Title :
Maintaining exact statistics counters with a multi-level counter memory
Author :
Roeder, Michael ; Lin, Bill
Author_Institution :
Dept. of Electr. & Comput. Eng., California Univ., San Diego, La Jolla, CA, USA
fDate :
29 Nov.-3 Dec. 2004
Abstract :
For monitoring and measuring high-speed networks accurately in real-time, a large number of statistics counters may need to be maintained at wirespeeds (e.g. 10 Gbit/s), Expensive, but fast. SRAM is needed for storing the counters to satisfy the speed requirements. However, high-density, but slower, DRAM is needed to provide the necessary storage capacity for storing all counter values exactly. Recent papers by Shah et al. (2003) and Ramabhadran and Varghese (2003) have addressed the problem using counter memory architectures based on one level of fast SRAM for storing partial counter values and a high-capacity DRAM for storing full counter values. In this paper, we propose to extend their work with a multi-level counter memory architecture to reduce the amount of fast memory required. Our multi-level counter memory architecture can reduce the amount of equivalent fast memory storage required by as much as 28%.
Keywords :
DRAM chips; SRAM chips; memory architecture; monitoring; statistics; telecommunication network management; 10 Gbit/s; DRAM; SRAM; exact statistics counters; high-speed network measurement; memory architecture; monitoring; multi-level counter memory; reduced fast memory storage; CADCAM; Computer aided manufacturing; Computer networks; Counting circuits; High-speed networks; Maintenance engineering; Memory architecture; Monitoring; Random access memory; Statistics;
Conference_Titel :
Global Telecommunications Conference, 2004. GLOBECOM '04. IEEE
Print_ISBN :
0-7803-8794-5
DOI :
10.1109/GLOCOM.2004.1378029