[alto] Review of draft-xiang-alto-exascale-network-optimization-01

[Yichen Qian <[e-mail]>]

Hi Qiao, all,

It’s interesting to introduce I/O and computation resources into ALTO to get not only the network resources for more accurate query. Here is my review of the draft:

>>> It is mentioned in chapter 4.2 (“Sharing raw site/cluster information would violate sites' privacy constraints.”) and 6 (“How much privacy, including… will be exposed?”) that the privacy is a problem in providing the information, but it seems that it is not addressed in this draft.

Abstract
"In this document, we propose that it is feasible to use existing ALTO services to provides not only network information, but also information about other resources in science networks including computing and storage."

>>> to provides -> to provide

"ExaO provides simple APIs for users to submit and manage dataset transfer and analytic requests and to monitor the status of each request, along with fine-grained local and global network and site state information in real-time. "

>>> Too many “and” to read and understand.
 
1.  Introduction
"Applications such as the Production ANd Distributed Analysis system (PanDA) in ATLAS and the Physics Experiment Data Export system (PhEDEX) in CMS have been developed to manage the data transfers among different cluster sites."
   
>>> have been -> has been

"Section 6 lists several key issues to address in order to realize the proposal of providng multi-resource information by ALTO topology services."

>>> providng -> providing

5.2.  Example: encode storage bandwidth into path vector
"Other than network resource, assume in this topology eh1 and eh3 are equipped with commodity hard drive disk (HDD) while eh2 and eh4 are equipped with SSD."

>>> There is the full name of HDD but no full name for SSD.
   
"In this example, if we see the end hosts as network elements, the storage I/O bandwidth of each host can also be encoded as an abstract element into the path-vector."

>>> It’s better to have a new figure with the storage I/O bandwidth as an abstract element in the topology.

6.  Key Issues
"a large dataset transfer or analytic application always involve many network elements in multiple clusters/sites and the absolute number of involved network elements keep increasing as the scale of clusters increase."

>>> involve -> involves

"There still lacks of an analytics or experimental understanding on the scalability of path-vector and RSA services."

>>> lacks of -> lacks

7.1.  Architecture
"including replica selection, routing path computation and bandwidth allocation, and request parallelization decisions, such as which cluster each sub-request should be placed at the multi-resource orchestrator."

>>> missing a preposition before “the multi-resource orchestrator” -> in the multi-resource…

7.2.1.  User API

>>> Another API that allows a user to query all the requests that the user has submitted can be considered in case of missing the requestID.

7.4.  ALTO Server
"Each ALTO server must provide basic information services as specified in [RFC7285] such as network map, cost map, endpoint cost service (ECS), etc.  To support the fine-grained multi-resource allocation in ExaO, each ALTO server should also provide more fine-grained information about different resources in clusters through ALTO extension services such as the routing state abstraction [DRAFT-RSA], path vector [DRAFT-PV], network graph [DRAFT-NETGRAPH],multi-cost [DRAFT-MC] and cost-calendar [DRAFT-CC] services."
				
>>> Providing basic information along with fine-grained information may has some overlap or redundant information such as information provided by cost map and RSA. It may increase the burden of data transfer. Some ways to reduce the redundant or send part of the information can be considered.

7.7.1.  Orchestration Algorithms
"The modular design of ExaO allows the adoption of different orchestration algorithms and methodologies, depending on the specific performance requirements."
>>> If using a specific algorithm, what interface (inputs/outputs…) should users specify?

7.7.3.  Example: A Max-Min Fairness Resource Allocation Algorithm

>>> As an example, I think there is too much introduction and description for MFRA. May be it can be seen as a default algorithm in the system?

8.3.  Constraints of the MFRA Algorithm
"Simply denoting the replica selection as a set of binary constraint will significantly increases the computation complexity of the scheduling process."

>>> increases -> increase