The Design and Implementation of a Next Generation Name Service for the Internet

1. The Design and Implementation of a Next Generation Name Service for the Internet Leo Bhebhe leo.bhebhe@nokia.com

2. Contents • Introduction • Name Servers and Name Resolution • Current Issues With Domain Name System (DNS) • Cooperative Domain Name System (CoDoNS) • Performance measurements • Summary/Conclusions

3. Introduction • Analysis of the current Domain Naming System (DNS) and looks at the limitations of the DNS structure, bottlenecks and performance issues • Proposal of a new Cooperative Domain Naming System (CoDoNS) to replace the old

4. Name Servers and Name Resolution VU CS Edu Yale Yale CS Originator name server name server name server name server 1 2 3 4 cs.yale.edu yale.edu cs.vu.nl edu.server.net 8 7 6 5 flits.cs.vu.nl

5. Current Issues With Domain Name System (DNS) Network failures • Susceptible to denial of service attacks (DoS) • Small number of nameservers serve domains, limited redundancy in name servers • At network level • 80% of domain names are served by just two servers • 0.8% by one • 32 servers are connected to the internet by a single gateway (serious outage or DoS) • At top level hierarchy • Small number of servers (targets for DoS) • Recent DoS attack severely affected the availability of Microsoft's web services for hrs. • Checked by performing trace routes to 10000 different nameservers which serve 5000 randomly chosen names from 50 global distributed sites on Planetlab

6. Current Issues With Domain Name System (DNS) Network failures • Failure Resilience-Implimentation Errors • 20% of name server implementations contains security flaws • 18% of servers don’t respond to version querries • 14% do not report valid BIND versions • 2% of the nameservers have the tsig bug which permits a buffer to overflow that can enable malicious agents to gain access to the system • 19% of the servers have the negache problem that can be exploited to launch the DoS attack by providing negative responses with large TTL value from a malicious server • Checked by surveying 150000 servers based on the Berkely Internet Name Daemon (BIND) to see if any known vulnerabilities are available

7. Current Issues With Domain Name System (DNS) Performance - System Latency makes it unsuitable for dynamic updates. • Name address translation in the DNS incurs long delays. The legacy DNS incorporates aggressive caching in order to reduce latency of query resolution • But explosive growth of namespace has decreased the effectiveness of caching • Short timeouts (TTL)-reduces DNS cache hit rates • Increase in load imbalance - at root and TLD nameservers handle a large load • Configuration error such as broken (lame) or inconsistance delegations - can introduce latent performance problems • Update propagation-Large scale caching -maintaining the consistance of cached records in the presence of dynamic change (TTL?)

8. Cooperative Domain Name System (CoDoNS) • CoDoNS derives its performance characteristics from a proactive caching layer called Beehive • Automatically replicates the DNS mappings throughout the network to match anticipated demand and provides a strong performance guarantee. • Achieves a targeted average lookup latency with a minimum number of replicas • Beehive is a proactive replication framework that enables prefix-matching DHTs to achieve O(1) lookup performance • Pastry and Tapestry are examples of structured DHTs that use prefix-matching to look up objects Example • If the identity of the record is e.g. 110011, Pastry would store the value in the host with the identity closest to that (the home node) • If proactive caching is used for one level, then the record is copied to all hosts whose identity begins with 11001* and • For level 2, 1100** etc. • The home node receives periodically information about the usage of the record and makes decisions on how many levels to cache the record. • Thus most often used records can be disseminated to almost all nodes while rarely used can be stored in relatively few places

9. Cooperative Domain Name System (CoDoNS) • Replicating every object at every node would achieve O(1) lookups BUTwould • Incur excessive space overhead • Consume significant bandwidth • Lead to large update latencies • Beehive minimize bandwidth by posing the following optimising problem • Minimize the total number of replicas subject to constraint that the aggregate looup latency less than a desired constant C • For power law (or Zipf-like) query distributions, Beehive analytically derives the optimal closed solution to this problem. • The final expression that minimizes the total number of replicas for Zipf-like query with parameter ά<1 is the following • b is the base of the underlying DHT, Xi is the fraction of most popular objects that get replicated at level I • Selecting the appropiate C enables applications to achieve any targeted average. lookup latency

10. Cooperative Domain Name System (CoDoNS) • CoDoNS architecture • Consists of globally distributed nodes that self organise to form a peer-to-peer network • CoDoNS associates the node whose identifier is closest to the consistent hash of the domain name as the home node for that domain name • The home node stores a permanent copy of the resource records owned by that domain name and manages their replication • If the home node fails, the closest node in the identifier space automatically becomes the new home node • CoDoNS replicates all records of several nodes adjacent to the home node in the identifier space in order to avoid data loss due to node failures • It can serve as a backup for legacy DNS, as well as a complete replacement

11. CoDoNS Deployment and The Process of Query Qesolution • Clients send DNS quiries to the local CoDoNS server • The local CoDoNS server obtains records from the home node or an intermediate node • The local CoDoNS server then responds to the client • In the background, the home node interacts with the legacy DNS to keep records fresh and propagate updates to cached copies

12. Issues and Implications • CoDoNS uses crypto-graphic delegations and self-verifying records based on the DNSSEC standard. • DNSSEC uses public key cryptography to enable authentication of resource records. • Every namespace operator has a public-private key pair • The private key is used to digitally sign DNS records managed by that operator • The corresponding public key is in turn certified by a signature from a domain higher up in the hierarchy. • The signature and the public key are stored in DNS as resource records of type sigand key respectively. • The use of cryptographic certificates enables any client to check the verity of a record independently, and keeps peers in the network from forging certificates. • To speed up certificate verification, CoDoNS servers cache the certificates along with the resource records and provide them to the clients.

13. Cumulative Distribution of Look Up Latency • CoDoNs achieves low latencies for name resolution • More than 50% of quries incur no network delay as they are answered from the local CoDoNS cache • This is because proactive replication pushes responses for the most popular domain names to all CoDoNS servers

14. Median latency • CoDoNS latency decreases significantly as proactive caching takes effect in the background • Initially, CoDoNS servers have an empty cache and redirect most of the queries to legacy DNS. • Consequently, they incur higher latencies than the legacy DNS. • But as resource records are fetched from legacy DNS and replication in the background pushes records to other CoDoNS servers, the latency decreases significantly.

15. Median resolution lookup latency in CoDoNS • Flash crowd is introduced at 6 hours • CoDoNS detects the flash crowd quickly and adapts to the amount of caching to counter it while continuing to provide high performance • Beehive’s proactive replication in the background detects the changes in popularity, adjusts the number of replicas, and decreases the lookup latency.

16. Load Balance quantifying load balancing using • CoDoNs handles flash-crowds by balancing the quiery load uniformly across node • The graph shows the standard deviation to the mean across all nodes • At the start of the experiment, the query load is highly unbalanced, since home nodes of popular domain names receive far greater number of queries than average. • The imbalance is significantly reduced as the records for popular domains get replicated in the system. • Even when a flash-crowd is introduced at the six hour mark, dynamic changes in caching keep the load balanced after a temporary increase in load variance.

17. Update Propagation Time • CoDoNS incurs low latencies for propagation updates. • 98% of replicas get updated within one second • It takes a few seconds longer to update some replicas due to high variance in network delays and loads at some hosts. • The latency to update 99% of replicas one hop from the home node is about one second. • Overall, update propagation latency in CoDoNS depends on the extent of replication of records

18. Summary/Conclusions • Performance measurements from a planetary-scale deployment against a real workload indicate that CoDoNS can provide low latencies for query resolution. • Massive replication for the most popular records, but a modest number of replicas per server, achieves high performance with low overhead. • Eliminating the static query processing hierarchy and shedding load dynamically onto peer nodes greatly decreases the vulnerability of CoDoNS to denial of service attacks. • Self organization and continuous adaptation of replication avoids bottlenecks in the presence of flash crowds. • Proactive update propagation ensures that unanticipated changes can be quickly disseminated and cached in the system • Cooperative Domain Name System (CoDoNS) is proposed as an alternative for DNS