diff --git a/src/core/compact_object.cc b/src/core/compact_object.cc
index c61506b28..82f80f8b4 100644
--- a/src/core/compact_object.cc
+++ b/src/core/compact_object.cc
@@ -205,7 +205,11 @@ struct TL {
 thread_local TL tl;
 
 constexpr bool kUseSmallStrings = true;
-constexpr bool kUseAsciiEncoding = true;
+
+/// TODO: Ascii encoding becomes slow for large blobs. We should factor it out into a separate
+/// file and implement with SIMD instructions.
+constexpr bool kUseAsciiEncoding = false;
+
 }  // namespace
 
 static_assert(sizeof(CompactObj) == 18);
diff --git a/src/core/segment_allocator.h b/src/core/segment_allocator.h
index b252e986b..05e89a9ca 100644
--- a/src/core/segment_allocator.h
+++ b/src/core/segment_allocator.h
@@ -63,8 +63,12 @@ class SegmentAllocator {
 };
 
 inline auto SegmentAllocator::Allocate(uint32_t size) -> std::pair<Ptr, uint8_t*> {
-  uint64_t ptr = (uint64_t)mi_heap_malloc(heap_, size);
-  uint64_t seg_ptr = ptr & kSegmentAlignMask;
+  void* ptr = mi_heap_malloc(heap_, size);
+  if (!ptr)
+    throw std::bad_alloc{};
+
+  uint64_t iptr = (uint64_t)ptr;
+  uint64_t seg_ptr = iptr & kSegmentAlignMask;
 
   // could be speed up using last used seg_ptr.
   auto [it, inserted] = rev_indx_.emplace(seg_ptr, address_table_.size());
@@ -73,7 +77,7 @@ inline auto SegmentAllocator::Allocate(uint32_t size) -> std::pair<Ptr, uint8_t*
     address_table_.push_back((uint8_t*)seg_ptr);
   }
 
-  Ptr res = (((ptr - seg_ptr) / 8) << kSegmentIdBits) | it->second;
+  Ptr res = (((iptr - seg_ptr) / 8) << kSegmentIdBits) | it->second;
   used_ += mi_good_size(size);
 
   return std::make_pair(res, (uint8_t*)ptr);
diff --git a/src/facade/op_status.h b/src/facade/op_status.h
index df5ea4e1d..cf1f9312a 100644
--- a/src/facade/op_status.h
+++ b/src/facade/op_status.h
@@ -17,6 +17,7 @@ enum class OpStatus : uint16_t {
   OUT_OF_RANGE,
   WRONG_TYPE,
   TIMED_OUT,
+  OUT_OF_MEMORY,
 };
 
 class OpResultBase {
diff --git a/src/server/dragonfly_test.cc b/src/server/dragonfly_test.cc
index 4279a6fd2..2ba45e3d0 100644
--- a/src/server/dragonfly_test.cc
+++ b/src/server/dragonfly_test.cc
@@ -325,6 +325,15 @@ TEST_F(DflyEngineTest, Memcache) {
   EXPECT_THAT(resp, ElementsAre("END"));
 }
 
+TEST_F(DflyEngineTest, LimitMemory) {
+  mi_option_enable(mi_option_limit_os_alloc);
+  string blob(1 << 15, 'a');
+  for (size_t i = 0; i < 1000; ++i) {
+    auto resp = Run({"set", absl::StrCat(blob, i), blob});
+    ASSERT_THAT(resp, RespEq("OK"));
+  }
+}
+
 // TODO: to test transactions with a single shard since then all transactions become local.
 // To consider having a parameter in dragonfly engine controlling number of shards
 // unconditionally from number of cpus. TO TEST BLPOP under multi for single/multi argument case.
diff --git a/src/server/transaction.cc b/src/server/transaction.cc
index 823dcfbc7..e5b3f8447 100644
--- a/src/server/transaction.cc
+++ b/src/server/transaction.cc
@@ -24,7 +24,6 @@ std::atomic_uint64_t op_seq{1};
 
 [[maybe_unused]] constexpr size_t kTransSize = sizeof(Transaction);
 
-
 }  // namespace
 
 struct Transaction::FindFirstProcessor {
@@ -359,16 +358,25 @@ bool Transaction::RunInShard(EngineShard* shard) {
 
   /*************************************************************************/
   // Actually running the callback.
-  OpStatus status = cb_(this, shard);
-  /*************************************************************************/
+  try {
+    OpStatus status = cb_(this, shard);
 
-  if (unique_shard_cnt_ == 1) {
-    cb_ = nullptr;  // We can do it because only a single thread runs the callback.
-    local_result_ = status;
-  } else {
-    CHECK_EQ(OpStatus::OK, status);
+    if (unique_shard_cnt_ == 1) {
+      cb_ = nullptr;  // We can do it because only a single thread runs the callback.
+      local_result_ = status;
+    } else {
+      CHECK_EQ(OpStatus::OK, status);
+    }
+  } catch (std::bad_alloc&) {
+    // TODO: to log at most once per sec.
+    LOG_FIRST_N(ERROR, 16) << " out of memory";
+    local_result_ = OpStatus::OUT_OF_MEMORY;
+  } catch (std::exception& e) {
+    LOG(FATAL) << "Unexpected exception " << e.what();
   }
 
+  /*************************************************************************/
+
   // at least the coordinator thread owns the reference.
   DCHECK_GE(use_count(), 1u);
 
@@ -532,7 +540,7 @@ OpStatus Transaction::ScheduleSingleHop(RunnableType cb) {
   // single hop -> concluding.
   coordinator_state_ |= (COORD_EXEC | COORD_EXEC_CONCLUDING);
 
-  if (!multi_) {   // for non-multi transactions we schedule exactly once.
+  if (!multi_) {  // for non-multi transactions we schedule exactly once.
     DCHECK_EQ(0, coordinator_state_ & COORD_SCHED);
   }
 
@@ -765,7 +773,15 @@ void Transaction::RunQuickie(EngineShard* shard) {
   DVLOG(1) << "RunQuickSingle " << DebugId() << " " << shard->shard_id() << " " << args_[0];
   CHECK(cb_) << DebugId() << " " << shard->shard_id() << " " << args_[0];
 
-  local_result_ = cb_(this, shard);
+  // Calling the callback in somewhat safe way
+  try {
+    local_result_ = cb_(this, shard);
+  } catch (std::bad_alloc&) {
+    LOG_FIRST_N(ERROR, 16) << " out of memory";
+    local_result_ = OpStatus::OUT_OF_MEMORY;
+  } catch (std::exception& e) {
+    LOG(FATAL) << "Unexpected exception " << e.what();
+  }
 
   sd.local_mask &= ~ARMED;
   cb_ = nullptr;  // We can do it because only a single shard runs the callback.