Estimation of resource for OSSEs

gsf   T382

 

Moorthi Dew/Mist

104 task, 13  node Th=2

8.5min /day

60min/ 7 day

gfs T382

3 hourly dump

Jim Jung Vapor

16 task  1 node  Th=4

20min for 6hr

 

gsf   T382

Jim Jung Vapor

16 task   1 node   Th=4

3.5 hour/7day 

gfs   T382

reanl Dew

224 task, 14 nodes, 1 thread

20 min/9hr

4hr20min/5day

(6 hr/7 day)

gfs   T126

reanl Dew

128 task 4 node Th=1

25 min /5day

35 min/7day

gfs  T62

reanl Dew

16 task 1 node Th=1

1 min/9 hour

9 min/5day

13min/7day

GSI T382

 

Operational analysis

Russ Treadon

Dew

160 task 10 nodes (16 tasks per node)

th=1

average run time:  1146.446  seconds
minimum run time:  1098.890 seconds
maximum run time:  1197.175 seconds

max-min:  98.285 seconds
standard deviation:  25.854 seconds

20 min/cycle

 

80min/day

GSI T382

Jim Jung  haze

64 task 4 node  TH=1

one hour/cycle

4 hr/day

GSI  T382

Jim Jung Vapor

32 task one node

real cpu only

Airs and IASI

one hour/cycle

4 hr/day

GSI T382

reanl Dew

224 task 14 nodes  Th=1

25 min/cycle

2hr/day

GSI T62

reanl Dew

16 task 1 node TH=1

10 min/cycle

40min/day

Simulation  of Radiance

Tong Zhu

Haze

one cpu

GOES-E/W, AMSU

 

0.5  hour/day

Simulation of Radiance

Ron

one cpu

HIRS, AMSU, AIRS

no GOES

3 hour/day

 

 

 

 

 

 

 


 

 

 

 

 

[Summary of estimate on vapor]

In general vapor is 30-50% faster than haze.

 

Rough estimate to process one day  of assimilation with two five day forecast

 

T62

one node

GSI 30min/day  (4 cycle)

two  5day fcst  15 min

 

45 min/day   one node

 

 T126

2 node

GSI  40min/day

two  5day fcst 20min  (rough guess from computation with 4 node)

 

1 hr/day   2 node

 

T382

one node

GSI  4 hr/day (4cycle)

one node

two 5day fcst  5 hour

 

9 hr/day   with one node

 

 

 

T62

Number of experiments.   100 exp for   one  year  =36500  day

                            1(node)x0.75(hr/day)x36500(day)=27400 hr*node=1142 day*node

            analysis only         200 exp  for one year    =73000  day

                            1(node)x0.5(hr/day)x73000 (day)=36500hr*node=1520 day*node

                                                                                                     

 

 

T126

Number of experiments.   100 exp for 6 week  =8400 day

                        2(node)x1(hr/day)x8400(day)= 32000 hr*node=1333 day*node

 

                                          20 experiments for one year=7300 day

                        2(node)x1(hr/day)x16000(day)= 32000 hr*node=1333 day*node

T382

100 exp for 6weeks=4200days

1 nodes  9 (hr/day)x4200 day =  37800 hr*node=1575day*node                                                                                                

 

 

200 exp   6wk with T126

20 exp  one year with T126

20 exp   6wk with  t382

6 hourly cycle

fcst twice a day

 

2 node    for 860 days 

do not include  resource for debug

 

Simulation

existing radiance data

365x3 hr  (existing + future=1095 hr x 1 task

                    34 hrx 64 task

 

If coefficient has to be calculated, it will take more time.

 

Time for simulation is negligible.

Each day produce 1GBfile

 

Run history of 15000day    15 TB

 

 

GOES-R

20km 10 times thining ->  60 km resolution  : 15MB/hr  360MB/day=150GB/year

20km                                                                                  3.6GB/day=1.5 TB/year

4km                                                                                     3.6x25=90GB/day=33TB/year

 

GOES-R CrIS, ATMS

We have tom mix  4km 20km and 60km resolution

 

 

50TB hpss space to save the simulated data for new satellite

50TB hpss space to save results

10TB disk space