Performing risk calculations on Google Cloud

The financial services industry faces a complex landscape of continuously evolving regulations, dynamic market conditions, and ever-increasing data volumes. To effectively manage risk and make informed decisions, financial institutions (FSIs) are increasingly turning towards high performance computing (HPC) technology.

HPC empowers FSIs to achieve the following:

  • Accelerate risk calculations: Perform complex simulations and calculations with increased speed and accuracy, enabling faster risk assessment and decision-making.
  • Scale computations: Handle massive data sets and run parallel computations across high-performance machines, reducing processing time and enhancing efficiency.
  • Optimize resource utilization: Leverage on-demand resources and elasticity of the cloud to optimize infrastructure costs and manage resource allocation effectively.

Overview

Here's what you can expect to find in the following sections:

  • Google Cloud for risk calculations: We'll explore the advantages of using Google Cloud for risk calculations, highlighting its unique capabilities and benefits for FSIs.
  • Reference architecture: This section lays out a reference architecture for building a typical risk calculation solution on Google Cloud using HPC resources.
  • Solution components: We delve deeper into the key components of the solution, including compute, storage, networking, and other essential tools and services.
  • Getting started with risk calculations: This section provides a practical guide for FSIs to begin implementing risk calculations on Google Cloud using HPC.
  • Benchmarks: We showcase performance benchmarks to demonstrate the efficiency and scalability of Google Cloud for risk calculation workloads.
  • Performance and cost optimization: This section provides helpful guidance on optimizing both performance and cost when utilizing HPC for risk calculations.
  • Case studies: We offer real-world examples of FSIs successfully utilizing Google Cloud and HPC technology for risk calculation purposes.

Google Cloud for risk calculations

In today's dynamic financial marketplace, risk calculations are increasingly critical for financial institutions (FSIs) to assess and manage their exposure to potential losses. Performing these calculations often involves complex simulations and modeling, demanding significant computational power and resources. This is where Google Cloud excels for FSIs seeking robust and efficient risk calculation capabilities.

Benefits of Google Cloud for risk calculation workloads

  • Ease of use: Google Cloud offers intuitive interfaces and APIs, along with comprehensive documentation and support resources, making it easier for FSIs to configure and manage their risk calculation workflows.
  • High performance: Google Cloud features powerful infrastructure equipped with advanced CPUs, GPUs, and TPUs, enabling FSIs to execute risk calculations rapidly and efficiently. VMs like H3 and C3D, specifically optimized for high-performance workloads, facilitate effortless scaling of risk calculations.
  • Scalability and flexibility: Google Cloud's elastic architecture allows FSIs to dynamically adjust compute resources based on changing workload demands, eliminating the need for expensive on-premises infrastructure and enabling a pay-as-you-go model.
  • Security and compliance: Google Cloud adheres to strict industry standards and regulations, ensuring the highest levels of security and compliance for sensitive financial data used in risk calculations. This provides FSIs with peace of mind knowing their data is safeguarded in a trusted cloud environment.
  • Faster time-to-market: Google Cloud's robust infrastructure empowers FSIs to rapidly perform risk calculations, enabling them to respond quickly to market shifts and accelerate the time-to-market for new offerings.
  • Cost optimization: Google Cloud's flexible pricing models allow FSIs to optimize resource utilization, leading to significant cost savings compared to traditional on-premises solutions.
  • Enhanced agility and innovation: By eliminating the complexities of managing on-premises infrastructure, Google Cloud empowers FSIs to focus on innovation and developing new risk management strategies. This agility allows them to stay ahead of market trends and gain a competitive advantage.

Let’s take a look at the reference architecture, and how it makes Google Cloud a compelling solution for FSIs seeking to improve their risk calculation processes. Our performance, scalability, and security, coupled with comprehensive tools and services, empower FSIs to achieve unprecedented agility, efficiency, and innovation in risk management.

Reference architectures

Google Cloud offers a selection of platforms to perform risk calculations. These depend on the requirements of your enterprise.

Google Compute Engine, enabled by the HPC Toolkit offers an extensive set of architectures where appropriate Infrastructure can be easily provisioned.

GKE offers a flexible robust platform to enable a wide variety of options for risk calculations.

Batch is available as a managed service on Google Compute Engine. Infrastructure is provisioned on demand and deprovisioned when the work is completed. 

Dataflow (Apache Beam) can group trades and perform aggregation across extremely large datasets. It can provide both batch processing and streaming analytics.

Solution components

Compute

Google Cloud's powerful compute options provide the foundation for tackling demanding risk calculation workloads, featuring versatile VM instances and accelerators for optimal performance.

  • High-performance VM instances: Leverage Google Cloud's diverse range of VM instances optimized for HPC workloads. Examples include H2 instances for memory-intensive calculations and N2 instances for balanced compute and network performance.
  • GPUs and TPUs: Enhance the performance of risk calculations by utilizing powerful GPU and TPU resources for accelerated task execution.
  • Managed instance groups: Utilize managed instance groups to automate scaling and ensure optimal resource utilization based on workload demands.

Network

A robust and secure network infrastructure ensures efficient data transfer and communication within the risk calculation environment.

  • VPC network: Establish a secure and isolated Virtual Private Cloud (VPC) network to host the risk calculation infrastructure.
  • High-performance networking: Utilize high-bandwidth networking options like Cloud Interconnect and VPC Network Peering for superior data transfer speeds between VMs and storage resources.

Storage

High-performance storage solutions facilitate fast data access and management for risk calculation models and generated outputs.

  • Block Storage: Employ persistent disk options like HDD and SSD, and Local SSD for fast data access during risk calculations.
  • Cloud Storage: Leverage Cloud Storage for vast amounts of historical data and model outputs, leveraging its scalability and cost-effectiveness.
  • Cloud Filestore: Use Cloud Filestore for high-performance shared storage accessible by multiple VMs within the risk calculation cluster.

Tools

Specialized tools simplify the deployment, management, and monitoring of the risk calculation infrastructure, ensuring smooth workflow operations.

  • Cloud HPC Toolkit: This toolkit simplifies the deployment and management of HPC clusters on Google Cloud, enabling flexible configuration and easy scaling.
  • Pre-configured VM Images: Utilize pre-configured VM images with popular risk calculation applications and tools pre-installed for faster environment setup.
  • Cloud Monitoring: Monitor and analyze the performance of risk calculation workflows using comprehensive monitoring tools for real-time insights and resource optimization.

Getting started with risk calculations

Embarking on your journey to leverage Google Cloud for risk calculations? Here's a roadmap to navigate the implementation process effectively.

Considerations

  • Define your goals: A well-defined set of objectives helps identify the specific risk calculations you need to perform and the desired outcomes.
  • Evaluate current workflows: Analyze your existing risk calculation processes to determine areas for improvement and assess potential migration to the cloud.
  • Identify your data needs: Map out the data sources required for your risk calculations and ensure data accessibility in the cloud environment.
  • Choose your tools: Select the appropriate Google Cloud HPC resources and tools that align with your risk calculation workflow requirements.
  • Security and compliance: Develop a robust security strategy for your cloud environment, adhering to relevant industry regulations for financial data protection.

Deployable Blueprints/Code

Benchmarks

To objectively assess the performance of Google Cloud for risk calculation workloads, we leverage industry-standard benchmarks. The following section highlights the STAC benchmark results, showcasing Google Cloud's capabilities in this domain.

STAC benchmarks

The Securities Technology Analysis Center (STAC) publishes a suite of industry-standard benchmarks designed to evaluate the performance and efficiency of technology solutions for financial services workloads. Google Cloud has participated in multiple STAC benchmarks, demonstrating its capabilities in risk calculation scenarios.

Following up on our previous STAC-M3 benchmark audit (SUT ID KDB181001), a redesigned Google Cloud architecture leveraged the most recent version of kdb+ 4.0, the time-series database from KX, and achieved significant improvements: 35 out of 41 benchmarks ran faster in the new cluster—by up to 18x faster than Google Cloud’s prior results. Key highlights include:

  • Compared to the previous STAC-M3 Antuco suite results on Google Cloud:
  • Was faster in 13 of 17 mean response-time benchmarks
  • Was 18x faster—a 94% reduction in run time—in the version of Year-High Bid that allows caching (STAC-M3.ß1.1T.YRHIBID-2.TIME), which also set an overall record for all published results
  • Had 9x higher throughput in Year-High Bid (STAC-M3.ß1.1T.YRHIBID.MBPS)
  • Compared to the previous STAC-M3 Kanaga suite results on Google Cloud:
  • Was faster in 22 of 24 mean response-time benchmarks
  • Was over 10x faster in all four Market Snapshot workloads (STAC-M3.ß1.10T.YR[2,3,4,5]-MKTSNAP.TIME)
  • Had 5x the throughput in Year-High Bid involving 2 years of data (STAC-M3.ß1.1T.2YRHIBID.MBPS)

Our recent STAC benchmark blog post includes an in-depth analysis of Google Cloud's performance in the STAC benchmarks.

Performance and cost optimization

Choosing the right machine type

Google Cloud provides a diverse range of machine types, each offering different CPU, GPU, and memory configurations. Selecting the right machine type for your risk calculation workload significantly impacts performance and cost.

Consider the following factors when choosing your machine type:

  • CPU performance: Choose instances with higher core counts and clock speeds for computationally intensive risk models.
  • Memory requirements: Select instances with sufficient memory to handle the data size and complexity of your risk calculations.
  • Cost optimization: Balance performance needs with cost by considering factors like burstable performance and preemptible instances.

Selecting optimal storage

Google Cloud offers a range of storage options with varying performance and cost characteristics. Choosing the right storage for your risk calculations is crucial for efficiency and cost optimization.

Here are different storage types to consider:

Home storage

  • Stores shared user files, configurations, scripts, and post-processing data
  • Persistent and typically built on the NFS protocol
  • Options include Google Cloud Filestore and partner offerings like NetApp Cloud Volumes Service

Scratch storage

  • Used for temporary files like intermediate results and simulation output data
  • May or may not be shared across different nodes
  • Typically built on high-performance systems like local flash storage or parallel file systems
  • Google Cloud-native options include Persistent Disk, Local SSD, Cloud Filestore, and Parallelstore
  • Partner offerings include NetApp Cloud Volumes Service, DDN EXAScaler Lustre, Sycomp (IBM Storage Scale), and Weka

Archive storage

  • Used for storing infrequently accessed data at a cost-effective price
  • Google Cloud Storage is our recommended option for archive storage

Leveraging Spot VMs for cost efficiency

Spot VMs offer a potential cost-effective solution for running risk calculations. These VMs come at a significant discount but can be terminated with short notice.

If your workflow can tolerate potential interruptions and runs within a short timeframe (< 4 hours), consider testing Spot VMs. Customers have found that the up to 91% cost savings versus on-demand saves them enough to tolerate minor interruptions.

Note that not all VM types, like H3, support the Spot model.

Case studies

To illustrate the practical application of Google Cloud and HPC technology for risk calculations, we present case studies highlighting successful implementations by leading FSIs.

HSBC

HPC Days Digital: HSBC customer spotlight

This video showcases how HSBC leverages Google Cloud HPC solutions for risk management and regulatory reporting.

HSBC STAR completes 4.4 billion risk calculations a day with Google Cloud solutions. This blog post details how HSBC successfully migrated its Strategic Trade Allocation and Rebalancing (STAR) platform to Google Cloud, achieving significant performance and efficiency improvements.

HSBC STAR completes 4.4 billion risk calculations a day with Google Cloud solutions

Goldman Sachs

Goldman Sachs customer spotlight: Delivering scale, agility, and trust

This video features how Goldman Sachs utilizes Google Cloud for building a highly scalable and secure infrastructure to support critical financial activities, including risk calculations.

Learn more

Learn more about Google Cloud HPC

Google Cloud HPC

Google Cloud HPC: Learn more about Google Cloud’s HPC solutions across industries.

Cloud HPC Toolkit: Get started with the Cloud HPC toolkit to easily configure and deploy your HPC infrastructure on Google Cloud.

Cloud HPC Toolkit blueprint catalog: Choose from a selection of pre-built HPC Toolkit blueprints to easily deploy complete HPC environments.

Google Cloud HPC blogs: Read more in Google’s HPC-focused blog channel.

Financial services

Value at risk (Wikipedia)

Google Cloud Solutions for financial services: Explore Google Cloud's range of solutions tailored to meet the unique needs of FSIs.

Google Cloud HPC for financial risk simulations one pager

Google Cloud