© GMV, 2011
Cloudy Inside:
Use of Cloud Computing
in Ground Systems
Development
Motivation
Business considerations for a Ground System COTS vendor
– Mission Critical applications
• Any deployment, even a bug fix, must be tested in an integrated
environment
– Multiple Projects
• The COTS model depends on having multiple customers to drive product
evolution with new license purchases and support
• Every customer’s concept of operations, and technical requirements, are
different
– Multiple Products
• To be competitive the product suite must cover several application areas
– Long support horizons
• Customers are on separate upgrade cycles
• Long periods of inactivity are punctuated by urgent requests
Bottom line:
GMV‘s Footprint
SCC: Satellite Control Centre (real time monitoring & control)
FDS / MPS: Flight dynamics / Mission Planning System
MA: Mission analysis
PMR: Payload management & reconfiguration
PDS/SOC/GMS: Payload Data Segment / Science Operations Centre, Ground Mission Segment
MEASAT 1 MEASAT 3 MEASAT 1R GALILEO IOV-1 GALILEO IOV-2 GALILEO IOV-3 GALILEO IOV-4 Galileo constellation EPS/METOP ERS-1/2 CRYOSAT GOCE ATV COSMO/Skymed FUEGO Envisat Helios SMOS ADM/Aeolus Earthcare EGPM ERM Minisob Smila Spectra SPOT Swarm Wales SEOSAT OCO BepiColombo Darwin Exomars Gaia Hesrchel/Planck LISA Venus Express Cluster Hipparcos Huygens Integral Mars Express Rosetta SMART-1 Soho COROT SCC & FDS / MPS FDS / MPS SCC MA PMR PDS/SOC/GMS XMM ISO LRO IS-IR IS-3R Intelsat 6B IS-9 Globalstar constelation Galaxy 10R Galaxy 11 Galaxy 13 Galaxy 3C Galaxy 4R SBS 6 DirecTV 10 DirecTV 11 DirecTV 1R DirecTV 4S Brasilesat B1 Brasilesat B2 Brasilesat B3 Brasilesat B4 Amazonas 1 Amazonas 2 Amazonas 3 Thor VI / W2A / W2M /Hot Bird 6 /Hot Bird 7 /Hot Bird 9 /Hot Bird 10 /Astra 3B KASAT Intelsat 5 Intelsat 10 NSS12 /Eurobird 4/ Eurobird 9 / Arabsat 4C / Arabsat 4C / Arabsat 4C / W5 / W6 / W7 Glory ACE SEOSAR 2 LEASAT 10 Optus C1 Optus D3 Intelsat 2 Giotto
Mars Sample Return Mars Next Ulyses Proba 3 LRO JWST Brasilesat C3 Hispasat 1E LDCM GLORY CRYOSAT Nilesat 201 Azersat/Africasat-1a MEASAT 3B
Example Configuration
Three simultaneous projects
– A (products X & Y, Red Hat Linux 4 32 bit) – B (products X & Z, Red Hat Linux 4 64 bit) – C (products Y & Z, SUSE Linux 10 32 bit)
Three versions per project
– Deployed
– Acceptance testing
– Next release
Three machines per version
– Build
– Server
– Client
One Year Later
Project A has entered long-term
maintenance
Project B is upgrading to SUSE
Linux 11 64 bit
– Deployed version will remain on
SUSE 10 for one more year
Project C has added redundancy
requirements and will require a
second server
Mismatch
The numbers of required environments increases monotonically
– Projects stay in support for a very long time
Computer resource requirements are proportional to current
workload
– Developers and testers transition to new projects as old ones ramp
down (i.e. team size is relatively constant)
– New software versions tend to require more resources than old
Space, power and money are scarce resources
– Filling the office space with computers is not sustainable
– Constantly reconfiguring the existing computers is not efficient
How do we scale the computing resources with the
workload, and not with the environment count?
Enter the cloud
Enabling factors
– All configurations run on compatible hardware
• 64 bit Intel processors (Core 2 and later)
• New hardware is backwards compatible (i.e. old instruction sets supported)
– Available hardware exceeds required performance for all roles
• Installable memory
• Number of processor cores
– Hardware price/performance grows as fast as the workload
• Lab workload grows with the requirements of new software versions, not
with the number of projects
Approach:
– Size hardware for the current workload
– Configure a set of virtual machines for each project/version
• Includes simulators for spacecraft and network link delays
– Run the appropriate sets of virtual machines for the current work
Lab Concept
Running Environments Stored EnvironmentsCloud
Results
•
Assumptions
• 4 new projects/year, 1 year duration
• Old projects require 1 month
maintenance every year
• 3 machines/project
• 2 GB + 20%/year memory requirement per machine
• Memory capacity doubles every two years (Moore’s law)
• Individual machine $700 (2GB), Cloud server $5000 (16 GB) 0 10000 20000 30000 40000 50000 60000 70000 80000 1 2 3 4 5 6 7 8 9
Cumulative Cost
Cloud cost Machine cost 0 50 100 150 200 250 300 1 2 3 4 5 6 7 8 9 memo ry (G B)
Moore's Law Wins
required (new & maint)
Conclusion
Ground Systems development has characteristics that cause
proliferation of lab environments (build and test)
– High degree of project-specific customization
– Long-lived deployments
Cloud computing enables efficient reuse of hardware
– Similarly to how the team’s resources are allocated to projects
according to their level of activity
– Saves space, power and money
