{"id":1196,"date":"2026-04-07T09:33:25","date_gmt":"2026-04-07T01:33:25","guid":{"rendered":"https:\/\/www.ndnlab.com\/?p=1196"},"modified":"2026-04-07T09:33:27","modified_gmt":"2026-04-07T01:33:27","slug":"swash-a-flexible-communication-framework-with-sliding-window-based-cache-sharing-for-scalable-dgnn-training","status":"publish","type":"post","link":"https:\/\/www.ndnlab.com\/?p=1196","title":{"rendered":"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training"},"content":{"rendered":"\n<p>Proceedings of the ACM on Management of Data, Volume 3, Issue 3 (June 2025)<\/p>\n\n\n\n<p>h\ue03cps:\/\/doi.org\/10.1145\/3725360<\/p>\n\n\n\n<p>EISSN: 2836-6573<\/p>\n\n\n\n<p><strong>1. \u6458\u8981\uff08Abstract\uff09<\/strong><\/p>\n\n\n\n<p>SWASH\u63d0\u51fa\u4e86\u4e00\u79cd\u4e13\u4e3a\u52a8\u6001\u56fe\u795e\u7ecf\u7f51\u7edc\uff08DGNN\uff09\u5206\u5e03\u5f0f\u8bad\u7ec3\u8bbe\u8ba1\u7684\u5168\u65b0\u901a\u4fe1\u6846\u67b6\uff0c\u6838\u5fc3\u76ee\u6807\u662f\u89e3\u51b3\u6ed1\u52a8\u7a97\u53e3\u8bad\u7ec3\u6a21\u5f0f\u4e0b\u901a\u4fe1\u5f00\u9500\u5927\u3001\u73b0\u6709\u5206\u533a\u7b56\u7565\u4e0d\u517c\u5bb9\u4ee5\u53ca\u7f13\u5b58\u590d\u7528\u6548\u7387\u4f4e\u4e09\u5927\u6311\u6218\u3002\u4f20\u7edf\u5206\u5e03\u5f0fDGNN\u8bad\u7ec3\u6846\u67b6\uff08\u5982\u5feb\u7167\u5206\u533a\u3001Chunk-based\u5206\u533a\u548cL-hop\u7f13\u5b58\u514d\u901a\u4fe1\u9876\u70b9\u5206\u533a\uff09\u5728\u6ed1\u52a8\u7a97\u53e3\u573a\u666f\u4e0b\u5b58\u5728\u4e25\u91cd\u4e0d\u5339\u914d\uff1a\u5feb\u7167\u5206\u533a\u96be\u4ee5\u652f\u6301\u8de8\u7a97\u53e3\u901a\u4fe1\u4f18\u5316\uff0cChunk-based\u5206\u533a\u867d\u51cf\u5c11\u5355\u5feb\u7167\u901a\u4fe1\u4f46\u5ffd\u7565\u6ed1\u52a8\u7a97\u53e3\u7684\u65f6\u5e8f\u7279\u6027\uff0c\u800c\u73b0\u6709\u7f13\u5b58\u6280\u672f\u4ec5\u9650\u4e8e\u5355\u673a\u5185\u90e8\u805a\u5408\u7ed3\u679c\u590d\u7528\uff0c\u65e0\u6cd5\u6709\u6548\u964d\u4f4e\u5206\u5e03\u5f0f\u73af\u5883\u4e0b\u7684\u7f51\u7edc\u901a\u4fe1\u91cf\u3002<\/p>\n\n\n\n<p>SWASH\u901a\u8fc7\u5f15\u5165\u6ed1\u52a8\u7a97\u53e3\u4e3a\u57fa\u7840\u7684\u7f13\u5b58\u5171\u4eab\uff08Sliding Window-based Cache Sharing\uff09\u6280\u672f\uff0c\u5b9e\u73b0\u4e86\u901a\u4fe1\u4e0e\u7f13\u5b58\u7684\u7075\u6d3b\u878d\u5408\u3002\u5177\u4f53\u800c\u8a00\uff0c\u7cfb\u7edf\u63d0\u51fa\u4e86\u4e00\u5957\u652f\u6301\u901a\u4fe1\u6bd4\u4f8b\u8c03\u6574\u3001\u65f6\u95f4\u9009\u62e9\u3001\u8d85\u53c2\u6570\u8bbe\u7f6e\u4ee5\u53ca\u81ea\u9002\u5e94\u8c03\u5ea6\u7684\u67d4\u6027\u901a\u4fe1\u6846\u67b6\uff0c\u540c\u65f6\u8bbe\u8ba1\u4e86\u8f7b\u91cf\u7ea7LightMetis\u5206\u533a\u7b56\u7565\uff0c\u4ec5\u9488\u5bf9\u6ed1\u52a8\u7a97\u53e3\u8bad\u7ec3\u7279\u70b9\u8fdb\u884c\u52a0\u6743\u56fe\u5212\u5206\uff0c\u5927\u5e45\u964d\u4f4e\u5206\u533a\u548c\u901a\u4fe1\u5f00\u9500\u3002\u4e3a\u4e86\u7f13\u89e3\u56e0\u51cf\u5c11\u901a\u4fe1\u5bfc\u81f4\u7684\u7cbe\u5ea6\u4e0b\u964d\uff0cSWASH\u8fdb\u4e00\u6b65\u63d0\u51fa\u57fa\u4e8e\u6ed1\u52a8\u7a97\u53e3\u7684\u8fb9\u754c\u9876\u70b9\u5d4c\u5165\u7f13\u5b58\u5171\u4eab\u673a\u5236\uff0c\u5728\u4e0d\u540c\u6ed1\u52a8\u7a97\u53e3\u4e4b\u95f4\u590d\u7528\u5386\u53f2\u5d4c\u5165\u4fe1\u606f\uff0c\u5b9e\u73b0\u901a\u4fe1\u91cf\u4e0e\u6a21\u578b\u7cbe\u5ea6\u7684\u5e73\u8861\u3002<\/p>\n\n\n\n<p>\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0cSWASH\u5728\u516d\u4e2a\u771f\u5b9e\u6570\u636e\u96c6\u548c\u4e09\u79cd\u771f\u5b9e\u73af\u5883\u4e0a\uff0c\u76f8\u6bd4\u73b0\u6709\u6700\u5148\u8fdb\u6ed1\u52a8\u7a97\u53e3\u6846\u67b6\u5e73\u5747\u52a0\u901f9.44\u500d\uff0c\u540c\u65f6\u4fdd\u6301\u4e86\u5168\u901a\u4fe1\u3001\u65e0\u7f13\u5b58\u8bad\u7ec3\u6846\u67b6\u7684\u7cbe\u5ea6\u3002\u8be5\u5de5\u4f5c\u4e0d\u4ec5\u586b\u8865\u4e86\u5206\u5e03\u5f0fDGNN\u8bad\u7ec3\u5728\u6ed1\u52a8\u7a97\u53e3\u6a21\u5f0f\u4e0b\u7684\u7cfb\u7edf\u7a7a\u767d\uff0c\u8fd8\u4e3a\u5927\u89c4\u6a21\u52a8\u6001\u56fe\u5b66\u4e60\u63d0\u4f9b\u4e86\u53ef\u6269\u5c55\u3001\u53ef\u7075\u6d3b\u914d\u7f6e\u7684\u901a\u4fe1\u4f18\u5316\u8303\u5f0f\uff0c\u5177\u6709\u91cd\u8981\u7684\u7406\u8bba\u521b\u65b0\u4ef7\u503c\u548c\u5de5\u7a0b\u5b9e\u7528\u610f\u4e49\u3002<\/p>\n\n\n\n<p><strong>2. \u5f15\u8a00\uff08Introduction\uff09<\/strong><\/p>\n\n\n\n<p>\u52a8\u6001\u56fe\u795e\u7ecf\u7f51\u7edc\uff08DGNN\uff09\u5728\u4ea4\u901a\u6d41\u91cf\u9884\u6d4b\u3001\u793e\u4ea4\u7f51\u7edc\u5206\u6790\u3001\u6d41\u884c\u75c5\u4f20\u64ad\u9884\u6d4b\u7b49\u5b9e\u9645\u5e94\u7528\u4e2d\u5c55\u73b0\u51fa\u5f3a\u5927\u80fd\u529b\uff0c\u5176\u6838\u5fc3\u4f18\u52bf\u5728\u4e8e\u80fd\u591f\u540c\u65f6\u6355\u6349\u56fe\u7684\u5c5e\u6027\u3001\u7ed3\u6784\u4ee5\u53ca\u65f6\u5e8f\u6f14\u5316\u7279\u6027\u3002\u7136\u800c\uff0c\u5927\u89c4\u6a21\u52a8\u6001\u56fe\u8bad\u7ec3\u9762\u4e34\u6781\u9ad8\u7684\u8ba1\u7b97\u548c\u901a\u4fe1\u5f00\u9500\uff1aDGNN\u9700\u901a\u8fc7RNN\u5efa\u6a21\u65f6\u5e8f\u52a8\u6001\uff0c\u5bfc\u81f4\u901a\u4fe1\u91cf\u548c\u8ba1\u7b97\u91cf\u662f\u9759\u6001GNN\u7684T\u500d\uff08T\u4e3a\u5feb\u7167\u6570\uff09\u3002\u6ed1\u52a8\u7a97\u53e3\u8bad\u7ec3\u867d\u80fd\u5c06\u590d\u6742\u5ea6\u4eceO(T)\u964d\u4f4e\u81f3O(Ns)\uff08Ns&lt;T\uff09\uff0c\u4f46\u5728\u5355\u673a\u65e0\u6cd5\u627f\u8f7d\u8d85\u5927\u89c4\u6a21\u52a8\u6001\u56fe\u65f6\uff0c\u5206\u5e03\u5f0f\u8bad\u7ec3\u6210\u4e3a\u5fc5\u7136\u9009\u62e9\u3002<\/p>\n\n\n\n<p>\u73b0\u6709\u5206\u5e03\u5f0f\u6846\u67b6\u4e3b\u8981\u9762\u4e34\u4e09\u5927\u6311\u6218\uff1a\u4e00\u662f\u5168\u901a\u4fe1\u5bfc\u81f4\u7684\u5de8\u5927\u7f51\u7edc\u5f00\u9500\uff1b\u4e8c\u662f\u73b0\u6709\u5206\u533a\u7b56\u7565\uff08\u5982\u5feb\u7167\u5206\u533a\u3001Chunk-based\u5206\u533a\uff09\u65e0\u6cd5\u6709\u6548\u9002\u914d\u6ed1\u52a8\u7a97\u53e3\u8bad\u7ec3\uff0c\u8de8\u5feb\u7167\u901a\u4fe1\u4f18\u5316\u56f0\u96be\uff1b\u4e09\u662f\u7f13\u5b58\u6280\u672f\u4ec5\u5c40\u9650\u4e8e\u5355\u5feb\u7167\u5185\u90e8\u590d\u7528\uff0c\u65e0\u6cd5\u5229\u7528\u6ed1\u52a8\u7a97\u53e3\u4e4b\u95f4\u7684\u65f6\u5e8f\u76f8\u5173\u6027\u8fdb\u4e00\u6b65\u964d\u4f4e\u901a\u4fe1\u3002SWASH\u6b63\u662f\u9488\u5bf9\u8fd9\u4e9b\u75db\u70b9\u800c\u8bbe\u8ba1\uff0c\u5b83\u9996\u6b21\u63d0\u51fa\u4e86\u4e00\u5957\u67d4\u6027\u901a\u4fe1\u6846\u67b6\uff0c\u652f\u6301\u4efb\u610f\u901a\u4fe1\u6bd4\u4f8b\u3001\u4efb\u610f\u901a\u4fe1\u65f6\u673a\u4ee5\u53ca\u81ea\u9002\u5e94\u8c03\u5ea6\u7b56\u7565\u3002\u540c\u65f6\uff0c\u7cfb\u7edf\u521b\u65b0\u6027\u5730\u5f15\u5165\u8f7b\u91cf\u7ea7\u5206\u533a\u7b56\u7565\u548c\u6ed1\u52a8\u7a97\u53e3\u7f13\u5b58\u5171\u4eab\u673a\u5236\uff0c\u5728\u4fdd\u8bc1\u6a21\u578b\u7cbe\u5ea6\u7684\u524d\u63d0\u4e0b\u5927\u5e45\u964d\u4f4e\u901a\u4fe1\u548c\u540c\u6b65\u5f00\u9500\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"813\" height=\"477\" src=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/1-6.png\"  class=\"wp-image-1197\" srcset=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/1-6.png 813w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/1-6-300x176.png 300w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/1-6-768x451.png 768w\" sizes=\"auto, (max-width: 813px) 100vw, 813px\" title=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe\" alt=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe\" \/><\/figure>\n\n\n\n<p>\u8bba\u6587\u4ee5\u771f\u5b9e\u52a8\u6001\u56fe\u6570\u636e\u96c6\u4e3a\u4f8b\uff0c\u7cfb\u7edf\u6027\u5730\u5206\u6790\u4e86\u73b0\u6709\u65b9\u6cd5\u7684\u5c40\u9650\uff0c\u5e76\u901a\u8fc7\u7406\u8bba\u5206\u6790\u4e0e\u5927\u91cf\u5b9e\u9a8c\u9a8c\u8bc1\u4e86SWASH\u7684\u6709\u6548\u6027\u3002\u5b9e\u9a8c\u8868\u660e\uff0cSWASH\u5728\u4fdd\u6301\u5168\u901a\u4fe1\u6846\u67b6\u7cbe\u5ea6\u7684\u540c\u65f6\uff0c\u5b9e\u73b0\u4e86\u5e73\u57479.44\u500d\u7684\u8bad\u7ec3\u52a0\u901f\uff0c\u5145\u5206\u8bc1\u660e\u4e86\u5176\u5728\u53ef\u6269\u5c55DGNN\u8bad\u7ec3\u4e2d\u7684\u91cd\u8981\u4ef7\u503c\u3002\u8be5\u5de5\u4f5c\u4e0d\u4ec5\u4e3a\u52a8\u6001\u56fe\u5b66\u4e60\u63d0\u4f9b\u4e86\u5b9e\u7528\u7cfb\u7edf\u6846\u67b6\uff0c\u4e5f\u4e3a\u5206\u5e03\u5f0f\u673a\u5668\u5b66\u4e60\u4e2d\u7684\u901a\u4fe1\u4f18\u5316\u63d0\u4f9b\u4e86\u65b0\u7684\u601d\u8def\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"807\" height=\"278\" src=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/2-6.png\"  class=\"wp-image-1198\" srcset=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/2-6.png 807w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/2-6-300x103.png 300w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/2-6-768x265.png 768w\" sizes=\"auto, (max-width: 807px) 100vw, 807px\" title=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe1\" alt=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe1\" \/><\/figure>\n\n\n\n<p><strong>3. \u80cc\u666f\u4e0e\u52a8\u673a\uff08Background and Motivation\uff09<\/strong><\/p>\n\n\n\n<p>\u52a8\u6001\u56fe\u901a\u5e38\u88ab\u5efa\u6a21\u4e3a\u4e00\u7cfb\u5217\u79bb\u6563\u5feb\u7167\u5e8f\u5217\uff0c\u6bcf\u4e2a\u5feb\u7167\u5305\u542b\u9876\u70b9\u96c6\u3001\u8fb9\u96c6\u4ee5\u53ca\u5bf9\u5e94\u7684\u7279\u5f81\u3002DGNN\u8bad\u7ec3\u9700\u540c\u65f6\u5904\u7406\u7a7a\u95f4\u90bb\u5c45\u805a\u5408\uff08GNN\u90e8\u5206\uff09\u548c\u65f6\u5e8f\u52a8\u6001\u5efa\u6a21\uff08RNN\u90e8\u5206\uff09\uff0c\u5bfc\u81f4\u901a\u4fe1\u5305\u62ecGNN\u901a\u4fe1\uff08\u540c\u5feb\u7167\u90bb\u5c45\u5d4c\u5165\u4ea4\u6362\uff09\u548cRNN\u901a\u4fe1\uff08\u8de8\u5feb\u7167\u9690\u85cf\u72b6\u6001\u4f20\u9012\uff09\u3002\u6ed1\u52a8\u7a97\u53e3\u8bad\u7ec3\u901a\u8fc7\u5728\u6bcf\u4e2a\u8fed\u4ee3\u4e2d\u4f7f\u7528\u8fde\u7eedNs\u4e2a\u5feb\u7167\u4ee3\u66ff\u5b8c\u6574T\u4e2a\u5feb\u7167\uff0c\u663e\u8457\u964d\u4f4e\u4e86\u8ba1\u7b97\u590d\u6742\u5ea6\uff0c\u540c\u65f6\u5f15\u5165\u6837\u672c\u591a\u6837\u6027\u63d0\u5347\u6cdb\u5316\u80fd\u529b\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"810\" height=\"450\" src=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/3-4.png\"  class=\"wp-image-1199\" srcset=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/3-4.png 810w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/3-4-300x167.png 300w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/3-4-768x427.png 768w\" sizes=\"auto, (max-width: 810px) 100vw, 810px\" title=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe2\" alt=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe2\" \/><\/figure>\n\n\n\n<p>\u7136\u800c\uff0c\u73b0\u6709\u5206\u5e03\u5f0fDGNN\u6846\u67b6\u5728\u6ed1\u52a8\u7a97\u53e3\u6a21\u5f0f\u4e0b\u5b58\u5728\u660e\u663e\u4e0d\u9002\u914d\uff1a\u5feb\u7167\u5206\u533a\u867d\u6d88\u9664GNN\u901a\u4fe1\uff0c\u4f46\u65e0\u6cd5\u652f\u6301\u8de8\u673a\u5668\u6ed1\u52a8\u7a97\u53e3\u8bad\u7ec3\uff1bChunk-based\u5206\u533a\u867d\u878d\u5408\u7ed3\u6784\u4e0e\u65f6\u5e8f\u90bb\u5c45\uff0c\u4f46\u4ecd\u9700\u4e24\u8f6e\u540c\u6b65\u4e14\u4e0d\u9002\u5408\u6ed1\u52a8\u7a97\u53e3\uff1b\u9876\u70b9\u5206\u533a\u867d\u652f\u6301\u6ed1\u52a8\u7a97\u53e3\uff0c\u5374\u56e0\u9759\u6001\u56fe\u8bbe\u8ba1\u5bfc\u81f4\u8de8\u5feb\u7167\u901a\u4fe1\u4f18\u5316\u4e0d\u8db3\u3002\u6b64\u5916\uff0c\u73b0\u6709\u7f13\u5b58\u6280\u672f\uff08\u5982CacheG\u3001PiPAD\uff09\u4e3b\u8981\u9488\u5bf9\u5355\u673a\u8ba1\u7b97\u4f18\u5316\uff0c\u65e0\u6cd5\u6709\u6548\u51cf\u5c11\u5206\u5e03\u5f0f\u573a\u666f\u4e0b\u7684\u7f51\u7edc\u901a\u4fe1\uff0c\u4e14\u5ffd\u7565\u6ed1\u52a8\u7a97\u53e3\u4e4b\u95f4\u7684\u5d4c\u5165\u5171\u4eab\u673a\u4f1a\u3002<\/p>\n\n\n\n<p>SWASH\u7684\u52a8\u673a\u6b63\u662f\u586b\u8865\u8fd9\u4e00\u7a7a\u767d\uff1a\u901a\u8fc7\u67d4\u6027\u901a\u4fe1\u6846\u67b6\u5b9e\u73b0\u901a\u4fe1\u6bd4\u4f8b\u548c\u65f6\u673a\u7684\u52a8\u6001\u8c03\u6574\uff0c\u901a\u8fc7\u8f7b\u91cf\u7ea7LightMetis\u5206\u533a\u964d\u4f4e\u8de8\u7a97\u53e3\u901a\u4fe1\uff0c\u901a\u8fc7\u6ed1\u52a8\u7a97\u53e3\u7f13\u5b58\u5171\u4eab\u673a\u5236\u590d\u7528\u5386\u53f2\u5d4c\u5165\uff0c\u4ece\u800c\u5728\u5206\u5e03\u5f0f\u73af\u5883\u4e0b\u5b9e\u73b0\u9ad8\u6548\u3001\u53ef\u6269\u5c55\u7684DGNN\u8bad\u7ec3\u3002\u8be5\u8bbe\u8ba1\u4e0d\u4ec5\u89e3\u51b3\u4e86\u73b0\u6709\u65b9\u6cd5\u7684\u4e09\u5927\u6311\u6218\uff0c\u4e5f\u4e3a\u540e\u7eed\u52a8\u6001\u56fe\u7cfb\u7edf\u7814\u7a76\u63d0\u4f9b\u4e86\u91cd\u8981\u53c2\u8003\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"804\" height=\"228\" src=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/4-3.png\"  class=\"wp-image-1200\" srcset=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/4-3.png 804w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/4-3-300x85.png 300w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/4-3-768x218.png 768w\" sizes=\"auto, (max-width: 804px) 100vw, 804px\" title=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe3\" alt=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe3\" \/><\/figure>\n\n\n\n<p><strong>4. \u7cfb\u7edf\u6846\u67b6\u4e0e\u5206\u533a\u7b56\u7565\uff08System Framework and Partitioning\uff09<\/strong><\/p>\n\n\n\n<p>SWASH\u6574\u4f53\u67b6\u6784\u5305\u542b\u4e94\u4e2a\u6838\u5fc3\u6a21\u5757\uff1a\u8f7b\u91cf\u7ea7\u5206\u533a\u6a21\u5757\u3001\u7f16\u7801\u6a21\u5757\u3001\u901a\u4fe1\u6a21\u5757\u3001\u7f13\u5b58\u6a21\u5757\u548c\u8c03\u5ea6\u7ba1\u7406\u6a21\u5757\u3002\u5206\u533a\u6a21\u5757\u9996\u5148\u5bf9\u52a8\u6001\u56fe\u8fdb\u884c\u8f7b\u91cf\u7ea7\u5212\u5206\uff0c\u751f\u6210\u9876\u70b9\u5230\u5de5\u4f5c\u8282\u70b9\u7684\u6620\u5c04\u8868\uff1b\u7f16\u7801\u6a21\u5757\u8d1f\u8d23\u672c\u5730\u6570\u636e\u52a0\u8f7d\u3001\u7d22\u5f15\u6620\u5c04\u53caGPU\u7f13\u5b58\u6784\u5efa\uff1b\u901a\u4fe1\u4e0e\u7f13\u5b58\u6a21\u5757\u5219\u6839\u636e\u8c03\u5ea6\u7b56\u7565\u51b3\u5b9a\u8fb9\u754c\u9876\u70b9\u662f\u8d70\u7f51\u7edc\u901a\u4fe1\u8fd8\u662f\u76f4\u63a5\u4ece\u7f13\u5b58\u8bfb\u53d6\uff1b\u8c03\u5ea6\u7ba1\u7406\u6a21\u5757\u8d1f\u8d23\u8bad\u7ec3\u6837\u672c\u751f\u6210\u3001\u901a\u4fe1\u65f6\u673a\u9009\u62e9\u4ee5\u53ca\u7f13\u5b58\u66f4\u65b0\u534f\u8c03\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"816\" height=\"320\" src=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/5-3.png\"  class=\"wp-image-1201\" srcset=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/5-3.png 816w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/5-3-300x118.png 300w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/5-3-768x301.png 768w\" sizes=\"auto, (max-width: 816px) 100vw, 816px\" title=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe4\" alt=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe4\" \/><\/figure>\n\n\n\n<p>\u5728\u5206\u533a\u7b56\u7565\u4e0a\uff0cSWASH\u63d0\u51faLightMetis\u7b97\u6cd5\uff0c\u4e13\u95e8\u9488\u5bf9\u6ed1\u52a8\u7a97\u53e3\u8bad\u7ec3\u7279\u70b9\u8bbe\u8ba1\u3002\u5b83\u9996\u5148\u8ba1\u7b97\u539f\u59cb\u5feb\u7167\u7684\u4f7f\u7528\u9891\u7387\uff08\u8003\u8651\u6ed1\u52a8\u7a97\u53e3\u957f\u5ea6Ns\u548c\u8fb9\u751f\u547d\u5468\u671fz\u7684\u53cc\u91cd\u5f71\u54cd\uff09\uff0c\u7136\u540e\u4ec5\u4fdd\u7559\u9ad8\u9891\u539f\u59cb\u5feb\u7167\uff0c\u6784\u5efa\u52a0\u6743\u5408\u5e76\u56fe\uff0c\u6700\u540e\u4f7f\u7528Metis\u8fdb\u884c\u52a0\u6743\u5212\u5206\u3002\u8be5\u7b56\u7565\u663e\u8457\u964d\u4f4e\u4e86\u5206\u533a\u65f6\u95f4\u548c\u540e\u7eed\u901a\u4fe1\u5f00\u9500\uff0c\u540c\u65f6\u4fdd\u8bc1\u540c\u4e00\u9876\u70b9\u5728\u6240\u6709\u5feb\u7167\u4e2d\u88ab\u5206\u914d\u5230\u540c\u4e00\u5de5\u4f5c\u8282\u70b9\uff0c\u5f7b\u5e95\u6d88\u9664RNN\u901a\u4fe1\u3002<\/p>\n\n\n\n<p>\u5b9e\u9a8c\u9a8c\u8bc1\u663e\u793a\uff0cLightMetis\u76f8\u6bd4\u4f20\u7edfMetis\u5206\u533a\u5728\u52a8\u6001\u56fe\u4e0a\u901a\u4fe1\u91cf\u66f4\u4f4e\u3001\u5206\u533a\u5f00\u9500\u66f4\u5c0f\uff0c\u4e3a\u540e\u7eed\u67d4\u6027\u901a\u4fe1\u548c\u7f13\u5b58\u5171\u4eab\u5960\u5b9a\u4e86\u9ad8\u6548\u57fa\u7840\u3002\u8be5\u6a21\u5757\u7684\u8bbe\u8ba1\u5145\u5206\u4f53\u73b0\u4e86SWASH\u5bf9\u6ed1\u52a8\u7a97\u53e3\u8bad\u7ec3\u7684\u6df1\u5ea6\u9002\u914d\uff0c\u662f\u7cfb\u7edf\u5b9e\u73b0\u9ad8\u6027\u80fd\u7684\u5173\u952e\u521b\u65b0\u4e4b\u4e00\u3002<\/p>\n\n\n\n<p><strong>5. \u67d4\u6027\u901a\u4fe1\u4e0e\u7f13\u5b58\u5171\u4eab\u6280\u672f\uff08Flexible Communication and Cache Sharing\uff09<\/strong><\/p>\n\n\n\n<p>SWASH\u7684\u6838\u5fc3\u521b\u65b0\u5728\u4e8e\u63d0\u51fa\u4e86\u4e00\u5957\u67d4\u6027\u901a\u4fe1\u6846\u67b6\uff0c\u652f\u6301\u4e09\u4e2a\u5173\u952e\u8d85\u53c2\u6570\uff1a\u901a\u4fe1\u95f4\u9694k\uff08\u6bcfk\u4e2aepoch\u901a\u4fe1\u4e00\u6b21\uff09\u3001\u901a\u4fe1\u7a97\u53e3\u6570s\uff08\u6bcf\u4e2aepoch\u5185\u901a\u4fe1s\u4e2a\u6ed1\u52a8\u7a97\u53e3\uff09\u4ee5\u53ca\u901a\u4fe1\u6bd4\u4f8bp\uff08\u6bcf\u4e2a\u5feb\u7167\u901a\u4fe1p\u6bd4\u4f8b\u7684\u8fb9\u754c\u5d4c\u5165\uff09\u3002\u901a\u8fc7\u8fd9\u4e9b\u53c2\u6570\uff0c\u7528\u6237\u53ef\u7075\u6d3b\u5e73\u8861\u901a\u4fe1\u5f00\u9500\u4e0e\u6a21\u578b\u7cbe\u5ea6\u3002<\/p>\n\n\n\n<p>\u4e3a\u964d\u4f4e\u901a\u4fe1\u65f6\u7684\u5b50\u56fe\u91cd\u6784\u5f00\u9500\uff0c\u7cfb\u7edf\u91c7\u7528\u5408\u5e76\u8fb9\u754c\u7f16\u7801\u6280\u672f\uff0c\u5c06\u6240\u6709\u5feb\u7167\u7684\u8fb9\u754c\u9876\u70b9\u96c6\u5408\u5e76\u7f16\u7801\u4e3a\u7edf\u4e00\u7d22\u5f15\uff0c\u51cf\u5c11\u91cd\u590d\u7f16\u7801\u548c\u5185\u5b58\u5360\u7528\u3002\u540c\u65f6\uff0cSWASH\u63d0\u51fa\u6ed1\u52a8\u7a97\u53e3\u4e3a\u57fa\u7840\u7684\u7f13\u5b58\u5171\u4eab\u673a\u5236\uff1a\u6bcf\u4e2a\u6ed1\u52a8\u7a97\u53e3\u7ef4\u62a4Ns\u4e2a\u5d4c\u5165\u77e9\u9635\uff0c\u540c\u4e00\u4f4d\u7f6e\u7684\u5feb\u7167\u56e0\u7ecf\u5386\u76f8\u540cRNN\u53d8\u6362\u800c\u5177\u6709\u8f83\u9ad8\u76f8\u4f3c\u6027\uff0c\u56e0\u6b64\u524d\u4e00\u4e2a\u7a97\u53e3\u7684\u5d4c\u5165\u53ef\u76f4\u63a5\u5171\u4eab\u7ed9\u540e\u7eed\u7a97\u53e3\uff0c\u4ece\u800c\u5927\u5e45\u51cf\u5c11\u8de8\u7a97\u53e3\u901a\u4fe1\u9891\u7387\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"795\" height=\"279\" src=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/6-4.png\"  class=\"wp-image-1202\" srcset=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/6-4.png 795w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/6-4-300x105.png 300w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/6-4-768x270.png 768w\" sizes=\"auto, (max-width: 795px) 100vw, 795px\" title=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe5\" alt=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe5\" \/><\/figure>\n\n\n\n<p>\u8c03\u5ea6\u6a21\u5757\u8fdb\u4e00\u6b65\u5f15\u5165\u81ea\u9002\u5e94\u7b56\u7565\uff0c\u6839\u636e\u5b9e\u65f6\u901a\u4fe1\u4ee3\u4ef7\u4e0e\u7f13\u5b58\u547d\u4e2d\u7387\u52a8\u6001\u8c03\u6574\u8d85\u53c2\u6570\uff0c\u5b9e\u73b0\u901a\u4fe1\u4e0e\u7f13\u5b58\u7684\u667a\u80fd\u5207\u6362\u3002\u8be5\u6280\u672f\u4e0d\u4ec5\u6709\u6548\u7f13\u89e3\u4e86\u7cbe\u5ea6\u4e0b\u964d\u95ee\u9898\uff0c\u8fd8\u663e\u8457\u63d0\u5347\u4e86\u6574\u4f53\u8bad\u7ec3\u6548\u7387\uff0c\u662fSWASH\u5728\u5206\u5e03\u5f0fDGNN\u8bad\u7ec3\u4e2d\u5b9e\u73b0\u9ad8\u52a0\u901f\u6bd4\u7684\u5173\u952e\u6240\u5728\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"792\" height=\"330\" src=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/7-5.png\"  class=\"wp-image-1203\" srcset=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/7-5.png 792w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/7-5-300x125.png 300w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/7-5-768x320.png 768w\" sizes=\"auto, (max-width: 792px) 100vw, 792px\" title=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe6\" alt=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe6\" \/><\/figure>\n\n\n\n<p><strong>6. \u5b9e\u9a8c\u8bc4\u4f30\uff08Evaluation\uff09<\/strong><\/p>\n\n\n\n<p>\u8bba\u6587\u5728\u516d\u4e2a\u771f\u5b9e\u52a8\u6001\u56fe\u6570\u636e\u96c6\u4e0a\u8fdb\u884c\u4e86\u5168\u9762\u5b9e\u9a8c\uff0c\u6db5\u76d6\u4ea4\u901a\u3001\u793e\u4ea4\u3001\u75ab\u60c5\u7b49\u591a\u79cd\u573a\u666f\uff0c\u5e76\u5728\u5355\u673a\u3001\u591a\u673aCPU\/GPU\u6df7\u5408\u73af\u5883\u4e0a\u9a8c\u8bc1\u7cfb\u7edf\u6027\u80fd\u3002\u76f8\u6bd4\u73b0\u6709\u6700\u5148\u8fdb\u6ed1\u52a8\u7a97\u53e3\u6846\u67b6\uff08DynaGraph\u3001DynaHB\u7b49\uff09\uff0cSWASH\u5e73\u5747\u5b9e\u73b09.44\u500d\u8bad\u7ec3\u52a0\u901f\uff0c\u540c\u65f6\u5728\u7cbe\u5ea6\u4e0a\u4e0e\u5168\u901a\u4fe1\u3001\u65e0\u7f13\u5b58\u6846\u67b6\u6301\u5e73\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"801\" height=\"765\" src=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/8-2.png\"  class=\"wp-image-1204\" srcset=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/8-2.png 801w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/8-2-300x287.png 300w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/8-2-768x733.png 768w\" sizes=\"auto, (max-width: 801px) 100vw, 801px\" title=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe7\" alt=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe7\" \/><\/figure>\n\n\n\n<p>\u6d88\u878d\u5b9e\u9a8c\u8fdb\u4e00\u6b65\u9a8c\u8bc1\u4e86\u5404\u6a21\u5757\u8d21\u732e\uff1aLightMetis\u5206\u533a\u663e\u8457\u964d\u4f4e\u901a\u4fe1\u91cf\uff0c\u67d4\u6027\u901a\u4fe1\u6846\u67b6\u901a\u8fc7\u8d85\u53c2\u6570\u8c03\u4f18\u5b9e\u73b0\u6700\u4f18\u5e73\u8861\uff0c\u6ed1\u52a8\u7a97\u53e3\u7f13\u5b58\u5171\u4eab\u5728\u51cf\u5c11\u901a\u4fe1\u7684\u540c\u65f6\u6709\u6548\u7ef4\u6301\u7cbe\u5ea6\u3002\u654f\u611f\u6027\u5206\u6790\u663e\u793a\uff0cSWASH\u5bf9\u4e0d\u540cNs\u3001z\u7b49\u53c2\u6570\u5177\u6709\u826f\u597d\u9c81\u68d2\u6027\uff0c\u5728\u5927\u89c4\u6a21\u52a8\u6001\u56fe\u4e0a\u6269\u5c55\u6027\u7a81\u51fa\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"843\" height=\"306\" src=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/9-1.png\"  class=\"wp-image-1205\" srcset=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/9-1.png 843w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/9-1-300x109.png 300w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/9-1-768x279.png 768w\" sizes=\"auto, (max-width: 843px) 100vw, 843px\" title=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe8\" alt=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe8\" \/><\/figure>\n\n\n\n<p>\u8fd9\u4e9b\u7ed3\u679c\u5145\u5206\u8bc1\u660e\u4e86SWASH\u5728\u5b9e\u9645\u90e8\u7f72\u4e2d\u7684\u5b9e\u7528\u4ef7\u503c\uff0c\u4e3a\u5927\u89c4\u6a21DGNN\u8bad\u7ec3\u63d0\u4f9b\u4e86\u9ad8\u6548\u3001\u53ef\u6269\u5c55\u7684\u7cfb\u7edf\u89e3\u51b3\u65b9\u6848\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"830\" height=\"756\" src=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/10.png\"  class=\"wp-image-1206\" srcset=\"https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/10.png 830w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/10-300x273.png 300w, https:\/\/www.ndnlab.com\/wp-content\/uploads\/2026\/03\/10-768x700.png 768w\" sizes=\"auto, (max-width: 830px) 100vw, 830px\" title=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe9\" alt=\"SWASH: A Flexible Communication Framework with Sliding Window-Based Cache Sharing for Scalable DGNN Training\u63d2\u56fe9\" \/><\/figure>\n\n\n\n<p><strong>7. \u8d21\u732e\u4e0e\u7ed3\u8bba\uff08Contributions and Conclusion\uff09<\/strong><\/p>\n\n\n\n<p>SWASH\u7684\u4e3b\u8981\u8d21\u732e\u5305\u62ec\uff1a\uff081\uff09\u63d0\u51fa\u67d4\u6027\u901a\u4fe1\u6846\u67b6\uff0c\u652f\u6301\u901a\u4fe1\u6bd4\u4f8b\u3001\u65f6\u673a\u548c\u81ea\u9002\u5e94\u8c03\u5ea6\uff0c\u6709\u6548\u89e3\u51b3\u901a\u4fe1\u5f00\u9500\u95ee\u9898\uff1b\uff082\uff09\u8bbe\u8ba1\u8f7b\u91cf\u7ea7LightMetis\u5206\u533a\u7b56\u7565\uff0c\u9488\u5bf9\u6ed1\u52a8\u7a97\u53e3\u8bad\u7ec3\u7279\u70b9\u4f18\u5316\u901a\u4fe1\uff1b\uff083\uff09\u521b\u65b0\u6ed1\u52a8\u7a97\u53e3\u7f13\u5b58\u5171\u4eab\u673a\u5236\uff0c\u5728\u5206\u5e03\u5f0f\u73af\u5883\u4e0b\u5b9e\u73b0\u8de8\u7a97\u53e3\u5d4c\u5165\u590d\u7528\uff1b\uff084\uff09\u901a\u8fc7\u5927\u91cf\u5b9e\u9a8c\u9a8c\u8bc1\u7cfb\u7edf\u5728\u52a0\u901f\u548c\u7cbe\u5ea6\u4e0a\u7684\u53cc\u91cd\u4f18\u52bf\u3002<\/p>\n\n\n\n<p>\u8bba\u6587\u7ed3\u8bba\u6307\u51fa\uff0c\u6ed1\u52a8\u7a97\u53e3\u8bad\u7ec3\u867d\u63d0\u5347\u4e86DGNN\u6548\u7387\uff0c\u4f46\u73b0\u6709\u5206\u5e03\u5f0f\u6846\u67b6\u7684\u901a\u4fe1\u4e0e\u5206\u533a\u4e0d\u9002\u914d\u5df2\u6210\u4e3a\u4e3b\u8981\u74f6\u9888\u3002SWASH\u901a\u8fc7\u7cfb\u7edf\u6027\u521b\u65b0\uff0c\u6210\u529f\u5b9e\u73b0\u4e86\u901a\u4fe1\u4f18\u5316\u4e0e\u7cbe\u5ea6\u4fdd\u6301\u7684\u5e73\u8861\uff0c\u4e3a\u5927\u89c4\u6a21\u52a8\u6001\u56fe\u5b66\u4e60\u63d0\u4f9b\u4e86\u5b9e\u7528\u6846\u67b6\u3002\u8be5\u5de5\u4f5c\u4e0d\u4ec5\u5177\u6709\u91cd\u8981\u7684\u5b66\u672f\u521b\u65b0\u4ef7\u503c\uff0c\u4e5f\u4e3a\u5de5\u4e1a\u754c\u90e8\u7f72DGNN\u5e94\u7528\u63d0\u4f9b\u4e86\u53ef\u843d\u5730\u7684\u7cfb\u7edf\u53c2\u8003\uff0c\u672a\u6765\u53ef\u8fdb\u4e00\u6b65\u6269\u5c55\u5230\u66f4\u591a\u52a8\u6001\u56fe\u573a\u666f\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Proceedings of the ACM on Management of Data, Volume 3, Issue 3 (June 2025) h\ue03cps:\/\/doi.org\/10.1145\/3725360 EISSN: 2836-6573 1. \u6458\u8981\uff08Abstract\uff09 SWASH\u63d0\u51fa\u4e86\u4e00\u79cd\u4e13\u4e3a\u52a8\u6001\u56fe\u795e\u7ecf\u7f51\u7edc\uff08DGNN\uff09\u5206\u5e03\u5f0f\u8bad\u7ec3\u8bbe\u8ba1\u7684\u5168\u65b0\u901a\u4fe1\u6846\u67b6\uff0c\u6838\u5fc3\u76ee\u6807\u662f\u89e3\u51b3\u6ed1\u52a8\u7a97\u53e3\u8bad\u7ec3\u6a21\u5f0f\u4e0b\u901a\u4fe1\u5f00\u9500\u5927\u3001\u73b0\u6709\u5206\u533a\u7b56\u7565\u4e0d\u517c\u5bb9\u4ee5\u53ca\u7f13\u5b58\u590d\u7528\u6548\u7387\u4f4e\u4e09\u5927\u6311\u6218\u3002\u4f20\u7edf\u5206\u5e03\u5f0fDGNN\u8bad\u7ec3\u6846\u67b6\uff08\u5982 &hellip; <a href=\"https:\/\/www.ndnlab.com\/?p=1196\">\u7ee7\u7eed\u9605\u8bfb <span class=\"meta-nav\">&rarr;<\/span><\/a><\/p>\n","protected":false},"author":5,"featured_media":1207,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[5,23,1],"tags":[],"class_list":["post-1196","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-rengongzhineng","category-23","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/www.ndnlab.com\/index.php?rest_route=\/wp\/v2\/posts\/1196","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.ndnlab.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.ndnlab.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.ndnlab.com\/index.php?rest_route=\/wp\/v2\/users\/5"}],"replies":[{"embeddable":true,"href":"https:\/\/www.ndnlab.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=1196"}],"version-history":[{"count":1,"href":"https:\/\/www.ndnlab.com\/index.php?rest_route=\/wp\/v2\/posts\/1196\/revisions"}],"predecessor-version":[{"id":1208,"href":"https:\/\/www.ndnlab.com\/index.php?rest_route=\/wp\/v2\/posts\/1196\/revisions\/1208"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.ndnlab.com\/index.php?rest_route=\/wp\/v2\/media\/1207"}],"wp:attachment":[{"href":"https:\/\/www.ndnlab.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=1196"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.ndnlab.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=1196"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.ndnlab.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=1196"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}