Anthropic’s mission is to develop safe and capable artificial intelligence. A critical part of this process is preparing the massive datasets needed to train our state-of-the-art language models - a task that requires high-performance storage and compute capabilities. We’ve found that Amazon EC2 i4i instances, with their combination of local NVMe storage (up to 30 TB of Nitro SSD) and powerful compute, are uniquely suited to handle our data preparation and experimentation workloads. The high I/O performance, security, and scalability of i4i have allowed us to iterate faster and more efficiently, ultimately accelerating our progress in AI research and development.

Indeed continuously incorporates many of AWS' latest innovations and solutions. Indeed frequently uses distributed databases, and scan times of those databases are extremely important. Indeed migrated from Amazon EC2 r6i.4xlarge instances to Amazon EC2 i4i.32xlarge instances. By migrating their instances, Indeed saw a 97% performance gain with scan time dropping from a high of 24.25 seconds to 0.65 seconds. Indeed is looking forward to future innovations from AWS, with a continued partnership and reliance on AWS's services for Indeed’s operations.

We are thrilled with the performance we are seeing from the Amazon EC2 I4i instances which use the new low latency AWS Nitro SSDs. Our testing shows I4i instances delivering an astonishing 2.9x higher query throughput than the previous generation I3 instances. We have also tested with various read and write mixes, and observed consistent and linearly scaling performance.

When we tested I4i instances, we observed up to 2.7x increase in throughput per vCPU compared to I3 instances for reads. With an even mix of reads and writes, we observed 2.2x higher throughput per vCPU, with a 40% reduction in average latency than I3 instances. We are excited for the incredible performance and value these new instances will enable for our customers going forward.

I4i instances have demonstrated superior performance over previous generation I3 instances, with a 30% improvement across operations. We look forward to using I4i to further elevate performance for our customers.

Aerospike running on EC2 I3en instances already delivers industry leading real-time performance for read/write workloads. With EC2 I4i instances, we have observed a 70% increase in performance on read workloads alone. This combined with the proven ability of Aerospike to deliver high throughput of writes means Aerospike running on I4i instances will deliver the best price/performance in the industry for real-time access to petabytes of data running on EC2.