100 likes | 196 Views
Archive Retrieval—Beta Test March 14, 2003. Introduction. Current archiving system is inadequate for several reasons Majority of the software runs under OpenVMS Maintenance of the OpenVMS-based code is extremely difficult, hence costly Code changes are difficult to make
E N D
Archive Retrieval—Beta TestMarch 14, 2003 Implementation Review
Introduction • Current archiving system is inadequate for several reasons • Majority of the software runs under OpenVMS • Maintenance of the OpenVMS-based code is extremely difficult, hence costly • Code changes are difficult to make • Recompilation of the code often exposes latent bugs unrelated to the intended fixes • The requirements and assumptions that shaped the system are antiquated • Ingest and distribution services operate near the limit of their capacity • Mismatch between older S/W and newer peripherals • Operations objectives no longer met by system functionality • Flexibility in handling requests that require manual intervention • Support for “lights out” operations Implementation Review
Architectural Objectives • Meet the needs of archive users today and the future • Improve operator control of the system • Provide more control over the running system • Increase automation options • Improve throughput & reliability • Relieve system bottlenecks in unreliable areas of the system • Staging disk space & outgoing network connections • Reduce system downtime • Adjust use case of MO jukebox more in line with vendor expectations • Lessen impedance mismatch at the OTFR interface • Support future data products and services • Increasing request sizes due to large data products require higher capacity shippable media like DVD • High degree of platform independence • UNIX is favored in the current implementation • Improve system maintainability Implementation Review
Release Strategy • Distribution will be ready for release prior to server availability • Demands placed on the compute infrastructure by the new distribution architecture disallow deployment on existing H/W • 1 Java process per active request + ~5 servers + complete OTFR system • Mitigate risk of deploying a new architecture by releasing as soon as possible in parallel with existing system • DADS 10.0 is planned for April 1 release as a beta system while DADS 9.X continues to run under OpenVMS • Will use Sun Fire 280R system (2 CPU) until 15K is available • Off-loading some archive users to beta system will increase headroom in current system Implementation Review
DADS 10.0 Deployment Strategy • Initial release on April 1 • Deploy with its own OTFR system on 280R • Server has the same architecture as the 15K • 2 CPU system will support some small number of concurrent requests • Run in parallel with the current OpenVMS DADS 9.x/OTFR system • S/W, database, and StarView changes already in place to support this mode of operations • A group of beta testers selected to exercise the new system • DADS 9.x will serve as a backup for these users in case of difficulty Implementation Review
Release Objectives • Construct a system test environment • Archive development and test systems today do not mirror operations • Provide a system test bed to resolve liens, fix problems, establish operations procedures before full-up release • Determine optimal full-up release system configuration • Verify functional correctness • Establish operations procedures Implementation Review
Beta Test User Selection • Will use a mix of current archive users • Internal, external users • Measure impact of delivery modes • Exercise priority reassignment • Low-, high-impact users • Tune parameters and controls for fair use of the system when a large dynamic range of request sizes is present • Multiple instruments & missions • Cover the range of available HST data • Exercises different paths through OTFR • FUSE archive users • Archive operators • Evaluate controllability of the system under a variety of scenarios • Exercise failure recovery Implementation Review
Test Plan Overview • Rapid development, build, test, and deploy on the beta test platform • Build cycle will be shortened to address issues quickly • Users will be made aware of the potential for downtime • Delivery of any liens against the system and features planned for later releases • Data depot capability will be added over the course of the test (will be designated DADS 10.1) • Enhanced capabilities not scheduled for the initial release will be added as they are implemented • The beta test platform will shift from the 280R to the 15K as it becomes available • Mitigates risk associated with the H/W platform in transitioning to the full operational release (DADS 10.2) Implementation Review
Beta Test Completion Criteria • 15K must be available before new system assumes full archive distribution load • System performance and loading characteristics during the test must be well understood • Will determine exact runtime configuration of full operational release • Support for data depot must be available • Archive operators must concur that the system is ready Implementation Review
Schedule Implementation Review