1 / 21

DiPIT : a Distributed Bloom-Filter based PIT Table

DiPIT : a Distributed Bloom-Filter based PIT Table. Authors: Wei You, Bertrand Mathieu, Patrick Truong, Jean-Francois Peltier ,and Gwendal Simon Publisher: ICCCN 2012 Presenter: Chai-Yi Chu Date: 2013/04/10. Outline. Introduction DiPIT Architecture PITi shared Bloom filter

cayla
Download Presentation

DiPIT : a Distributed Bloom-Filter based PIT Table

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. DiPIT: a Distributed Bloom-Filter based PIT Table Authors: Wei You, Bertrand Mathieu, Patrick Truong, Jean-Francois Peltier ,and GwendalSimon Publisher:ICCCN 2012 Presenter: Chai-Yi Chu Date: 2013/04/10

  2. Outline • Introduction • DiPITArchitecture • PITi • shared Bloom filter • Evaluation

  3. Introduction • Propose a new implementation for one of the main components of CCN nodes, the Pending Interest Table (PIT). • PIT needs a large memory space to store the pending Interests. • Cut the PIT table into several sub-tables, which we call PITi. • Call DiPITfor Distributed PIT

  4. DiPIT Architecture • Use counting Bloom filters • faster than hash tables for lookup and update operations. • be more efficient in memory space. • deal with information removal. • when a Data packet is forwarded, the related Interest entry should be deleted from the PIT table.

  5. PITi: One Bloom filter per CCN face • works independently. • records in a counting Bloom filter the footprints of the Interests packets that come from the associated face. • Incoming Data packets are checked in parallel on all PITis and forwarded on the faces when the associated PITi has a matching footprint.

  6. A shared Bloom filter: Deal with false positive • Adding another filter after the first Interest verification process. • The total false positive ratio is the product of the two individual false positive ratios of the two filters. • Returned Data packets do not remove entries in the shared Bloom filter • refresh the shared Bloom filter sometimes. • The number of inserted elements reaches a threshold • A regular basis

  7. Main DiPITalgorithms • Incoming Interest: • (1) a negative result means that the Interest never came in. It is forwarded to the FIB and its footprint is added in the filter; • (2) a positive result means that either the Interest has already come, or it is a false positive. It is then checked in the secondary shared Bloom filter; • (2a) if the second filter gives a negative answer, the Interest is forwarded to the FIB and its footprint is added in the second filter; • (2b) on the contrary, a positive result means that this Interest is a duplicated emission, the CCN node blocks it.

  8. Incoming Data: • Being verified in all the PITis, except the one where the Data comes. • If a PITicontains a matching Interest footprint, the Data packet is forwarded through this face and the footprint should be deleted from the PITi.

  9. Evaluation • Assume one networking line card holding 16 interfaces. • The average Interest arrival rate ranges from 20 Mpcksto 200 Mpcks.(Million packet per second) • Data packet RTT time as 80 ms. • 5 hash functions for the counting Bloom filters, each counter has 3 bits. • The shared Bloom filter has a size of 1 Mbits. • Acceptable false positive in the range from 0.1% to 1%.

  10. According to Little’s Law, the biggest number of elements in the PIT table is , which can be represented by . • Thus for the centralized hash table we used H-bit of 28 bits. We also add 32 bits H-bit in the valuations. • Assume 40 Bytes Interest packet. • Take the middle value and give 128 bits for each ContentName matching in hash table. • 2 Bytes to memorize the incoming interface identifiers.

More Related