Hi Qin, thanks for the discussion on this proposal. Please see answers inline. Any feedback on this work in progress is more than welcome. Best regards, Sabine and Luis
From: alto <[email protected]> On Behalf Of Qin Wu Sent: Wednesday, November 2, 2022 8:26 AM To: IETF ALTO <[email protected]> Cc: [email protected] Subject: [alto] Review of draft-contreras-alto-service-edge-06 Hi, Luis: Thank for the update to this draft, it is well written. but I have a few comments which help you prepare slides for discussion in IETF 115 1. when one single entity belongs to two different domains, such entity in different domain can have different properties, e.g., when in IPv4 domain, we use IPv4 address to locate a server, when in compute domain, we use identifier such as DC10-HOST1 to locate the same server, I am wondering whether you can obtain all these properties in different domains using Cloud Infrastructure Managers? Indeed, "ane:DC10-HOST1" is unknown in the ipv4 domain while "ipv4:X.X.X.X" is just a string in the ane domain. This draft precisely proposes a way for a CIM to retrieve information on a single entity, say for example a server, that is identified in several different entity domains, by: - (i) introducing property "entity domain mappings" that exposes the other entity domains where this entity is identified, - (ii) introducing specific properties to directly provide network addresses such as "network-address". It is not intended to provide all the specific properties of other domains, but since ALTO is about network exposure, systematic use of "network-address" seems to be desirable. This is illustrated in example of section 4.1 2. Can you provide an example of such application which integrate compute info with network information, e.g., network resource can be allocated based on scale up or down of compute resource in the data center? The current use case this drafts focuses on is the selection of an edge server based on both path cost to this server and compute capabilities of this server 3. For section 4.2, how do you deal with computing resource dynamic change in the run time? or you just focus on service placement scenario, where you only care about the total capacity of the compute resource which is static? For the moment/in this draft focuses on service placement. Compute resource changes can be reported by an ALTO server with incremental updates (RFC 8895) on properties or the extensions defined in [draft-ietf-alto-new-transport]. -Qin
_______________________________________________ alto mailing list [email protected] https://www.ietf.org/mailman/listinfo/alto
