You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 9 Next »

Scope:

  • The purpose of this page is to provide a process for submitting requests to Product Management to determine next steps, request inclusion in the product, or escalate for further approval. Requests will be submitted in a standard format (illustrated below) to the Product Manager, at which time they will be added to the agenda of the Product Management Meeting. Please make sure to include any stakeholders that will be impacted, as well as those necessary for reaching a clear decision.

New requests can be submitted to the Product Manger via email (micah@intel.com).

 

Product Management Meeting Agenda:

TBA

Product Management Current Initiatives:

Table of Contents:

NameDescriptionPriority
SAM (Service Assurance Management) Integration for IMLMonitoring features for IML that tie into Intel PlatformLow
Lustre Client for Xeon PhiAllows Xeon Phi to connect to lustre storage cluster

Med

Differentiated Storage ServicesAllows SSD Caching based on defined client profiles, caching done at MDS/OSSHigh
IML Installer ImprovementsAdd checks to validate install environment, improve robustness, and output functional error messages when issues are encountered 
   
   

 

 

Title

SAM (Service Assurance Management) integration for IML
TypeProof of Concept
DescriptionService Assurance Management is designed to provide QoS monitoring, metering, and performance analysis of nodes on a number of metrics (CPU, memory, IO) using a common platform node agent when used with IA platforms.
PurposeProvides increased feature set when used with Intel hardware, as well as giving more data about bottlenecks and server usage.
Next StepsAllocate a resource to work with the SAM team to setup a SAM scheduler and Node Collectors on an IEEL system and test basic functionality. Intended outcome would be to gather information about the usefulness of the data and potention benefits to IEEL product line.
PriorityLow
CreatorMicah Bhakti
Required AttendeesMicah Bhakti, Mrittika Ganguli (SAM team), Dan Ferber (or SE team resource)
StatusOpen

Title

Xeon Phi Supported Lustre Client
TypeFeature
DescriptionXeon Phi uses an embeded Linux OS on the card. We would like to add support for the Phi to access the Lustre FS based on Phi addoption in HPC as well as the Enterprise Technical Computing potential. To do this we will either need to support the Lustre client on the Phi Linux OS, or on the host and allow client Phi's to access the FS through the host.
PurposeProvides Lustre FS access to Xeon Phi co-processors.
Next Steps

Lustre 2.4 working on Phi with Ethernet (MPSS 2.1), IB support waiting on new OFED stack (WW39). TT-1346 will enable automatic Lustre build with RPM's, under review now.

Performance is still not where we want through virtual ethernet.

PriorityMed
CreatorMicah Bhakti
Required AttendeesMicah Bhakti, Dmitry Eremin, Oleg Droken, Peter Jones
StatusOpen

Title

Differentiated Storage Services
TypeFeature
DescriptionDSS runs as a service on the storage server (MDS or OSS) allocating data to tiered storage based on metadata. Profiles are created and client-specific, which allow data to be sorted to SSD or HDD storage based on the performance-optimized profiles.
PurposeImproves lustre performance for critical workloads, as well as potentially improving small-file IO.
Next Steps

DSS needs to be backported to current RHEL/CentOS kernel, followed by benchmarking and evaluation of test profiles for specific workloads. Following this we can define the impact and required engineering to integrate into the product.

PriorityHigh
CreatorMicah Bhakti
Required AttendeesMicah Bhakti, Michael Mesnier, Christian Black, Mikhail Pershin
StatusOpen

Title

IML Installer Improvements
TypeFeature
DescriptionProductization of IML installer including environment validation (checking to make server meets requirements) and error handling with clear user output.
PurposeIML install was not developed holistically and is not fully integrated into testing. The result is that it can be error-prone and difficult to troubleshoot when an issue is encountered. The purpose of these improvements would then be to deliver a robust installer which is fully validated, and has clear and "easy to troubleshoot" error reporting mechanisms.
Next StepsAdd to IEEL v1.1 scope for review and assignment.

 

PriorityHigh
CreatorMicah Bhakti
Required AttendeesMicah Bhakti, Michael Macdonald, Keith Mannthey
StatusOpen
  • No labels