diff --git a/bilix/download/base_downloader_part.py b/bilix/download/base_downloader_part.py
index c8458de7..12ea055c 100644
--- a/bilix/download/base_downloader_part.py
+++ b/bilix/download/base_downloader_part.py
@@ -16,6 +16,8 @@
 
 __all__ = ['BaseDownloaderPart']
 
+domain_scores = {}
+
 
 class BaseDownloaderPart(BaseDownloader):
     """Base Async http Content-Range Downloader"""
@@ -199,6 +201,7 @@ async def get_file(self, url_or_urls: Union[str, Iterable[str]], path: Union[Pat
 
     async def _get_file_part(self, urls: List[str], path: Path, part_range: Tuple[int, int],
                              task_id) -> Path:
+        global domain_scores
         start, end = part_range
         part_path = path.with_name(f'{path.name}.{part_range[0]}-{part_range[1]}')
         exist, part_path = path_check(part_path)
@@ -208,10 +211,20 @@ async def _get_file_part(self, urls: List[str], path: Path, part_range: Tuple[in
             await self.progress.update(task_id, advance=downloaded)
         if start > end:
             return part_path  # skip already finished
-        url_idx = random.randint(0, len(urls) - 1)
 
         for times in range(1 + self.stream_retry):
+            # find domain with min score in domain_scores
+            domains = [urlparse(url_).netloc for url_ in urls]
+            domain_idx = min(range(len(domains)), key=lambda i: domain_scores.get(domains[i], 0))
+            url_idx = domain_idx
+            # update domain_scores
+            domain_min_score = domain_scores.get(domains[domain_idx], 0)
+            for domain in domain_scores.keys():
+                domain_scores[domain] = domain_scores[domain] - domain_min_score
+
             try:
+                # parse domain from url
+                domain = urlparse(urls[url_idx]).netloc
                 async with \
                         self.client.stream("GET", urls[url_idx], follow_redirects=True,
                                            headers={'Range': f'bytes={start}-{end}'}) as r, \
@@ -227,6 +240,7 @@ async def _get_file_part(self, urls: List[str], path: Path, part_range: Tuple[in
                         await self._check_speed(len(chunk))
                 break
             except (httpx.HTTPStatusError, httpx.TransportError):
+                domain_scores[domain] = domain_scores.get(domain, 0) + 1
                 continue
         else:
             raise Exception(f"STREAM 超过重复次数 {part_path.name}")