Recently,　small　satellites　have　attracted　much　attention　in　both　academia　and　industry.　These　satellites　can　form　a　swarm　and　communicate　via　Inter-Satellite　Links　(ISLs)　and　help　each　other　download　data　to　one　or　more　Ground　Stations　(GSs).　A　key　challenge　is　that　each　satellite　has　a　short　contact　period　with　different　GSs　over　time.　Henceforth,　this　paper　considers　the　following　objective:　minimize　the　download　time　of　a　given　amount　of　data　from　each　satellite　to　ground　stations.　The　main　challenging　issues　are　that　satellites　have　different　energy　and　buffer　constraints.　We　model　the　problem　as　a　Mixed　Integer　Linear　Program　(MILP).　Its　objective　is　to　minimize　the　total　number　of　time　slots　taken　to　download　a　fixed　amount　of　data　from　each　satellite.　Its　key　decision　variables　relate　to　(i)　routing,　(ii)　link　scheduling,　and　(iii)　data　aggregation　rate　in　each　time　slot.　Our　results　show　that　in-network　data　aggregation　is　able　to　significantly　reduce　the　total　download　time　by　52%　to　84%　when　the　data　aggregation　rate　is　between　10%　to　50%.　©　2018　IEEE.