A Ceph-based storage strategy for big gridded remote sensing data

Figures & data

Figure 1. A typical rowkey design.

Figure 2. An example of ideal query and degraded query.

Figure 3. An example of an I/O process.

Table

Algorithm 1 A pseudo-code example of CRUSH algorithm

Input: ObjcetID Output: OSDList:[OSD1, OSD2, OSD3,..] result = empty_list for num_reps do bucket = take; item = random_pick_from_list(ObjcetID, in bucket); if item.type not failure_domain then in_bucket.append(item); if item not in result then result[rep] = item; end if end if end for return result

Figure 4. The processes of metadata query.

Table 1. Environment parameters.

Parameters	Value
Number of nodes	3
Operating System	Ubuntu16.04
Memory	32 G
Number of CPU	4
Hadoop Version	3.1.2
HBase Version	2.2.2
Ceph Version	10.2.11

Figure 5. Creation time differences between nodes in the different storage strategy scenarios.

Figure 6. Inquiry time differences between nodes in the different storage strategy scenarios.

Data availability statement

Data used in this paper are available upon request from the corresponding author. http://gaofenplatform.com/.