TODO

[DONE] 1. Deploy linux kernel 3.0 (or above) on gr3.

[DONE] 2. Code review for GreenDM. 
		- basic comments [DONE]
    	- add special mechanism for disk initialization [DONE]

[DONE] 3. Documentation for software stack. 

[DONE] 4. Review on the cache design and implementation.

5. Debug on virtual device when it is full by using RAM as temp 
   storage for swapping between extents. 
	- The reason is that eviction needs addtional temp space

[DONE] 6. Kernel control of disk spin down/up

7. Consult on translating workloads to dataseries. 
	- extent vs. sector tradeoff

8. Involved issues 
	- metadata management by DM kernel API (meet with Vasily, Mandar,
	  Atul, Ming, Rajesh) [Half-DONE]
	- Evolve to kernel version 3.3 from 3.0 because of new DM features
	- Consistency issue doing data migration and updating the mapping
	  table [Transaction Manager?]
	- disks reliability concern, like how the user specifies the
	  tradeoff between power, performance, and reliability [NOT CSE595]

9. Evaluation of different workloads (web server, file server,
   database server, email server, etc) with three different approaches
	- Naive sectors mapping without knowledge about workload [CSE595]
	- Design test scenario by "dd" [CSE595]
	- Offline trace analysis based [NOT CSE595]
	- Online trace analysis based  [NOT CSE595]

10. Optimization. [NOT CSE595 <-- Time limit]
	- IO by-passing
	- sequential pre-fetch by trace analysis
	- reliability optimization with reliability aware policy
	- extent overhead optimization
	- explore workloads with different IO patterns 
	- Two levels Cache (RAM)
	- Online Trace study with more aggressive domain knowledge
	- etc. 

11. Report submission[CSE595]
	- By May 1, Ming and Rajesh should write a report summarizing 
	  what has been done so far. 

12. Paper submission[NOT CSE595]
	- Tweak for better results. And write one potential paper in a 
	  perspective different from the report, carefully distinguishing 
	  from several other approaches with more domain knowledge,
	  optimization, and tradeoffs between power, performance, and 
	  reliability.