9. Benchmarks

UltiHash is a high-performance, resource-efficient solution built for AI and advanced analytics applications. In the fast-paced tech industry, organizations race to develop AI-powered technologies and leaders have less time to make informed decisions on their data infrastructure, increasing resource consumption. UltiHash optimizes data volume out of the box through a built-in deduplication algorithm that eliminates redundancies at a byte level, regardless of data type or format. This results in significant space savings of up to 60% on the entire data volume, depending on various factors including:

• compressed vs uncompressed data format: UltiHash generates up to 75% space savings on uncompressed formats (e.g. RAW, TIFF) and up to 51% on compressed formats (e.g. JPG, PNG)

• the similarity between the objects: the higher the similarity level, the more space savings generated

Space savings benchmarks

This section documents the space savings generated by UltiHash on different datasets, giving a fair demonstration of UltiHash’s capabilities. The results can be reproduced on any UltiHash cluster. Our benchmark is updated with new data on a frequent basis, you can submit a request with a desired data source to [e-mail].

Format	Description & link	Size	Space savings
RAW	Human and animal scans	1.48 GB	74%
JPG	Self-driving cars	2.6 GB	53%
PNG	Defect detection on texture surfaces (DAGM 2007)	6 GB	50%
TIFF	Images of climate data	16 GB	46%
CSV	Disease prediction symptoms	1.4 MB	42%

Performance benchmarks

UltiHash Object Storage was built for AI and advanced analytics workloads, which require high performance to give users the best time-to-value possible. Given this, UltiHash utilizes an optimized, lightweight algorithm and architecture tailored for high throughput and low latency.

This section provides users an overview of UltiHash throughput performance, including a comparison to S3 Standard and MinIO.

Benchmarks were run on a UltiHash cluster with 4 storage nodes of the type m5dn.24xlarge, in an AWS Virtual-Private Cloud. The network bandwidth in the cluster is configured at 100 Gb/s. The storage disks are physically attached to the instances (instance store) to provide optimal performance. The test setup is writing and reading to a single m5dn instance.

Our performance test was performed on a 110 GB dataset of 64MB average object size: UltiHash version 0.5.4 achieves an average throughput of 500.27 MB/s for PUT and 1578.59 MB/s GET operations.

To set UltiHash in comparison with other solutions, the benchmark includes performance measurements of PUT and GET operations with the same dataset against S3 Standard and MinIO.

To test the performance of S3 with our reference dataset, the setup used was one EC2 instance of type m5dn.24xlarge that connects to S3 via VPC Gateway Endpoint. The results of the test indicate that the read operation provided by UltiHash is 248% faster than S3, while the write throughput is similar between both solutions. Here is a summary of the results:

Benchmark	PUT	GET
S3 Standard	496.02 MB/s	587.64 MB/s
UltiHash	500.27 MB/s	1578.59 MB/s

MinIO is an object storage solution designed for performance optimization. In direct comparison to UltiHash it must be noted that their team made an early a design decision that resource-efficiency achieved by deduplication would not be part of their product. The results of the benchmarking test demonstrate that UltiHash and Min.io offer similar performance.

Benchmark	PUT	GET
MinIO	587.64 MB/s	1597.05 MB/s
UltiHash	500.27 MB/s	1578.59 MB/s

In conclusion, these performance benchmarks show how UltiHash is differentiated from managed services such as AWS S3, providing a read operation almost 250% higher. The throughput comparison to MinIO highlights that UltiHash reaches similar level of performance, demonstrating that throughput and resource efficiency need not be mutually exclusive with the right solution.