I am sponsoring this case for Bing Zhao. It requests a patch
binding for iSER, updating PSARC 2008/395 as a result.
I do not expect this to be controversial so I have marked it
Self Review, but am happy to start a timer if an ARC member
so desires or significant discussion ensues.
-- mark
Template Version: @(#)onepager.txt 1.35 07/11/07 SMI
Copyright 2008 Sun Microsystems
1. Introduction
1.1. Project/Component Working Name:
iSER: iSCSI Extensions for RDMA
1.2. Name of Document Author/Supplier:
Bing Zhao
1.3. Date of This Document:
1/08/10
1.3.1. Date this project was conceived:
8/1/07
1.4. Name of Major Document Customer(s)/Consumer(s):
1.4.1. The PAC or CPT you expect to review your project:
Solaris PAC
1.4.2. The ARC(s) you expect to review your project:
PSARC
1.4.3. The Director/VP who is "Sponsoring" this project:
scott.tracy at sun.com
1.4.4. The name of your business unit:
Software
1.5. Email Aliases:
1.5.1. Responsible Manager: peter.buckingham at sun.com
1.5.2. Responsible Engineer: peter.dunlap at sun.com
1.5.3. Marketing Manager: margaret.hamburger at sunc.om
1.5.4. Interest List: iser-dev at opensolaris.org
2. Project Summary
2.1. Project Description:
iSCSI Extensions for RDMA (iSER) accelerates the iSCSI protocol by
mapping the data transfer phases to Remote DMA (RDMA) operations.
As a result an iSER initiator should be able to read and write data
from an iSER target at high data rates with relatively low CPU
utilization compared to iSCSI using TCP/IP. This project will implement
both an iSER initiator and an iSER target. In addition, the project will
implement an iSCSI port provider (iscsit) for the SCSI Target Mode
Framework (STMF) introduced by PSARC 2007/523 adding an iSCSI transport
option to the existing Fibre Channel support. The iSER target will be an
extension of this iSCSI port provider.
2.2. Risks and Assumptions:
This project does not replace the existing iSCSI target (PSARC 2005/441)
although it is intended to become the preferred iSCSI target
implementation since future SCSI target development will be focused
around STMF.
3. Business Summary
Solaris currently lacks a high performance low latency storage protocol
for Infiniband networks. This is inhibiting the adoption of Solaris in
the HPC space which relies heavily on Infiniband.
3.1. Problem Area:
Need an Infiniband storage protocol
3.2. Market/Requester:
Software
3.3. Business Justification:
3.4. Competitive Analysis:
Linux and Windows provide both iSER and iSCSI initiators. Linux
provides an iSCSI target implementation as well.
3.5. Opportunity Window/Exposure:
3.6. How will you know when you are done?:
Both iSER and software iSCSI transport should provide comparable
performance to Linux and Windows offerings on the same hardware as
well as interoperating with those same Linux and Windows
implementations.
4. Technical Description:
4.1. Details:
iSER will be implemented as a pluggable transport capability so
that a modified iSCSI initiator and the new iSCSI target can use
either TCP/IP transport or iSER (via Infiniband) transport.
The case materials describe the approach in more detail.
4.2. Bug/RFE Number(s):
6702584 Need iSCSI port provider for COMSTAR
6702591 COMSTAR iSCSI port provider needs to support iSER transport
4.3. In Scope:
Standards compliant iSER initiator and target implementations
Standards compliant iSCSI target implementation
4.4. Out of Scope:
Integration between the new iSCSI target and ZFS (a future project
should address this)
4.5. Interfaces:
Imported:
libstmf
STMF port provider API
Exported:
itadm CLI for iSCSI target management
libiscsit
4.6. Doc Impact:
COMSTAR/STMF documentation will need modifications to describe the
new iSCSI transport capability.
iSCSI documentation will need to be modified to describe
configuring iSER connections.
4.7. Admin/Config Impact:
The iSCSI administrative tools (iscsiadm(1m) and itadm(1m) in this
case) are not expected to require any modifications for iSER.
iSER transport is automatically selected for iSCSI connections
established on an iSER-capable network.
4.8. HA Impact:
N/A
4.9. I18N/L10N Impact:
N/A
4.10. Packaging& Delivery:
No modifications to existing packages
4.11. Security Impact:
The iSCSI target will listen for new iSCSI connections on any port
configured by the user (default is 3260).
4.12. Dependencies:
No dependencies on other projects
5. Reference Documents:
iSER/iSCSI COMSTAR OpenSolaris project page:
http://www.opensolaris.org/os/project/iser
RFC3720: Internet Small Computer Systems Interface (iSCSI)
RFC5046: Internal Small Computer System Interface (iSCSI)
Extensions for Remote Direct Memory Access (RDMA)
RFC5047: DA: Datamover Architecture for the Internal Small
Computer System Interface (iSCSI)
PSARC 2007/523 COMSTAR: Common Multiprotocol SCSI Target
PSARC 2007/153 iSCSI target provider
PSARC 2006/622 iSCSI/ZFS integration
PSARC 2005/441 iSCSI Target Project
6. Resources and Schedule:
6.1. Projected Availability:
Q1FY09
6.2. Cost of Effort:
5 people for 12 months
6.3. Cost of Capital Resources:
$100,000
6.4. Product Approval Committee requested information:
6.4.1. Consolidation or Component Name:
NWS
6.4.3. Type of CPT Review and Approval expected:
Standard
6.4.4. Project Boundary Conditions:
N/A
6.4.5. Is this a necessary project for OEM agreements:
No
6.4.6. Notes:
6.4.7. Target RTI Date/Release:
9/22/08, targeting build 100
Solaris 10 Update 9
The iSER initiator is required to be backported to S10U9 and
needs a patch binding
6.4.8. Target Code Design Review Date: 7/31/08
6.4.9. Update approval addition:
N/A
6.5. ARC review type:
Standard
6.6. ARC Exposure:
Open
7. Prototype Availability:
7.1. Prototype Availability:
iSCSI target prototype was available 4/25/08
iSER initiator and target should be available 7/1/08
7.2. Prototype Cost:
N/A
