Solutions

Overview

GraudFrame, provided by Tocharian, is committed to meeting customers' observability needs. Our solution is based on the Elastic Stack (including Elasticsearch, Kibana, Logstash, Beats, Fleet/Agent) and software processes that incorporate custom machine learning and large language models (LLM). GraudFrame has undergone extensive customization and optimization at the hardware, operating system, and software levels to ensure optimal performance and data security.

Technical Architecture

GraudFrame adopts a layered architecture, including the following levels:

• Data Ingestion Layer: Efficiently collects data using open-source components of the Elastic Stack, such as MetricBeat, FileBeat, and Elastic Agent.
• Storage and Retrieval Layer: Ensures efficient data storage and fast retrieval through highly customized hardware configurations and layered storage strategies.
• AI Analysis Layer: Integrates custom machine learning and AI models to provide intelligent data analysis and decision support.
• User Interface and API Layer: Provides real-time visualization using tools like Kibana and enables flexible data access and integration through APIs.

With this architectural design, GraudFrame can efficiently process and analyze large volumes of data, offering real-time visualization and intelligent decision support.

Hardware Customization

GraudFrame has undergone extensive hardware customization to ensure optimal performance at different data levels:

• Hot Nodes
  • CPU: Default use of high-performance AMD CPUs, providing powerful multi-core processing capabilities.
  • Storage: NVMe SSDs to ensure high-speed read/write performance, supporting high-frequency data access and real-time queries.
• Warm Nodes
  • CPU: Flexible choice between Intel or AMD processors, optimizing performance and cost based on specific needs.
  • Storage: SSD disks, balancing storage density and access speed, meeting medium-frequency data access needs.
• Cold Nodes
  • Storage: Utilizes distributed S3 and redundancy mechanisms to separate archived data from the Elasticsearch environment. Each storage retains at least one replica to ensure high availability and durability. All Elasticsearch data is synchronized to S3, providing multi-layer security and protection against single points of failure.

In addition, GraudFrame places great emphasis on the physical environment of data centers, ensuring stability in temperature, humidity, and power supply to ensure the long-term reliable operation of hardware.

Operating System Optimization

• TCP communication and NIC configuration: Optimize network settings to reduce latency and improve data transmission efficiency.
• Firewall strategy: Employ efficient firewall strategies to ensure system security.
• CPU clock and configuration: Adjust CPU clock frequency and core allocation to maximize software performance.
• Network optimization: Optimize TCP stack and NIC drivers to ensure efficient data transmission and low-latency communication.

Software Components & Data Ingestion

GraudFrame utilizes open-source components of the Elastic Stack for data ingestion and manages agents and environmental permission isolation through Fleet to ensure data ingestion is secure and efficient. Specific measures include:

• Sensitive information anonymization: Automates processing of sensitive data such as personal names, IDs, addresses, and health information to ensure compliance with GDPR and other industry standards.
• Industry-specific configurations: Customizes data collection and processing for specific industries, such as finance and healthcare, to ensure data compliance.
• Fleet management: Uses Fleet for agent management and permission isolation based on the customer's environment and data sources, ensuring secure and efficient data ingestion.

Data Security & Compliance

GraudFrame strictly adheres to industry security standards, including SOC2 and GDPR. Specific measures include:

• Internal network security: All Elasticsearch cluster nodes run in an internal network environment, providing secure encrypted external access through a Coordinating Node, with internal communication encrypted entirely via TLS.
• Authentication integration: Supports LDAP, PKI, and other authentication methods supported by the paid version of Elasticsearch to ensure secure user access.
• Granular permission management: Provides document-level and field-level security, achieving strict permission isolation and management.
• Data encryption: All data is encrypted using advanced encryption techniques during transmission and storage to prevent data leakage and unauthorized access.
• Compliance support: Provides customized compliance configurations and data processing workflows for specific industries, such as finance and healthcare, based on customer needs.

AI Integration

• We offer robust AI integration capabilities, supporting external AI model APIs such as OpenAI and Claude, allowing seamless integration with top AI technologies.
• Our platform also supports custom model deployments with GPU cluster configurations, enabling optimized performance and resource efficiency.
• For clients requiring enhanced security and isolation, we provide fully local deployment of AI models with capabilities close to GPT-4, offering high-level inference performance in an isolated environment.
• This allows you to integrate your own ChatGPT-like models into our platform, gaining valuable AI-driven insights into observability and alerting while maintaining complete control over your data and environment.

Alerts & Monitoring

GraudFrame utilizes tools such as Kibana for visualization and machine learning jobs and integrates custom AI models to provide data insights. Our alerting system supports integration with various notification channels, such as Email, Webhook, Jira, MS Teams, OpsGenie, PagerDuty, Slack, IBM Resilient, ServiceNow®, Tines, Torq, etc. Alert rules are based on machine learning, KQL, and the latest ESQL rules, ensuring timely responses to critical events, helping customers quickly address potential issues and ensuring business continuity.

Data Storage & Lifecycle Management

Based on the Elasticsearch cluster architecture and hardware configuration, GraudFrame adopts a Hot-Warm-Cold-Archive data lifecycle strategy:

• Hot Storage (7 or 30 days): Suitable for high-frequency accessed data, ideal response time ≤ 2 seconds.
• Warm Storage (90 days): Suitable for medium-frequency accessed data, response time 2-5 seconds.
• Cold Storage (180 or 360 days): Suitable for low-frequency accessed data, response time 5-15 seconds.
• Archive Storage (in searchable snapshot form): Used for long-term storage and occasional access data.

This layered storage strategy ensures efficient management and access to data at different lifecycle stages, optimizing the use of storage resources.

Pricing Strategy

GraudFrame's pricing is determined based on the specific needs of customers (daily data ingestion volume, storage time for different lifecycles, response requirements, etc.) and is divided into two packages: Basic and Prime, to meet the needs of customers of different scales. Specific pricing factors include:

• Daily data ingestion volume: Determines the required resource configuration and storage needs based on the data volume to be ingested daily.
• Storage lifecycle: Adjusts the allocation of storage resources and costs based on how long data is stored at hot, warm, cold, and archive levels.
• Response requirements: Determines the pricing range based on customer needs for query response times and suitable storage levels and hardware configurations.

Data Center & Legal Compliance

GraudFrame's data centers are located in Nuremberg, Germany, Falkenstein, Germany, and Helsinki, Finland, strictly adhering to GDPR and the strictest German digital laws, ensuring data is stored and processed in a secure and compliant environment. Specific measures include:

• Physical security: Data centers have strict physical security measures to prevent unauthorized access.
• Environmental control: Ensure stability in temperature, humidity, and power supply in data centers to guarantee the normal operation of hardware devices.
• Legal compliance: Fully complies with data protection regulations and industry standards in the area, ensuring the legality and compliance of data processing.

Scalability & Flexibility

• Horizontal and vertical scaling: GraudFrame can easily accommodate data growth by adding more nodes or upgrading existing hardware.
• Flexibly adapting to different environments: GraudFrame can be deployed flexibly in local, cloud, or hybrid cloud environments to meet diverse business needs.

Reliability & High Availability

• High availability architecture: GraudFrame utilizes Elasticsearch's Hot-Warm-Cold architecture with multiple replicas across different nodes, ensuring data safety and protection against single points of failure.
• Customized storage strategies: Based on customer-specific requirements, we implement RAID1 disk deployments to enhance redundancy and data protection at the hardware level.
• Cold and archival data management: For cold and archived data, we leverage distributed object storage using S3's multi-node storage, integrated with MinIO's erasure coding mechanism for efficient data durability and fault tolerance.
• Disaster recovery plans: Provides comprehensive data backup and recovery strategies tailored to safeguard data and ensure business continuity in the event of a disaster.

Support & Maintenance

• Technical support: GraudFrame offers 24/7 technical support services, equipped with dedicated account managers to ensure rapid response to customer needs.
• System maintenance: Regularly conducts system maintenance, updates, and security patch management to ensure long-term stable operation of the system.

Training & Onboarding

• Training programs: GraudFrame offers training courses for customer teams to help them quickly master system usage and management.
• Documentation & resources: Provides detailed user manuals, online documentation, tutorial videos, and knowledge base resources to support self-learning and problem-solving.

Compliance & Certifications

• Industry certifications: GraudFrame has received multiple industry certifications and awards, enhancing customer trust.
• Additional compliance: Adheres to HIPAA, PCI DSS, and other regulations and standards to meet compliance needs in various industries.

Sustainability

Tocharian is committed to environmental protection and sustainability, with all data centers for GraudFrame powered by wind and hydropower, ensuring the renewable and eco-friendly nature of energy sources. Our data centers have relevant certifications demonstrating compliance with environmental standards and sustainability principles. This not only reduces carbon footprints but also ensures stable energy supply, supporting our efficient and green operational model.