diff --git a/python/pyspark/rdd.py b/python/pyspark/rdd.py
index cd1f64e8aaa48..8978f028c5928 100644
--- a/python/pyspark/rdd.py
+++ b/python/pyspark/rdd.py
@@ -115,7 +115,7 @@ def _parse_memory(s):
     2048
     """
     units = {'g': 1024, 'm': 1, 't': 1 << 20, 'k': 1.0 / 1024}
-    if s[-1] not in units:
+    if s[-1].lower() not in units:
         raise ValueError("invalid format: " + s)
     return int(float(s[:-1]) * units[s[-1].lower()])
 
diff --git a/python/pyspark/tests.py b/python/pyspark/tests.py
index a5a83c7e38b3c..40fccb8c0090c 100644
--- a/python/pyspark/tests.py
+++ b/python/pyspark/tests.py
@@ -1966,6 +1966,18 @@ def test_startTime(self):
             self.assertGreater(sc.startTime, 0)
 
 
+class ConfTests(unittest.TestCase):
+    def test_memory_conf(self):
+        memoryList = ["1T", "1G", "1M", "1024K"]
+        for memory in memoryList:
+            sc = SparkContext(conf=SparkConf().set("spark.python.worker.memory", memory))
+            l = list(range(1024))
+            random.shuffle(l)
+            rdd = sc.parallelize(l, 4)
+            self.assertEqual(sorted(l), rdd.sortBy(lambda x: x).collect())
+            sc.stop()
+
+
 @unittest.skipIf(not _have_scipy, "SciPy not installed")
 class SciPyTests(PySparkTestCase):