Cloud Infrastructures for In Silico Drug Discovery: Economic and Practical Aspects

Table 1

Characteristics of the operations of the drug discovery pipeline.

	TI	VS	ER	LO
Functional
Number ofcores	≥16	≥8	≥32	≥4
Network	—	—	Infiniband	—
RAM (GB)	≥32	≥16	≥64	≥16
HDD (GB)	—	≥500	≥1000	—
Reference Amazon EC2^1 virtual cluster	20 × D	10 × 2 × D	1–10 × 2 × D	1 × C
Non functional
Security	Low	Medium	Medium	High
Availability/Resiliency	Low	Medium	Medium	Low
Operations parameters
Input data and size	10 K–100 K Parameter simulations	5 K–10 K Compounds	1–10 Simulations 50 K–250 K Atoms; 5–300; Nanoseconds	5–10 Compounds
Service demand time (𝜇𝑘) with respect to the reference Amazon EC2 virtual cluster	16–160 h	8–16 h	2–1200 h	1–6 h
Arrival rate (𝜆𝑘) per year per research group	4	3/4	2/3	1/2
Packages
Commercial	MATLAB	FlexX, Glide, ICM	AMBER, CHARMM	MacroModel, Medstere
Free/open source	R	Dock, Autodock	GROMACS, NAMD	Omega

1See Table 2.