The　densification　of　small　cells　(SCs)　and　caching　at　the　edge　are　two　promising　approaches　to　improve　the　network　throughput,　reduce　end-to-end　delay　and　backhaul　cost　in　future　5G　wireless　networks.　However,　current　analysis　and　design　only　focus　on　the　system　state　at　a　certain　time,　neglecting　the　temporal-spatial　fluctuation　of　traffic　and　the　content　popularity　variations,　which　are　often　affected　by　users＇　behaviors.　In　this　paper,　we　propose　a　novel　social-aware　cache　information　processing　approach,　where　the　social-tie　factor　(STF)　is　modelled　on　the　basis　of　the　data　collected　from　practical　cellular　networks.　Limited　caching　capacity　is　deployed　in　a　few　selected　very　important　base　stations　(VIBSs),　which　have　higher　average　STF　values.　Normal　SCs　are　linked　to　VIBSs　with　limited　microwave　fronthaul.　By　adopting　the　stochastic　geometry　process,　key　performance　indicators,　e.　g.,　throughput,　delay,　energy　efficiency　(EE),　are　all　derived　as　functions　of　transmission　power,　cache　ability,　file　popularity,　density　of　SCs　and　users.　Numerical　results　show　that　the　throughput　of　the　proposed　social-aware　caching　method　is　nearly　250%　larger　than　heterogeneous　network　(HetNet)　with　no　cache　and　48%　larger　than　the　homogeneous　cache,　by　better　utilizing　STF.　The　number　of　selected　VIBSs　can　be　optimized　to　maximize　the　network　throughput　and　EE　under　the　constraint　of　cache　and　backhaul　capacity.　Our　study　provides　insights　into　the　efficient　use　of　cache　utilizing　BS　social　ties　in　ultra-dense　networks　(UDNs).　©　2016　IEEE.