Q4 Product updates and launches

Streamlined Microservices Workflow: from article management to order fulfilment

1. Article Database Setup: Establishing the foundational database for storing article-related information.
**2. Stock Queue with SQS:**Implementing a stock queue using Amazon Simple Queue Service (SQS) to manage inventory levels.
3. Secret Manager and Configuration: Utilizing a secret manager for secure storage and management of application configurations and sensitive data.
4. Article App and Namespace Configuration: Setting up the article application, including defining the namespace for organizational and isolation purposes within the Kubernetes environment.
**5. SQS Trigger Integration:**Integrating SQS Trigger to automate processes based on messages in the SQS queue.
**6. Order Microservices - Database Setup:**Creating and setting up a database specifically for order management within the microservices architecture.
**7. Secret Manager for Order Service:**Configuring the secret manager to handle sensitive data related to the order service.
8. Container Setup for Order Service: Preparing and deploying the container for the order microservice.
9. Testing with Postman:

• Creating an Article: Using Postman to simulate the creation of an article in the system.
• Creating an Order: Using Postman to simulate placing an order in the system.

**10. Order Validation and Stock Check:**Final step involves validating the order and checking stock levels to ensure the order can be fulfilled.


q4-2023-use-case-1-v1 (1).mp4

Exploring advanced Block Storage capabilities
In this second use case, we delve into the performance comparison of our Block Storage Basic with Block Storage Low Latency. This analysis is crucial for understanding how various storage options can influence the efficiency and speed of data processing, key elements in high-performance computing.

Block Storage performance: Basic vs. Low Latency

Setup for the test:

• A PRO2-XXS instance was used for both testing scenarios.
• One instance was equipped with Block Storage Basic, while the other utilized Block Storage Low Latency (15K).
• Both storage types were mounted on /mount/sdb to maintain a uniform testing environment.
  
  Executing the JEDEC219 Test:
• Performed a heavy workload test focused mainly on write operations with random data, utilizing the FIO tool.
• This test simulates intense data workloads, mimicking real-world scenarios.
  
  Analysis of Results:
• Block Storage Basic demonstrated around 5000 IOPS (Input/Output Operations Per Second).
• A notable increase in performance was observed with Block Storage Low Latency, reaching nearly 10000 IOPS.

This test underscores the enhanced performance of our Low Latency Block Storage, highlighting its suitability for applications requiring rapid data access and processing.


q4-2023-use-case-2-v1.mp4

In these examples, we demonstrate the real-world implications and benefits of using advanced computing technologies. Through practical comparisons, we highlight the significant performance differences between various hardware configurations and technologies, providing insights into how these can be leveraged for more efficient and powerful computing solutions.

Inference x86 vs Inference ARM
Setup for inference tests: Utilizing a GP1-XL and an AMP2-C48, both equipped with 48 vCPUs, to conduct the tests.
Running inference: Employing Tensorflow along with Huggingface Transformers for inference. The pre-trained model 'bert-base-uncased' is used with small batch sizes (8) on short sequences (32/128/512).
Observations: Notable faster performance on ARM with reduced cost and energy consumption compared to x86.

Elastic Metal SSD vs Elastic Metal NVME performance test
Hardware selection: Conducting tests on Elastic Metal SSD (EM-B112X-SSD) and Elastic Metal NVME (EM-I210E-NVME).
Executing JEDEC219 Test: Performing a heavy workload test, predominantly write, with random data using FIO.
Results analysis: A significant increase in IOPS observed on NVME (reaching approximately 80k) compared to SSD (around 4k), highlighting the superior performance of NVME.


q4-2023-use-case-3-v1 (1).mp4