analyze existing record data in netflix zones

[codefromthecrypt]

Relating to issue #7 we need to analyze all resource record types used across Route 53, UltraDNS and DynECT to ensure the model we design is as simple as possible yet covers all existing DNS records in Netflix. In this case, we should categorize washed information from live production.

Let's take care to hunt down the purpose of any outlying record types, such as those with private or provider-specific types (ex. 65280).

[codefromthecrypt]

Here's an analysis of records in UltraDNS

Type	Value	Count
A	1	1366
NS	2	749
CNAME	5	387
SOA	6	111
PTR	12	31
MX	15	287
TXT	16	65
AAAA	28	7
SRV	33	3

54s duration 3006 total records 113 POST requests

• 1 getAccountsListOfUser
• 1 getZonesOfAccount
• 111 getResourceRecordsOfZone

[codefromthecrypt]

Here's an analysis of records in DynECT

Type	Value	Count
A	1	4080
NS	2	2471
CNAME	5	54223
SOA	6	610
PTR	12	1608
MX	15	7
TXT	16	6
AAAA	28	976

Note that the api currently only works single threaded and often requires following a large number of redirects. Here are the statistics of commands needed to produce above:

6hrs 9mins duration 63981 total records 65823 GET requests

• 2 were getZone queries to enumerate the 610 zones
• 610 were AllRecord queries that enumerated the record ids in the zones
• 1219 noise from following redirects (meaning the results were not immediately available)
• 11 fudge I haven't looked into

[codefromthecrypt]

note the ultradns details above are only for one zone. I need to still do an exhaustive analysis.

[codefromthecrypt]

Here's an analysis of record sets in Route53

Type	Value	Count
A	1	25
NS	2	24
CNAME	5	335
SOA	6	6
TXT	16	18
A - Alias	1	1

The last record refers to an alias record set to IP addresses vs enumerating them directly.

4.5s duration 409 total records 9 GET requests

• 1 get all hosted zones
• 8 get rrset (2 of which continued list above 100 record sets)

[codefromthecrypt]

updated w/ultradns analysis