SPARK, more robust with Sequoia

Spark AdminNews

Since migration to Equinox as host, SPARK has been up and available 99.99% of the time. Great, right? And now SPARK is about to become even more stable and robust! Along with the move to Evergreen 2.6, PaILS will migrate to Sequoia to host SPARK this summer.

Sequoia Technical Description

Physical Structure
– Redundant, over-provisioned power
– 20TB of redundant iSCSI-based SAN space for application and configuration storage
– Redundant Sequoia Controller servers, managing configuration and customization
– Minimum eight-way redundant Sequoia cloud servers per cluster
– 32+ CPU cores per server
– 2.3GHz+ AMD Opteron processors
– 128GB+ RAM
– Redundant ingress servers for customization access
– Massive three-way redundant Database Servers available per cluster
– 48-64 CPU Cores
– 2.2-3.lGHz AMD Opteron processors o256GB+ RAM
– SSD-backed database storage

Virtual Structure
– Service-specific provisioning toavoid negative inter-service interaction
– Vertically provisioned middleware stack toavoid hardware-based outages
– Horizontally provisioned front-end stack toavoid load-balancing outages
– Single code base per Sequoia cluster; Single cluster per software version
– Automated resource deployment toavoid mismatched or incorrect configuration
– Database-backed cluster and application configuration management
– Git-based customization management

Health and Capacity Monitoring
– Continuous utilization monitoring toenable capacity analysis and planning
– SNMP-based
– MRTG visualization
– Historical and predictive data
– Continuous health and status monitoring toenable immediate reactive support responses o Nagios-based
– Topological dependency-aware monitoring
– Direct alerting
– Situation-appropriate automated initial response
– Configuration and Customization Access
– SSH, SFfP and SCP access tocustom template files
– Simple, menu-driven text interface for Command and Control
– Service status and control
– Template and configuration deployment
– Instance configuration
– Backend service output, such as notices and reports

Data Security
– Data is continuously streamed toredundant database servers
– Allows for restoration of data tothe moment of failure.
– Full nightly backups
– Weekly offsite tape rotation

Spark AdminSPARK, more robust with Sequoia