Loading...

Design and Implementation of Hardware Accelerator for Domain Name Service

Jahandar, Ebrahim | 2016

651 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 49056 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Jahangir, Amir Hossein
  7. Abstract:
  8. In this project we have designed & implemented a hardware accelerator for domain name service. This hardware accelerator is compatible with existing designs and it could be used standalone as an authoritative DNS server or a hardware accelerator in series of an existing DNS server facility. Two goals, are achieved in this thesis: increasing total DNS throughput and decreasing its response time.In this project, we have surveyed about domain name service, its scientific measurements, theory of caching and its effectiveness, name lookup methods and finally some of the similar designs. Domain name lookup in memory is one of the most challenging operations in every DNS server. We have researched about the name lookup methods and their implementation challenges and proposed a new name lookup method based on randomized algorithms and hash table. We have compared our proposed design with similar methods such as cuckoo hash, bloom filter and legacy hash tables. We showed that our method in comparision with “cuckoo hash” has a higher throughput in multicore systems and the same throughput in single core system, but with 2 to 5 times less memory usage.We have implemented our proposed name lookup method in a general and configurable way. For storing 100000 domain names with their 32 bit IP addresses, our method needs 8 Mbit of memory, about 3 times less than cuckoo hash, Our design has achieved about 1.5 million lookup per second with low price CPU and average name lengths of 128 characters, which is the same as cuckoo hash, but with high-end multicore system our method has achieved lookup per second 2 times higher than “cuckoo hash”.We have implemented the lookup method in GPU and tested it with 3 NVidia GPU’s. Evaluations showed that GPU could increase lookup method 5 to 20 times compared with high-end CPU configurations.Our GPU based DNS hardware accelerator communicates directly with network interface cards and receives Ethernet frames directly. After deconstructing DNS packet it resolves a batch of query domain names in GPU and finally, respondes back to sender. It could meet 400k to 3M query per second with single high-end system, ie. 5 to 20 times better than its software alternative
  9. Keywords:
  10. Hash Table ; Hardware Accelerator ; Domain Name System (DNS) ; Graphic Processing ; Graphics Procssing Unit (GPU) ; Memory Lookup ; Cluster Computer

 Digital Object List

 Bookmark

No TOC