diff --git a/primitives/primproc/primproc.cpp b/primitives/primproc/primproc.cpp
index 672659d1d..ad3459081 100644
--- a/primitives/primproc/primproc.cpp
+++ b/primitives/primproc/primproc.cpp
@@ -220,7 +220,7 @@ class QszMonThd
 };
 #endif
 
-#define DUMP_CACHE_CONTENTS
+// #define DUMP_CACHE_CONTENTS
 #ifdef DUMP_CACHE_CONTENTS
 void* waitForSIGUSR1(void* p)
 {
diff --git a/tests/counting_allocator.cpp b/tests/counting_allocator.cpp
index 94d69690d..c3f3571d9 100644
--- a/tests/counting_allocator.cpp
+++ b/tests/counting_allocator.cpp
@@ -17,25 +17,28 @@
 #include <gtest/gtest.h>
 #include <atomic>
 #include <cstddef>
+#include <cstdint>
 #include <memory>
 #include <thread>
 
+#include <boost/smart_ptr/allocate_shared_array.hpp>
+
 #include "countingallocator.h"
-#include "rowgroup.h"
 
 using namespace allocators;
 
 // Example class to be managed by the allocator
 struct TestClass
 {
-  int value;
+  int value[1024];
 
   TestClass(int val) : value(val)
   {
   }
 };
 
-static const constexpr int64_t MemoryAllowance = 10 * 1024 * 1024;
+static const constexpr int64_t MemoryAllowance = 1 * 1024 * 1024;
+static const constexpr int64_t MemoryLimitStep = MemoryAllowance / 100;
 
 // Test Fixture for AtomicCounterAllocator
 class CountingAllocatorTest : public ::testing::Test
@@ -49,7 +52,8 @@ class CountingAllocatorTest : public ::testing::Test
 
   // Constructor
   CountingAllocatorTest()
-   : allocatedMemory(MemoryAllowance), allocator(&allocatedMemory, MemoryAllowance / 100)
+   : allocatedMemory(MemoryAllowance)
+   , allocator(&allocatedMemory, MemoryAllowance / 100, MemoryAllowance / 100)
   {
   }
 
@@ -63,7 +67,15 @@ TEST_F(CountingAllocatorTest, Allocation)
   const std::size_t numObjects = 5;
   TestClass* ptr = allocator.allocate(numObjects);
   EXPECT_NE(ptr, nullptr);
-  EXPECT_EQ(allocatedMemory.load(), MemoryAllowance - numObjects * static_cast<int64_t>(sizeof(TestClass)));
+  if (MemoryLimitStep > numObjects * static_cast<int64_t>(sizeof(TestClass)))
+  {
+    EXPECT_EQ(allocatedMemory.load() - allocator.getCurrentLocalMemoryUsage(),
+              MemoryAllowance - numObjects * static_cast<int64_t>(sizeof(TestClass)));
+  }
+  else
+  {
+    EXPECT_EQ(allocatedMemory.load(), MemoryAllowance - numObjects * static_cast<int64_t>(sizeof(TestClass)));
+  }
   allocator.deallocate(ptr, numObjects);
 }
 
@@ -72,8 +84,15 @@ TEST_F(CountingAllocatorTest, Deallocation)
 {
   const std::size_t numObjects = 3;
   TestClass* ptr = allocator.allocate(numObjects);
-  EXPECT_EQ(allocatedMemory.load(), MemoryAllowance - numObjects * static_cast<int64_t>(sizeof(TestClass)));
-
+  if (MemoryLimitStep > numObjects * static_cast<int64_t>(sizeof(TestClass)))
+  {
+    EXPECT_EQ(allocatedMemory.load() - allocator.getCurrentLocalMemoryUsage(),
+              MemoryAllowance - numObjects * static_cast<int64_t>(sizeof(TestClass)));
+  }
+  else
+  {
+    EXPECT_EQ(allocatedMemory.load(), MemoryAllowance - numObjects * static_cast<int64_t>(sizeof(TestClass)));
+  }
   allocator.deallocate(ptr, numObjects);
   EXPECT_EQ(allocatedMemory.load(), MemoryAllowance);
 }
@@ -94,34 +113,42 @@ TEST_F(CountingAllocatorTest, AllocatorEquality)
 TEST_F(CountingAllocatorTest, AllocateSharedUsesAllocator)
 {
   // Create a shared_ptr using allocate_shared with the custom allocator
-  std::shared_ptr<TestClass> ptr = std::allocate_shared<TestClass>(allocator, 100);
+  CountingAllocator<TestClass> allocatorSmallerStep(&allocatedMemory, MemoryAllowance / 100,
+                                                    MemoryAllowance / 1000);
+  std::shared_ptr<TestClass> ptr1 = std::allocate_shared<TestClass>(allocatorSmallerStep, 100);
+  std::shared_ptr<TestClass> ptr2 = std::allocate_shared<TestClass>(allocatorSmallerStep, 100);
+  std::shared_ptr<TestClass> ptr3 = std::allocate_shared<TestClass>(allocatorSmallerStep, 100);
 
   // Check that the counter has increased by the size of TestClass plus control block
   // Exact size depends on the implementation, so we verify it's at least sizeof(TestClass)
-  EXPECT_LE(allocatedMemory.load(), MemoryAllowance - static_cast<int64_t>(sizeof(TestClass)));
+  EXPECT_LE(allocatedMemory.load(), MemoryAllowance - 3 * static_cast<int64_t>(sizeof(TestClass)));
 
   // Reset the shared_ptr and check that the counter decreases appropriately
-  ptr.reset();
+  ptr1.reset();
+  ptr2.reset();
+  ptr3.reset();
   // After deallocation, the counter should return to zero
   EXPECT_EQ(allocatedMemory.load(), MemoryAllowance);
 
-  auto deleter = [this](TestClass* ptr) { this->allocator.deallocate(ptr, 1); };
-  ptr.reset(allocator.allocate(1), deleter);
-  EXPECT_LE(allocatedMemory.load(), MemoryAllowance - static_cast<int64_t>(sizeof(TestClass)));
+  // auto deleter = [&allocatorSmallerStep](TestClass* ptr) { allocatorSmallerStep.deallocate(ptr, 1); };
+  // ptr1.reset(allocatorSmallerStep.allocate(3), deleter);
+  // EXPECT_LE(allocatedMemory.load(), MemoryAllowance - static_cast<int64_t>(sizeof(TestClass)));
 
-  ptr.reset();
+  ptr1.reset();
   EXPECT_EQ(allocatedMemory.load(), MemoryAllowance);
 
-  size_t allocSize = 16ULL * rowgroup::rgCommonSize;
-  auto buf = boost::allocate_shared<rowgroup::RGDataBufType>(allocator, allocSize);
+  using RGDataBufType = uint8_t[];
+  size_t allocSize = 16ULL * 8192;
+  auto buf = boost::allocate_shared<RGDataBufType>(allocator, allocSize);
+
   EXPECT_LE(allocatedMemory.load(), MemoryAllowance - allocSize);
 
   buf.reset();
   EXPECT_EQ(allocatedMemory.load(), MemoryAllowance);
 
-  CountingAllocator<rowgroup::RGDataBufType> allocator1(&allocatedMemory, MemoryAllowance / 100);
-  std::optional<CountingAllocator<rowgroup::RGDataBufType>> allocator2(allocator1);
-  auto buf1 = boost::allocate_shared<rowgroup::RGDataBufType>(*allocator2, allocSize);
+  CountingAllocator<RGDataBufType> allocator1(&allocatedMemory, MemoryAllowance / 100, MemoryAllowance / 100);
+  std::optional<CountingAllocator<RGDataBufType>> allocator2(allocator1);
+  auto buf1 = boost::allocate_shared<RGDataBufType>(*allocator2, allocSize);
   EXPECT_LE(allocatedMemory.load(), MemoryAllowance - allocSize);
 
   buf1.reset();
@@ -136,11 +163,27 @@ TEST_F(CountingAllocatorTest, ThreadSafety)
 
   auto worker = [this]()
   {
+    std::vector<TestClass*> ptrs;
+    CountingAllocator<TestClass> allocatorLocal(&allocatedMemory, MemoryAllowance / 100,
+                                                MemoryAllowance / 1000);
     for (std::size_t i = 0; i < allocationsPerThread; ++i)
     {
-      TestClass* ptr = allocator.allocate(1);
-      allocator.deallocate(ptr, 1);
+      ptrs.push_back(allocatorLocal.allocate(1));
     }
+
+    int64_t usedMemory = allocationsPerThread * sizeof(TestClass);
+    EXPECT_EQ(allocatorLocal.getCurrentLocalMemoryUsage(), allocationsPerThread * sizeof(TestClass));
+    EXPECT_GE(usedMemory - allocatorLocal.getlastMemoryLimitCheckpoint(), 0LL);
+    EXPECT_LE(allocatedMemory.load(), MemoryAllowance - allocatorLocal.getlastMemoryLimitCheckpoint());
+
+    for (auto* ptr : ptrs)
+    {
+      allocatorLocal.deallocate(ptr, 1);
+    }
+
+    EXPECT_EQ(allocatorLocal.getCurrentLocalMemoryUsage(), 0);
+    EXPECT_EQ(allocatorLocal.getlastMemoryLimitCheckpoint(), 0);
+    EXPECT_GE(allocatedMemory.load(), allocationsPerThread * sizeof(TestClass));
   };
 
   std::vector<std::thread> threads;
@@ -156,7 +199,7 @@ TEST_F(CountingAllocatorTest, ThreadSafety)
     th.join();
   }
 
-  // After all allocations and deallocations, the counter should be zero
+  // After all allocations and deallocations, the counter should be zero minus the remainder
   EXPECT_EQ(allocatedMemory.load(), MemoryAllowance);
 }
 
@@ -172,8 +215,8 @@ TEST_F(CountingAllocatorTest, AllocateZeroObjects)
 
 TEST_F(CountingAllocatorTest, CopyAssignable)
 {
-    CountingAllocator<TestClass> allocator1(&allocatedMemory);
-    CountingAllocator<TestClass> allocator2(&allocatedMemory);
-    allocator1 = allocator2;
-    EXPECT_EQ(allocator1, allocator2);
+  CountingAllocator<TestClass> allocator1(&allocatedMemory);
+  CountingAllocator<TestClass> allocator2(&allocatedMemory);
+  allocator1 = allocator2;
+  EXPECT_EQ(allocator1, allocator2);
 }
\ No newline at end of file
diff --git a/utils/common/countingallocator.h b/utils/common/countingallocator.h
index 4d9a36f6f..d0c140ee8 100644
--- a/utils/common/countingallocator.h
+++ b/utils/common/countingallocator.h
@@ -17,99 +17,217 @@
 
 #pragma once
 
+#include <cassert>
 #include <cstdint>
-#include <limits>
-#include <memory>
 #include <atomic>
 #include <cstddef>
-#include <iostream>
-#include <utility>
+#include <cstdlib>
 
-namespace allocators 
+#include <iostream>
+
+namespace allocators
 {
 
-// const constexpr std::uint64_t CounterUpdateUnitSize = 4 * 1024 * 1024; 
-const constexpr std::int64_t MemoryLimitLowerBound = 100 * 1024 * 1024;  // WIP
+// WIP placement
+
+
+// const constexpr std::uint64_t CounterUpdateUnitSize = 4 * 1024 * 1024;
+const constexpr std::int64_t MemoryLimitLowerBound = 500 * 1024 * 1024;  // WIP
+const constexpr std::int64_t CheckPointStepSize = 100 * 1024 * 1024;     // WIP
 
 // Custom Allocator that tracks allocated memory using an atomic counter
 template <typename T>
-class CountingAllocator {
-public:
-    using value_type = T;
+class CountingAllocator
+{
+ public:
+  using value_type = T;
 
-    // Constructor accepting a reference to an atomic counter
-    explicit CountingAllocator(std::atomic<int64_t>* memoryLimit, const uint64_t lowerBound = MemoryLimitLowerBound) noexcept
-        : memoryLimit_(memoryLimit), memoryLimitLowerBound(lowerBound) {}
+  bool needCheckPoint(const int64_t sizeChange, const int64_t diffSinceLastCheckPoint,
+                      const int64_t checkPointStepSize)
+  {
+    return std::llabs(sizeChange + diffSinceLastCheckPoint) > checkPointStepSize;
+  }
 
-    // Copy constructor (template to allow conversion between different types)
-    template <typename U>
-    CountingAllocator(const CountingAllocator<U>& other) noexcept
-        : memoryLimit_(other.memoryLimit_),  memoryLimitLowerBound(other.memoryLimitLowerBound) {}
+  int64_t int_distance(const int64_t x, const int64_t y)
+  {
+    return (x > y) ? x - y : y - x;
+  }
 
-    // Allocate memory for n objects of type T
-    template <typename U = T>
-    typename std::enable_if<!std::is_array<U>::value, U*>::type
-    allocate(std::size_t n) 
+  // INVARIANT: sizeChange > 0
+  void changeLocalAndGlobalMemoryLimits(const int64_t sizeChange)
+  {
+    // This routine must be used for mem allocation accounting path only!
+    // The case Current > last checkpoint(we deallocated mem since the last checkpoint), sizeIncrease is
+    // negative b/c we now move into the opposite direction. The case Last Checkpoint > Current (we allocated
+    // mem since the last checkpoint), sizeIncrease is positive
+    int64_t sizeChangeWDirection =
+        (currentLocalMemoryUsage_ <= lastMemoryLimitCheckpoint_) ? -sizeChange : sizeChange;
+    int64_t diffSinceLastCheckPoint = int_distance(currentLocalMemoryUsage_, lastMemoryLimitCheckpoint_);
+    if (needCheckPoint(sizeChangeWDirection, diffSinceLastCheckPoint, checkPointStepSize_))
     {
-        auto memCounted = memoryLimit_->fetch_sub(n * sizeof(T), std::memory_order_relaxed);
-        if (memCounted < memoryLimitLowerBound) {
-            memoryLimit_->fetch_add(n * sizeof(T), std::memory_order_relaxed);
-            throw std::bad_alloc();
-        }
-        
-        T* ptr = static_cast<T*>(::operator new(n * sizeof(T)));
-        // std::cout << "[Allocate] " << n * sizeof(T) << " bytes at " << static_cast<void*>(ptr)
-        //           << ". current timit: " << std::dec << memoryLimit_.load() << std::hex << " bytes.\n";
-        // std::cout << std::dec;
-        return ptr;
+      // std::cout << "changeLocalAndGlobalMemoryLimits " << sizeChange << " bytes at "
+      //           << " diffSinceLastCheckPoint " << diffSinceLastCheckPoint << ". current timit: " << std::dec
+      //           << memoryLimit_->load() << std::hex << " bytes.\n";
+      // std::cout << std::dec;
+
+      // auto lastMemoryLimitCheckpointDiff = diffSinceLastCheckPoint + sizeChangeWDirection;
+      int64_t lastMemoryLimitCheckpointDiff = (currentLocalMemoryUsage_ <= lastMemoryLimitCheckpoint_)
+                                                  ? sizeChange - diffSinceLastCheckPoint
+                                                  : sizeChange + diffSinceLastCheckPoint;
+      assert(lastMemoryLimitCheckpointDiff > 0);
+      // {
+      //   std::cout << "[Allocate::changeLocalAndGlobalMemoryLimits!!!] lastMemoryLimitCheckpoint_ "
+      //             << lastMemoryLimitCheckpoint_ << " currentLocalMemoryUsage_ " << currentLocalMemoryUsage_
+      //             << " sizeChangeWDirection " << sizeChangeWDirection << " lastMemoryLimitCheckpointDiff " << lastMemoryLimitCheckpointDiff
+      //             << std::endl;
+      // }
+
+      // lastMemoryLimitCheckpointDiff sign signifies a direction we move allocating memory.
+      auto currentGlobalMemoryLimit =
+          memoryLimit_->fetch_sub(lastMemoryLimitCheckpointDiff, std::memory_order_relaxed);
+      if (currentGlobalMemoryLimit < memoryLimitLowerBound_)
+      {
+        memoryLimit_->fetch_add(lastMemoryLimitCheckpointDiff, std::memory_order_relaxed);
+        // ? what to do with local counters here
+        throw std::bad_alloc();
+      }
+      lastMemoryLimitCheckpoint_ += lastMemoryLimitCheckpointDiff;
     }
 
-       template <typename U = T>
-    typename std::enable_if<std::is_array<U>::value, typename std::remove_extent<U>::type*>::type
-    allocate(std::size_t n) 
+    currentLocalMemoryUsage_ += sizeChange;
+  }
+
+  // Constructor accepting a reference to an atomic counter
+  explicit CountingAllocator(std::atomic<int64_t>* memoryLimit,
+                             const uint64_t lowerBound = MemoryLimitLowerBound,
+                             const uint64_t checkPointStepSize = CheckPointStepSize) noexcept
+   : memoryLimit_(memoryLimit), memoryLimitLowerBound_(lowerBound), checkPointStepSize_(checkPointStepSize)
+  {
+  }
+
+  // Copy constructor (template to allow conversion between different types)
+  template <typename U>
+  CountingAllocator(const CountingAllocator<U>& other) noexcept
+   : memoryLimit_(other.memoryLimit_)
+   , memoryLimitLowerBound_(other.memoryLimitLowerBound_)
+   , checkPointStepSize_(other.checkPointStepSize_)
+  {
+  }
+
+  // Allocate memory for n objects of type T
+  template <typename U = T>
+  typename std::enable_if<!std::is_array<U>::value, U*>::type allocate(std::size_t n)
+  {
+    auto sizeAllocated = n * sizeof(T);
+
+    changeLocalAndGlobalMemoryLimits(sizeAllocated);
+
+    T* ptr = static_cast<T*>(::operator new(sizeAllocated));
+    // std::cout << "[Allocate] non-array " << n * sizeof(T) << " bytes at " << static_cast<void*>(ptr)
+    //           << ". current timit: " << std::dec << memoryLimit_->load() << std::hex << " bytes.\n";
+    // std::cout << std::dec;
+    return ptr;
+  }
+
+  template <typename U = T>
+  typename std::enable_if<std::is_array<U>::value, typename std::remove_extent<U>::type*>::type allocate(
+      std::size_t n)
+  {
+    auto sizeAllocated = n * sizeof(T);
+
+    changeLocalAndGlobalMemoryLimits(sizeAllocated);
+
+    T ptr = static_cast<T>(::operator new[](n));
+    // std::cout << "[Allocate] array " << n * sizeof(T) << " bytes at " << static_cast<void*>(ptr)
+    //           << ". current timit: " << std::dec << memoryLimit_->load() << std::hex << " bytes.\n";
+    return ptr;
+  }
+
+  // Deallocate memory for n objects of type T
+  void deallocate(T* ptr, std::size_t n) noexcept
+  {
+    ::operator delete(ptr);
+
+    int64_t sizeToDeallocate = n * sizeof(T);
+
+    // std::cout << "[Deallocate start] " << sizeToDeallocate << " bytes from " << static_cast<void*>(ptr)
+    //           << ". current timit: " << std::dec << memoryLimit_->load() << std::hex << " bytes.\n";
+    // std::cout << std::dec;
+
+    int64_t sizeChangeWDirection =
+        (currentLocalMemoryUsage_ >= lastMemoryLimitCheckpoint_) ? -sizeToDeallocate : sizeToDeallocate;
+    int64_t diffSinceLastCheckPoint = int_distance(currentLocalMemoryUsage_, lastMemoryLimitCheckpoint_);
+
+    if (needCheckPoint(sizeChangeWDirection, diffSinceLastCheckPoint, checkPointStepSize_))
     {
-        auto memCounted = memoryLimit_->fetch_sub(n * sizeof(T), std::memory_order_relaxed);
-        if (memCounted < memoryLimitLowerBound) {
-            memoryLimit_->fetch_add(n * sizeof(T), std::memory_order_relaxed);
-            throw std::bad_alloc();
-        }
-        
-        T ptr = static_cast<T>(::operator new[](n));
-        // std::cout << "[Allocate] " << n * sizeof(T) << " bytes at " << static_cast<void*>(ptr)
-        //           << ". current timit: " << std::dec << memoryLimit_.load() << std::hex << " bytes.\n";
-        return ptr;
+      // Invariant is lastMemoryLimitCheckpoint_ >= currentLocalMemoryUsage_ - sizeToDeallocate
+      // and lastMemoryLimitCheckpoint_ value must be negative.
+      // int64_t lastMemoryLimitCheckpointDiff =
+      //     labs(lastMemoryLimitCheckpoint_ - currentLocalMemoryUsage_ - sizeToDeallocate);
+      // auto lastMemoryLimitCheckpointDiff = diffSinceLastCheckPoint + sizeChangeWDirection;
+      int64_t lastMemoryLimitCheckpointDiff =
+          (currentLocalMemoryUsage_ >= lastMemoryLimitCheckpoint_)
+              ? sizeToDeallocate - (currentLocalMemoryUsage_ - lastMemoryLimitCheckpoint_)
+              : diffSinceLastCheckPoint + sizeToDeallocate;
+
+      assert(lastMemoryLimitCheckpointDiff > 0);
+
+      // std::cout << "[Deallocate checkpoint!!!] lastMemoryLimitCheckpoint_ " << lastMemoryLimitCheckpoint_
+      //           << " currentLocalMemoryUsage_ " << currentLocalMemoryUsage_ << " sizeChangeWDirection "
+      //           << sizeChangeWDirection << " lastMemoryLimitCheckpointDiff " << lastMemoryLimitCheckpointDiff
+      //           << std::endl;
+
+      // assert(lastMemoryLimitCheckpointDiff < 0);
+      memoryLimit_->fetch_add(lastMemoryLimitCheckpointDiff, std::memory_order_relaxed);
+
+      lastMemoryLimitCheckpoint_ -= (lastMemoryLimitCheckpoint_ == 0) ? 0 : lastMemoryLimitCheckpointDiff;
     }
+    currentLocalMemoryUsage_ = currentLocalMemoryUsage_ - sizeToDeallocate;
 
-    // Deallocate memory for n objects of type T
-    void deallocate(T* ptr, std::size_t n) noexcept 
-    {
-        ::operator delete(ptr);
-        memoryLimit_->fetch_add(n * sizeof(T), std::memory_order_relaxed);
-        // std::cout << "[Deallocate] " << n * sizeof(T) << " bytes from " << static_cast<void*>(ptr)
-        //           << ". current timit: " << std::dec << memoryLimit_.load() << std::hex << " bytes.\n";
-        // std::cout << std::dec;
-    }
+    // std::cout << "[Deallocate end] " << n * sizeof(T) << " bytes from " << static_cast<void*>(ptr)
+    //           << ". current timit: " << std::dec << memoryLimit_->load() << std::hex << " bytes.\n";
 
-    // Equality operators (allocators are equal if they share the same counter)
-    template <typename U>
-    bool operator==(const CountingAllocator<U>& other) const noexcept 
-    {
-        return memoryLimit_ == other.memoryLimit_;
-    }
+    // std::cout << std::dec;
+  }
 
-    template <typename U>
-    bool operator!=(const CountingAllocator<U>& other) const noexcept 
-    {
-        return !(*this == other);
-    }
+  // Equality operators (allocators are equal if they share the same counter)
+  template <typename U>
+  bool operator==(const CountingAllocator<U>& other) const noexcept
+  {
+    return memoryLimit_ == other.memoryLimit_;
+  }
 
-private:
-    std::atomic<int64_t>* memoryLimit_ = nullptr;
-    int64_t memoryLimitLowerBound = 0;
+  template <typename U>
+  bool operator!=(const CountingAllocator<U>& other) const noexcept
+  {
+    return !(*this == other);
+  }
 
-    // Grant access to other instances of CountingAllocator with different types
-    template <typename U>
-    friend class CountingAllocator;
+  int64_t getMemoryLimitLowerBound() const noexcept
+  {
+    return memoryLimitLowerBound_;
+  }
+
+  int64_t getlastMemoryLimitCheckpoint() const noexcept
+  {
+    return lastMemoryLimitCheckpoint_;
+  }
+
+  int64_t getCurrentLocalMemoryUsage() const noexcept
+  {
+    return currentLocalMemoryUsage_;
+  }
+
+ private:
+  std::atomic<int64_t>* memoryLimit_ = nullptr;
+  int64_t memoryLimitLowerBound_ = MemoryLimitLowerBound;
+  int64_t checkPointStepSize_ = CheckPointStepSize;
+  int64_t lastMemoryLimitCheckpoint_ = 0;
+  int64_t currentLocalMemoryUsage_ = 0;
+
+  // Grant access to other instances of CountingAllocator with different types
+  template <typename U>
+  friend class CountingAllocator;
 };
 
 }  // namespace allocators
\ No newline at end of file
diff --git a/utils/common/stlpoolallocator.h b/utils/common/stlpoolallocator.h
index 0864deb08..8a10c6e93 100644
--- a/utils/common/stlpoolallocator.h
+++ b/utils/common/stlpoolallocator.h
@@ -102,7 +102,7 @@ STLPoolAllocator<T>::STLPoolAllocator(joblist::ResourceManager* rm)
   if (rm) 
   {
     auto alloc = rm->getAllocator<PoolAllocatorBufType>();
-    pa.reset(new PoolAllocator(alloc));
+    pa.reset(new PoolAllocator(alloc, DEFAULT_SIZE));
   }
   else
   {
diff --git a/utils/joiner/tuplejoiner.cpp b/utils/joiner/tuplejoiner.cpp
index 58883244e..c76bcfc0a 100644
--- a/utils/joiner/tuplejoiner.cpp
+++ b/utils/joiner/tuplejoiner.cpp
@@ -67,9 +67,9 @@ TupleJoiner::TupleJoiner(const rowgroup::RowGroup& smallInput, const rowgroup::R
     // _pool.reset(new boost::shared_ptr<PoolAllocator>[bucketCount]);
     for (i = 0; i < bucketCount; i++)
     {
-      STLPoolAllocator<pair<const long double, Row::Pointer>> alloc(resourceManager_);
+      // STLPoolAllocator<pair<const long double, Row::Pointer>> alloc(resourceManager_);
       // _pool[i] = alloc.getPoolAllocator();
-      // auto alloc = resourceManager_->getAllocator<pair<const long double, Row::Pointer>>();
+      auto alloc = resourceManager_->getAllocator<pair<const long double, Row::Pointer>>();
       ld[i].reset(new ldhash_t(10, hasher(), ldhash_t::key_equal(), alloc));
     }
   }
@@ -79,9 +79,9 @@ TupleJoiner::TupleJoiner(const rowgroup::RowGroup& smallInput, const rowgroup::R
     _pool.reset(new boost::shared_ptr<PoolAllocator>[bucketCount]);
     for (i = 0; i < bucketCount; i++)
     {
-      STLPoolAllocator<pair<const int64_t, Row::Pointer>> alloc(resourceManager_);
+      // STLPoolAllocator<pair<const int64_t, Row::Pointer>> alloc(resourceManager_);
       // _pool[i] = alloc.getPoolAllocator();
-      // auto alloc = resourceManager_->getAllocator<pair<const int64_t, Row::Pointer>>();
+      auto alloc = resourceManager_->getAllocator<pair<const int64_t, Row::Pointer>>();
       sth[i].reset(new sthash_t(10, hasher(), sthash_t::key_equal(), alloc));
     }
   }
@@ -91,9 +91,9 @@ TupleJoiner::TupleJoiner(const rowgroup::RowGroup& smallInput, const rowgroup::R
     _pool.reset(new boost::shared_ptr<PoolAllocator>[bucketCount]);
     for (i = 0; i < bucketCount; i++)
     {
-      STLPoolAllocator<pair<const int64_t, uint8_t*>> alloc(resourceManager_);
+      // STLPoolAllocator<pair<const int64_t, uint8_t*>> alloc(resourceManager_);
       // _pool[i] = alloc.getPoolAllocator();
-      // auto alloc = resourceManager_->getAllocator<pair<const int64_t, uint8_t*>>();
+      auto alloc = resourceManager_->getAllocator<pair<const int64_t, uint8_t*>>();
       h[i].reset(new hash_t(10, hasher(), hash_t::key_equal(), alloc));
     }
   }
@@ -184,9 +184,9 @@ TupleJoiner::TupleJoiner(const rowgroup::RowGroup& smallInput, const rowgroup::R
   ht.reset(new boost::scoped_ptr<typelesshash_t>[bucketCount]);
   for (i = 0; i < bucketCount; i++)
   {
-    STLPoolAllocator<pair<const TypelessData, Row::Pointer>> alloc(resourceManager_);
+    // STLPoolAllocator<pair<const TypelessData, Row::Pointer>> alloc(resourceManager_);
     // _pool[i] = alloc.getPoolAllocator();
-    // auto alloc = resourceManager_->getAllocator<pair<const TypelessData, Row::Pointer>>();
+    auto alloc = resourceManager_->getAllocator<pair<const TypelessData, Row::Pointer>>();
     ht[i].reset(new typelesshash_t(10, hasher(), typelesshash_t::key_equal(), alloc));
   }
   m_bucketLocks.reset(new boost::mutex[bucketCount]);
@@ -1833,9 +1833,9 @@ void TupleJoiner::clearData()
 
   for (uint i = 0; i < bucketCount; i++)
   {
-    STLPoolAllocator<pair<const TypelessData, Row::Pointer>> alloc(resourceManager_);
+    // STLPoolAllocator<pair<const TypelessData, Row::Pointer>> alloc(resourceManager_);
     // _pool[i] = alloc.getPoolAllocator();
-    // auto alloc = resourceManager_->getAllocator<pair<const TypelessData, Row::Pointer>>();
+    auto alloc = resourceManager_->getAllocator<pair<const TypelessData, Row::Pointer>>();
     if (typelessJoin)
       ht[i].reset(new typelesshash_t(10, hasher(), typelesshash_t::key_equal(), alloc));
     else if (smallRG.getColTypes()[smallKeyColumns[0]] == CalpontSystemCatalog::LONGDOUBLE)
diff --git a/utils/joiner/tuplejoiner.h b/utils/joiner/tuplejoiner.h
index 282642b1f..c8cb8e498 100644
--- a/utils/joiner/tuplejoiner.h
+++ b/utils/joiner/tuplejoiner.h
@@ -478,37 +478,37 @@ class TupleJoiner
   }
 
  private:
-   typedef std::unordered_multimap<int64_t, uint8_t*, hasher, std::equal_to<int64_t>,
-                                  utils::STLPoolAllocator<std::pair<const int64_t, uint8_t*> > >
-      hash_t;
-  typedef std::unordered_multimap<int64_t, rowgroup::Row::Pointer, hasher, std::equal_to<int64_t>,
-                                  utils::STLPoolAllocator<std::pair<const int64_t, rowgroup::Row::Pointer> > >
-      sthash_t;
-  typedef std::unordered_multimap<
-      TypelessData, rowgroup::Row::Pointer, hasher, std::equal_to<TypelessData>,
-      utils::STLPoolAllocator<std::pair<const TypelessData, rowgroup::Row::Pointer> > >
-      typelesshash_t;
-  // MCOL-1822 Add support for Long Double AVG/SUM small side
-  typedef std::unordered_multimap<
-      long double, rowgroup::Row::Pointer, hasher, LongDoubleEq,
-      utils::STLPoolAllocator<std::pair<const long double, rowgroup::Row::Pointer> > >
-      ldhash_t;
+    // typedef std::unordered_multimap<int64_t, uint8_t*, hasher, std::equal_to<int64_t>,
+    //                                 utils::STLPoolAllocator<std::pair<const int64_t, uint8_t*> > >
+    //     hash_t;
+    // typedef std::unordered_multimap<int64_t, rowgroup::Row::Pointer, hasher, std::equal_to<int64_t>,
+    //                                 utils::STLPoolAllocator<std::pair<const int64_t, rowgroup::Row::Pointer> > >
+    //     sthash_t;
+    // typedef std::unordered_multimap<
+    //     TypelessData, rowgroup::Row::Pointer, hasher, std::equal_to<TypelessData>,
+    //     utils::STLPoolAllocator<std::pair<const TypelessData, rowgroup::Row::Pointer> > >
+    //     typelesshash_t;
+    // // MCOL-1822 Add support for Long Double AVG/SUM small side
+    // typedef std::unordered_multimap<
+    //     long double, rowgroup::Row::Pointer, hasher, LongDoubleEq,
+    //     utils::STLPoolAllocator<std::pair<const long double, rowgroup::Row::Pointer> > >
+    //     ldhash_t;
 
-  // typedef std::unordered_multimap<int64_t, uint8_t*, hasher, std::equal_to<int64_t>,
-  //                                 allocators::CountingAllocator<std::pair<const int64_t, uint8_t*> > >
-  //     hash_t;
-  // typedef std::unordered_multimap<int64_t, rowgroup::Row::Pointer, hasher, std::equal_to<int64_t>,
-  //                                 allocators::CountingAllocator<std::pair<const int64_t, rowgroup::Row::Pointer> > >
-  //     sthash_t;
-  // typedef std::unordered_multimap<
-  //     TypelessData, rowgroup::Row::Pointer, hasher, std::equal_to<TypelessData>,
-  //     allocators::CountingAllocator<std::pair<const TypelessData, rowgroup::Row::Pointer> > >
-  //     typelesshash_t;
-  // // MCOL-1822 Add support for Long Double AVG/SUM small side
-  // typedef std::unordered_multimap<
-  //     long double, rowgroup::Row::Pointer, hasher, LongDoubleEq,
-  //     allocators::CountingAllocator<std::pair<const long double, rowgroup::Row::Pointer> > >
-  //     ldhash_t;
+    typedef std::unordered_multimap<int64_t, uint8_t*, hasher, std::equal_to<int64_t>,
+                                    allocators::CountingAllocator<std::pair<const int64_t, uint8_t*> > >
+        hash_t;
+    typedef std::unordered_multimap<int64_t, rowgroup::Row::Pointer, hasher, std::equal_to<int64_t>,
+                                    allocators::CountingAllocator<std::pair<const int64_t, rowgroup::Row::Pointer> > >
+        sthash_t;
+    typedef std::unordered_multimap<
+        TypelessData, rowgroup::Row::Pointer, hasher, std::equal_to<TypelessData>,
+        allocators::CountingAllocator<std::pair<const TypelessData, rowgroup::Row::Pointer> > >
+        typelesshash_t;
+    // MCOL-1822 Add support for Long Double AVG/SUM small side
+    typedef std::unordered_multimap<
+        long double, rowgroup::Row::Pointer, hasher, LongDoubleEq,
+        allocators::CountingAllocator<std::pair<const long double, rowgroup::Row::Pointer> > >
+        ldhash_t;
 
   typedef hash_t::iterator iterator;
   typedef typelesshash_t::iterator thIterator;