WISDOM: A Grid-Enabled Drug Discovery Initiative against Malaria - Grid Computing: Infrastructure, Service, and Applications - page 369

• Once the job is i nished, the outputs are stored back on the grid

storage elements via the data management system and the
useful docking results are inserted directly from the grid to a
relational database where they can later be more easily queried
and analyzed.

14.4.3 Data Challenge Deployment

The deployment was performed on several grid infrastructures (Auvergrid
[24], EELA, EGEE, EUChinaGrid) and involved at least one manager to
oversee the process on each of them. The three groups of targets (GST,
P. v iv a x, and P. falciparum DHFR) were docked against the whole ZINC
database (4.3 million ligands). The database was actually cut into 2422
chunks of 1800 ligands each. This splitting was chosen because we wanted
to have an approximated processing time ranging from 20 to 30 hours for
each job (one docking process takes from 40 seconds to 1 minute, depend-
ing on CPU power). The subsets had to be stored on the grid infrastruc-
tures. They were basically copied on a storage element and registered on
the grid i le catalog (LFC) and were also replicated on several locations
whenever possible to improve fault tolerance. We dei ned a WISDOM
instance as being a target structure docked against the whole ZINC data-
base, with a given parameter set. Table 14.3 shows the instances deployed
on the different infrastructures.
A total number of 32 instances were deployed, corresponding to an over-
all workload of 77,504 jobs, and up to 140 million docking operations.
As shown in Figure 14.5 , the environment included a FlexLm server that
provided the l oating licenses for the FlexX commercial software. During
the i rst WISDOM deployment in 2005, the license server was identii ed as
a potential bottleneck and point of failure because we had just one server

TABLE 14.3
Instances Deployed on the Different Infrastructures during the
WISDOM-II Data Challenge
Target Structures Number of Instances Deployed

GST (A chain) 4 on EGEE

GST (B chain) 4 on EGEE

2BL9 ( P. vivax wild-type DHFR) 3 on EGEE, 1 on EELA

2BLC ( P. vivax double mutant DHFR) 3 on EGEE, 1 on AuverGrid

Dm_vivax ( P. vivax DHFR 2BLC minimized) 4 on EGEE

Wt_vivax ( P. vivax DHFR 2BL9 minimized) 4 on EGEE

1J3K ( P. falciparum quadruple mutant DHFR) 4 on EGEE

1J3I ( P. falciparum wild-type DHFR) 3 on EGEE, 1 on EuChinaGRID