Flush after every 2MB merge of replace ops.
This will be in sync with incremental OTA's where the sync
is done every 2MB. This improves performance on devices
with low memory.
Merge times for full OTA may increase by couple of seconds but
that is ok given it decreases the memory footprint.
Bug: 237490659
Test: OTA
Ignore-AOSP-First: cherry-pick from aosp
Signed-off-by: Akilesh Kailash <akailash@google.com>
Change-Id: Ic2c8d2ffdbdb677e0c4d44e5de68ce8ccf86df34
diff --git a/fs_mgr/libsnapshot/snapuserd/user-space-merge/snapuserd_merge.cpp b/fs_mgr/libsnapshot/snapuserd/user-space-merge/snapuserd_merge.cpp
index 17f3c4f..63f47d6 100644
--- a/fs_mgr/libsnapshot/snapuserd/user-space-merge/snapuserd_merge.cpp
+++ b/fs_mgr/libsnapshot/snapuserd/user-space-merge/snapuserd_merge.cpp
@@ -71,16 +71,16 @@
}
bool Worker::MergeReplaceZeroOps() {
- // Flush every 8192 ops. Since all ops are independent and there is no
+ // Flush after merging 2MB. Since all ops are independent and there is no
// dependency between COW ops, we will flush the data and the number
- // of ops merged in COW file for every 8192 ops. If there is a crash,
- // we will end up replaying some of the COW ops which were already merged.
- // That is ok.
+ // of ops merged in COW block device. If there is a crash, we will
+ // end up replaying some of the COW ops which were already merged. That is
+ // ok.
//
- // Why 8192 ops ? Increasing this may improve merge time 3-4 seconds but
- // we need to make sure that we checkpoint; 8k ops seems optimal. In-case
- // if there is a crash merge should always make forward progress.
- int total_ops_merged_per_commit = (PAYLOAD_BUFFER_SZ / BLOCK_SZ) * 32;
+ // Although increasing this greater than 2MB may help in improving merge
+ // times; however, on devices with low memory, this can be problematic
+ // when there are multiple merge threads in parallel.
+ int total_ops_merged_per_commit = (PAYLOAD_BUFFER_SZ / BLOCK_SZ) * 2;
int num_ops_merged = 0;
SNAP_LOG(INFO) << "MergeReplaceZeroOps started....";