained models for classification and transfer learning. For details, see the white paper Deep Learning with MATLAB and Multiple GPUs, and Pretrained Convolutional Neural Networks (Deep Learning Toolbox).

• Support for new P3 instance types

  • Cloud Center supports the new P3 instance family. P3s have up to 8 NVIDIA Tesla V100 GPUs and are well suited for high performance general computation including deep learning, modeling and data analysis.

  • For more details on newly added compute instances and regional availability, see Choose Supported EC2 Instance Machine Types.

• Updated IAM Role creation workflow

  • Cloud Center on-screen instructions are updated to help you through the latest workflow in the Amazon console to create IAM Roles.

  • For more details, see Create New IAM Role.

• Functionality being removed or changed

  Functionality	Result	Use Instead	Compatibility Considerations
  Cloud Center support for running MATLAB version R2015b has been removed.	Errors	MATLAB versions R2017b, R2017a, R2016b, or R2016a.	As newer versions of MATLAB are supported, the support for older versions will be removed in future releases.

September 2017

• Support for MATLAB Release R2017b including pretrained convolutional neural networks (CNN) for deep learning

  • You can run MATLAB Release R2017b on Cloud Center instances, including support for pretrained convolutional neural network (CNN) models. These networks are now available on all instances for MATLAB Release R2017a and MATLAB Release R2017b: AlexNet, VGG-16, and VGG-19. You can access the models using the functions alexnet, vgg16, and vgg19. You can use these pretrained models for classification and transfer learning. For details, see the white paper Deep Learning with MATLAB and Multiple GPUs, and Pretrained Convolutional Neural Networks (Deep Learning Toolbox).

• Support for new G3 instance types

  • Cloud Center supports the new G3 instance family. G3s have multiple GPUs with high single-precision performance well suited for deep learning, image processing and computer vision.

  • For more details on newly added compute instances and regional availability, see Choose Supported EC2 Instance Machine Types.

• Streamlined workflow for using data in the cloud

  • To work with data in the cloud, you can upload to Amazon S3, then use datastores to access the data in S3 from the workers in your cluster. For details, see Transfer Data To Amazon S3 Buckets.

• Functionality being removed or changed

  Functionality	Result	Use Instead	Compatibility Considerations
  Cloud Center support for running MATLAB version R2015a has been removed.	Errors	MATLAB versions R2017b, R2017a, R2016b, or R2016a.	As newer versions of MATLAB are supported, the support for older versions will be removed in future releases.
  Cloud Center support for running MATLAB versions R2015b will be removed in a future release.	Warns

June 2017

• Deep Learning: Pretrained convolutional neural network (CNN) models AlexNet, VGG-16, and VGG-19 are now available on all instances

  • Three pretrained convolutional neural network (CNN) models are now available on all instances for MATLAB Release R2017a only: AlexNet, VGG-16, and VGG-19. You can access the models using the functions alexnet, vgg16, and vgg19. These models are SeriesNetwork objects. You can use these pretrained models for classification and transfer learning.

    For more details, see Pretrained Convolutional Neural Networks (Deep Learning Toolbox).

• New “Getting Started with Cloud Center” documentation

  • For details, see Getting Started with Cloud Center

• Support for new Compute Optimized instance types

  • c4.xlarge, with 2 physical cores and 7.5 GB Memory

  • c4.2xlarge, with 4 physical cores and 15 GB Memory

  • c4.4xlarge, with 8 physical cores and 30 GB Memory

  • c4.8xlarge, with 16 physical cores and 60 GB Memory

  Note

  c4.8xlarge is the suggested default instance for clusters in a VPC.

  For more details on newly added compute optimized instances and regional availability, see Choose Supported EC2 Instance Machine Types.

• Support for new Memory Optimized instance types

  • r4.xlarge, with 2 physical cores and 30.5 GB Memory

  • r4.2xlarge, with 4 physical cores and 61 GB Memory

  • r4.4xlarge, with 8 physical cores and 122 GB Memory

  • r4.8xlarge, with 16 physical cores and 244 GB Memory

  • r4.16xlarge, with 32 physical cores and 488 GB Memory

  For more details on newly added compute optimized instances and regional availability, see Choose Supported EC2 Instance Machine Types.

• Support for new Storage Optimized instance types

  • i3.xlarge, with 2 physical cores, 30.5 GB Memory, and 950 GB ephemeral storage

  • i3.2xlarge, with 4 physical cores, 61 GB Memory, and 1900 GB ephemeral storage

  • i3.4xlarge, with 8 physical cores, 122 GB Memory, and 3800 GB ephemeral storage

  • i3.8xlarge, with 16 physical cores, 244 GB Memory, and 7600 GB ephemeral storage

  • i3.16xlarge, with 32 physical cores, 488 GB Memory, and 15200 GB ephemeral storage

  For more details on newly added compute optimized instances and regional availability, see Choose Supported EC2 Instance Machine Types.

• MATLAB Worker Amazon Machine Images (AMI) now support Ubuntu 16.04 LTS

• Functionality being removed or changed

  Functionality	Result	Use Instead	Compatibility Considerations
  Cloud Center support for running MATLAB version R2015aSP1 has been removed.	Errors	MATLAB versions R2015b, R2016a, R2016b or R2017a.	As newer versions of MATLAB are supported, the support for older versions will be removed in future releases.

March 2017

• Deep Learning in the cloud

  You can use MATLAB with Cloud Center to perform deep learning in the cloud using Amazon Elastic Compute Cloud (Amazon EC2) with new P2 instances. These instances provide access to NVIDIA^® Tesla K80 Accelerators with NVIDIA GK210 GPUs that include Error Correcting Code (ECC) memory protection and double precision floating point operations. For details, see the Deep Learning in the Cloud with MATLAB White Paper.

• MATLAB worker Amazon Machine Images (AMI) now support CUDA 8.0

• Functionality being removed or changed

  Functionality	Result	Use Instead	Compatibility Considerations
  Cloud Center support for running MATLAB versions R2015a and R2015aSP1 will be removed in a future release.	Warns	MATLAB versions R2015b, R2016a, R2016b or R2017a.	As newer versions of MATLAB are supported, the support for older versions will be removed in future releases.

January 2017

• Add local storage on each cluster node

  You can now use an Amazon EBS volume of type “General Purpose Solid-State Drive” ( "gp2") as local storage. You can specify existing snapshots to instantiate the volumes. For more information, see Create a Cloud Cluster.

November 2016

• Support for new GPU compute instance types

  • p2.xlarge, with 2 physical cores and 1 GPU

  • p2.8xlarge, with 16 physical cores and 8 GPUs

  • p2.16xlarge, with 32 physical cores and 16 GPUs

For more details on newly added GPU compute instances and regional availability, see Choose Supported EC2 Instance Machine Types.

• Support for new Memory Optimized instance types

  • r3.xlarge, with 2 physical cores and 30.5 GiB Memory

  • r3.2xlarge, with 4 physical cores and 61 GiB Memory

  • r3.4xlarge, with 8 physical cores and 122 GiB Memory

  • r3.8xlarge,with 16 physical cores and 244 GiB Memory

For more details on newly added memory optimized instances and regional availability, see Choose Supported EC2 Instance Machine Types.

October 2016

• MATLAB worker Amazon Machine Images (AMI) have been patched to mitigate CVE-2016-5195

September 2016

• Functionality being removed or changed

  Functionality	Result	Use Instead	Compatibility Considerations
  Cloud Center support for running MATLAB version R2014b has been removed.	Errors	MATLAB versions R2015a, R2015b, R2016a, or R2016b.	As newer versions of MATLAB are supported, the support for older versions will be removed in future releases.

August 2016

• Use Cloud Center with Amazon Virtual Private Cloud (VPC)

  • VPC enables Cloud Center users to launch clusters in a virtual network that you define. For more information, see Create a Cloud Cluster.

• Functionality being removed or changed

  Functionality	Result	Use Instead	Compatibility Considerations
  Cloud Center support for running MATLAB version R2014b will be removed.	Warns	MATLAB versions R2015a, R2015b, R2016a, or R2016b.	As newer versions of MATLAB are supported, the support for older versions will be removed in future releases.

March 2016

• MATLAB worker Amazon Machine Images (AMI) now support CUDA 7.5

• Functionality being removed or changed

  Functionality	Result	Use Instead	Compatibility Considerations
  Cloud Center support for running MATLAB version R2014a has been removed.	Errors	MATLAB versions R2014b, R2015a, R2015b, or R2016a.	As newer versions of MATLAB are supported, the support for older versions will be removed in future releases.