Profile petascope, rasj for excessive RAM usage

[Dimitar Misev]

It seems that petascope uses more memory than it theoretically should. The following query selects a 1GB PNG from rasdaman:

for b4 in(s2a_l2a_t32tps_B04_10m), b8 in (s2a_l2a_t32tps_B08_10m)
return encode(
 (b8[ansi("2016-11-08T15:47:44.000Z")] - b4[ansi("2016-11-08T15:47:44.000Z")]) / 
 (b8[ansi("2016-11-08T15:47:44.000Z")] + b4[ansi("2016-11-08T15:47:44.000Z")])
,"png")

however, Tomcat consumes at least 4GB, even though petascope needs to transfer only 1GB in total from rasdaman to the client.

[Bang Pham Huu]

The problem is not from Petascope as Rasserver takes 4.5 GB Ram (monitored by htop and System Monitor) and the result returns only 40 MB.

the coverages were imported with 1 band for 1 coverage (band04: 113.6 MB and band08: 135.3 MB) JPG2 and the WCPS query is

for c in (sentinel2_10m_B04), d in (sentinel2_10m_B08)
return encode(
 (c - d) / 
 (c + d)
,"png")

and Rasql query is

SELECT encode( ( c - d )  /  ( c + d ) , "PNG" , "{\"geoReference\":{\"crs\":\"EPSG:32633\",\"bbox\":{\"xmin\":300000.0,\"ymin\":5290200.0,\"xmax\":409800.0,\"ymax\":5400000.0}}}") FROM sentinel2_10m_B04 AS c,sentinel2_10m_B08 AS d

[Bang Pham Huu]

There is a problem with memory usage in rasj library for Tomcat which occupies the byte array multiple times from the rasql result (e.g: a rasql returns 600 MB Tiff file, Tomcat will need to set heap maximum memory at least 2.4 GB). It will need to remove the initialization unnecessarily in rasj for storing result.