BlockServe is presented, a continuous batching framework that integrates block-grained scheduling -- immediately evicting completed requests at block boundaries -- with mixed-state execution that extends dual cache and parallel decoding to heterogeneous batches via gather-scatter indexing.
Abstract
Efficient serving of diffusion large language models (dLLMs) is hindered by convergence heterogeneity: when batching multiple requests, different sequences converge at different rates, causing faster requests to stall behind slower stragglers and introducing compute bubbles and tail latency. We present BlockServe, a continuous batching framework that integrates block-grained scheduling -- immediately evicting completed requests at block boundaries -- with mixed-state execution that extends dual cache and parallel decoding to heterogeneous batches via gather-scatter indexing. Furthermore, a compute-aware admission controller expands effective batch capacity through token-budgeted refill. On Dream and LLaDA across five benchmarks, BlockServe achieves 1.9--10.6$\times$ throughput over Fast-dLLM with comparable generation quality, establishing block-grained scheduling as a foundation for high-throughput offline dLLM inference.
DynamoServe is presented, a multi-tenant LLM serving framework that addresses challenges through three key innovations: leveraging stranded GPU memory to offload model weights and KV caches, mitigating resource fragmentation in multi-workload environments, and improving memory locality through coordinated data placemen...
Diman Zad Tootaghaj, Khaled Diab, Bob Lantz et al.· Conference on Applications,...· 0 citations
ExpertPlex is presented, which shares massive MoE experts across phases while disaggregating lightweight attention modules to eliminate over 95% of duplicate model weights and multiplexes dynamically sparse computation, while attention disaggregation reduces attention communication cost.
The authors co-design WIFiV-LPDDR, a wide-I/O LPDDR system for precision-tagged reads, BRQ-KV for a canonical low-rank-plus-INT8-residual prefix with query-dependent per-entry precision, and DAT-FFN for drift-mapped canonical replacement, adjacent-stage-corrected low-bit delta, or cached-state carry while keeping live...
Wei-Hsing Huang, Kiseok Lee, Ming-Yen Lee et al.· 0 citations
With rising popularity of LLMs, the performance, scalability, and resource-efficiency of inferences become a crucial challenge. The core part of the inference process is the KV cache, which avoids recomputing intermediate attention states, and the batching strategy that batches multiple requests per forward pass to lev...
Jie Ye, Avinash Maurya, Krishna Teja Chitty-Venkata et al.· IEEE International Symposium...· 1 citation
A mathematical scheduling model that connects within-batch resource fairness to system throughput and provides a bi-criterion scheduling policy, ISJL, which maintains high throughput while aligning max-driven batch cost with token-metered revenue.
Or OrionInfer, an adaptive LLM serving system that aligns inference strategies with real-time demand and introduces three key techniques: runtime switching between data parallelism and tensor parallelism with negligible overhead, an efficient inference pipeline that preserves batching efficiency during parallelism tran...
Jingqi Feng, Guang Yang, Yukai Huang et al.· Proceedings of the 32nd ACM...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
