Züs Cache Preliminary Datasheet

This section introduces Züs Cache, the AWS-hosted deployment of the Züs storage platform available via the AWS Marketplace. It provides an end-to-end overview of the product’s architecture, key capabilities, configuration parameters, and benchmark results.

Züs Cache enables users to deploy a zero-trust, blockchain-secured storage cluster within their own AWS account, offering scalability, data ownership, and encrypted collaboration.

This datasheet serves as a technical reference for solution architects, DevOps engineers, and IT administrators who wish to understand the capabilities, performance characteristics, and operational parameters of Züs Cache on AWS before provisioning or onboarding the service.

High Level Architecture

ZS3Server – a gateway interface server that allows users to interact with the Züs storage cluster using standard S3-compatible APIs. Check out the Github here → https://github.com/0chain/zs3server
Blobber – a storage server which can mount EBS volumes or connect to an EFS file system. Check out the Github here → https://github.com/0chain/eblobber

Figure 1: Züs storage architecture of a single cluster on AWS

Recommended Allocation Parameters

Settings

Description

Value

Notes

Data/Parity (erasure code)

Data and Parity Shards

4 data, 1 parity or 8 data, 1 parity

Select 8/1 for higher throughput, lower latency, 4/1 for lower cost

EBS Provisioned storage

64 TB or 128 TB

Example: 4 data × 16 TB EBS volume = 64 TB for a 4/1 cluster

EFS

Usage-based storage

Unlimited

Automatically tiers to EFS-IA or Archive

Performance Results

The objective is to benchmark a single node Züs cluster performance for different data/parity combination and object sizes. These configurations and DevOps processes will be available on AWS Marketplace via Blimp.Software soon.

Test Setup

Eblobber HW/instance type: c6i.2xlarge Storage: Mounted EBS (gp3): 500 GiB; IOPS 3000; Throughput 1000 MiB/s

ZS3Server Version used: https://github.com/0chain/zs3server/pull/177 HW/instance type: c5n.18xlarge zs3server.json file:

{
  "encrypt": false,
  "compress": false,
  "max_batch_size": 40,
  "batch_wait_time": 70,
  "batch_workers": 72,
  "upload_workers": 72,
  "download_workers": 72,
  "max_concurrent_requests": 400
}

Allocation – created through zboxcli Check out the Github here → https://github.com/0chain/zboxcli Data/parity: 4/1 and 8/1 Size: 800 GB

Test Procedure

Used Warp to benchmark the performance. Instructions can be found here.: https://docs.zus.network/zus-devops/zus-network-testing/performance-test/zs3server-warp-benchmark-testing
To generate benchmark data, each test was conducted 3 times in a row. Commands used were as follows:

./warp put --host <zs3server-ip:port> --access-key rootroot --secret-key rootroot \
--bucket benchmark --duration 3m --concurrent 16 --obj.size 128MB

./warp get --host <zs3server-ip:port> --access-key rootroot --secret-key rootroot \
--bucket benchmark --duration 3m --concurrent 16 --obj.size 128MB --objects 32

./warp put --host <zs3server-ip:port> --access-key rootroot --secret-key rootroot \
--bucket benchmark --duration 3m --concurrent 16 --obj.size 512MB

./warp get --host <zs3server-ip:port> --access-key rootroot --secret-key rootroot \
--bucket benchmark --duration 3m --concurrent 16 --obj.size 512MB --objects 32

Warp Results

Results are generated based on the Warp configuration above.

GET

Metric

4/1 (128 MB)

4/1 (512 MB)

8/1 (128 MB)

8/1 (512 MB)

Throughput (GiB/s)

2.5

4.6

Latency (s)

0.8

3.1

0.4

1.7

TTFB (s)

0.3

1.0

0.2

0.4

PUT