Computer Sci./20023521/Hyoundeock Yim/hdyim@cosmos.kaist.ac.kr/011-9047-4109

1. Title 
   "Performance measurement of Internationalized  Top Domain Name"

2. Motivation
 1) K root name server(NS) processes maximum about 4,100 queries/s. [RIPE 02]
     Generic top level domain(gTLD) NS processes more queries than root NSs. [N. Brownlee et al. 01] 
 2) The size of a root NS zone file : 51KB [Internic 02]
     The size of a gTLD NS .com zone file : over 2GB [N. Brownlee et. al. 01]
 3) Internationalized domain name(IDN) can support 32 languages. [VeriSign 02]
     The length of encoded domain name become longer. [Draft-IDNA 02] 
     => The size of a root/TLD NS zone file can be increased by over a factor of 32
 4) The number of hosts increases by 10% every year. [ISC 02]
     IPv6 with more IP addresses than IPv4 will be introduced in the future.
     => The Root/TLD NS will receive more queries from increased clients by a factor of about 2. 
 5) Approximately, burden of the root NS will be increased by over a factor of 64(=32 * 2)
     => A search time for queries of clients will be increased in the root/TLD NS

3.Problem Definition
 1) IDN may make the existing root/TLD NS performance worse due to the increase of the size of zone file and clientsĄŻ query.
     Ex) How fast does the gTLD NS reload 128GB .com zone file twice a day for update?
           How long does it take to resolve client's query in 128GB .com zone file?

 => This simulation will measure and analyze this problem of IDN comparing with the existing domain name system.

4.Approach
 1) Assumption
   i. All root/TLD NSs use BIND(Berkeley Internet Name Domain)
   ii. NS itself performance test without consideration of network state such as congestion.
 2) Considerations
   i. Query pattern ? maximum, average, and minimum traffic as time changes
   ii. Queries population
 3) Testbed setup
   i. Root/TLD/local NS, IDN-supporting Client, and zone file design
 4) List of experiments to compare existing DNS with IDN system
   (All tests will be experimented as TTL = 0 to prevent positive caching  and use DIG for scropt)
   i. Resolution time by change of zone file size(the number of resource records)
   ii. Resolution time per the number of queries
   iii. The number of processed queries per the number of queries
   iv. Resolution time per the length of encoding domain name
   v. The number of processed per the length of encoding domain name
   vi. H/W ? CPU and RAM
    vii. BIND version

5.Schedule
   10/28 Second presentation
   11/03 NS setup
   12/14 Measurement and analysis
   12/15 Term paper due

6.Deliverable
   A measurement  report

7.Reference list
1) [N. Brownlee et al. 01] Nevil Brownlee, kc claffy, Evi Nemeth, DNS Root/gTLD Performance Measurements, LISA 2001
2) [RIPE 02] RIPE-NCC summary for k.root-servers.net, Oct 2002
    http://k.root-servers.org/ktotal/
3) [Internic 02] Root zone file archiving site, Oct 2002
    ftp://ftp.rs.internic.net/domain/root.zone.gz
4) [VeriSign 02] VeriSign IDN datasheet, 2002
    http://www.verisign-grs.com/idn/idn_datasheet.pdf
5) [ISC 02] Internet Domain Survey, Jul 2002
     http://isc.org/ds/www-200207/index.html
6) [Draft-IDNA 02] Internationalizing Domain Names in Applications (IDNA), Oct 4, 2002 
     http://www.ietf.org/internet-drafts/draft-ietf-idn-idna-13.txt