Specifications include, but are not limited to: 1. Can provide a multi-year contract agreement and usage credits with a one-time, upfront payment a. Pre-purchased usage credits do not expire 2. Cloud agnostic a. The solution can be deployed/provisioned on AWS, Azure, or Google clouds with no additional cloud accounts or cloud resources management/administration required b. Ability to replicate and/or failover to another region/zone or another cloud c. Provide automatic multi-zone system/data availability 3. Fully managed solution with near-zero administration (1 FTE or less) a. No cloud infrastructure to manage b. No outages for system maintenance or upgrades c. No partitioning, indexing, or vacuuming d. No manual backups 4. Provide an SLA of 99.95% uptime at a minimum 5. Ability to easily create unlimited operational accounts under one master organizational account to allow for separate business and data domains for storage/processing costs management, usage tracking, and billing 6. Provide consumption-based pricing a. Only pay for data storage and data processing resources that are used i. Provide resource consumption pricing with per-second metering b. Able to analyze costs by compute cluster, by query, by user, by role, etc. c. Provide graphical and SQL interfaces for understanding resource and storage consumption d. Able to set resource monitors and query statement timeouts to facilitate cost controls 7. Provide near-infinite scalability and elasticity a. Full separation of storage and compute with all compute clusters having access to all stored data b. Unlimited data storage with data automatically compressed, encrypted, and replicated across multiple availability zones i. Storage costs are based on compressed storage volume not uncompressed/raw data volume ii. Data processing performance should not be impacted by data compression/decompression when data are used iii. Native end-to-end data encryption (AES256 strong encryption or better) when data are in motion and at rest c. Ability to create unlimited number of custom sized compute clusters with the appropriate resources for desired, and often variable sized, workloads d. Full workload separation (able to scale compute resources) to eliminate concurrency issues i. Ability to scale up/down to process large jobs to meet data processing time constraints ii. Ability to scale in and out for high concurrency 1. Concurrency means being able to run multiple simultaneous processing jobs concurrently on the same data without competition for computational resources iii. Scaling must be non-disruptive in that resources can be scaled without stopping compute clusters or rerunning queries e. Compute resources should automatically suspend after a short time of non-use and resume when called upon f. Provide the ability to reuse query result sets without re-executing queries
