The　architecture　and　fundamental　components　of　information　transmission-based　flexible　remote　diagnosing　mode　are　discussed.　The　implementation　of　the　embedded　intelligent　agent　model　for　the　flexible　remote　diagnosing　system　is　presented.　The　system　makes　use　of　ARM　+　DSP　master　slave　structure　and　embedded　LINUX.　Key　techniques　related　to　the　performance　degradation　prediction　and　the　embedded　intelligent　agent　are　also　discussed.　Using　the　system　based　on　the　proposed　model　results　in　near-zero-downtime　performance　for　devices/products.