SlideShare a Scribd company logo
1 of 24
Download to read offline
© Hitachi, Ltd. 2018. All rights reserved.
06/05/2018
Keiichi Matsuzawa*†
Mitsuo Hayasaka*
Takahiro Shinagawa†
The Quick Migration of File Servers
* Hitachi, Ltd.
† University of Tokyo
© Hitachi, Ltd. 2018. All rights reserved.
Background
• Most cases of installing new file servers require server
and data migration across file servers
Source: Storage Magazine, Vol.16, No.2, 2017
20%
24%
32%
39%
56%
1
2
3
4
5
What are the top drivers
for your new NAS system?
Improve performance for existing apps
Replace old hardware
Deploy new application
Support virtual server better
Use as a backup target
Migrate from
existing old server
1 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Challenges for migration scheme
1. Independent of the source server
– Heterogenous file server migration is needed to:
• Change to a more attractive vendor and product
• Dispose of End-of-Life products
• Aggregate small file servers to a enterprise file server
– Should not modify the source server
2. No long downtime for clients (<30 seconds)
– Common OSs define network file access timeout about tens
of seconds
3. No performance overhead after migration
– Improving performance is one of top drivers
2 / 23
© Hitachi, Ltd. 2018. All rights reserved.
• Pre-copy data per file [rsync][robocopy]
– Tracking the last updates takes long time to traverse tree
• Copy disk volume [SnapMirror]
– Source and target must share the same disk layout
Existing solutions
Source Target
Cannot mount
volume
>1day downtime
to track the last
update
rsync
Source Target
3 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Our approach
• Post-Copy per file
– Clients change connection before copying data
– Target server copies metadata and data on demand
• Transfer data only via common interface (ex. NFS)
X
5: return
3: read(“/A/X”)
from target
Source Target
A
root
A
X
root
B
Y
4: copy(“/A/X”)
Client
1: read(“/A/X”)
from source
2: remount
4 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Data migration method
• Migrate whole file namespace by:
1. On-demand migration
– Copies objects accessed by clients on-demand
2. Background migration
– Copies un-accessed objects in background
3. Stub-based file management:
– Enables intermediate migration state and supports above
migrations
5 / 23
© Hitachi, Ltd. 2018. All rights reserved.
On-demand migration
• Copy directories and files recursively when accessed
• After preparing root directory, any files are accessible
– Accessing directory only generates intermediate state
objects to reduce time to copy
root
A B C
root root
A B C
X Y Z
root
A B C
X Y Z
lookup “/” lookup “/A/” lookup “/A/Y”
read( “/A/Y” )
Create objects without
data
6 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Background migration (1)
• Copy un-accessed files by clients
• Crawling program traverses file namespace and
induces on-demand migration
– Should run when access load is low (ex. midnight or
weekend)
root
A B C
X Y Z
root
A B C
X Y Z
root
A B C
X Y Z
Run crawling
program
7 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Background migration (2)
• Crawling program may run intermittently
– Copy huge file share across several nights
– Need to handle updated files in the daytime
• Restart from the root directory each time
– Count intermediate objects until reaching zero
root
A B C
X Y Z
root
A B C
X Y Z
root
A B C
X Y Z
count: 4 count: 2 count: 0
8 / 23
© Hitachi, Ltd. 2018. All rights reserved.
• Should consider characteristics of file system:
• Data structure
– Hierarchical structure
– Protocol dependent metadata (ex. ACL, uid)
• Operations
– Rename/move objects
– Hardlinks
– Truncate
• Client behaviors (ex. Return code handling, thumbnail
preview)
• Product dependent features (ex. Compression, quota)
Issues of post-copy per file (1)
Scope of this
presentation
9 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Issues of post-copy per file (2)
• Cause timeout to create inodes under huge directory
– Retrieving metadata requires many packet exchanges and
takes long time
A
X
read(“/A/Y”)
Source Target
Y Z …
A
Target
A
X Y Z …
Create all at once?
10 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Issues of post-copy per file (3)
• Keep consistency of source and target objects even
after move operation
root
X
move(“/B”,”/A/C”)
Source Target
A B
Y
root
X
A B
Target
root
X
A
C
Different path for
source and target
Y
C
11 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Stub management
• Concept used in Hierarchical Storage Management
– Stub is a virtual object does not have data contents but
stores actual data location
– File system intercepts stub accesses and inserts additional
operation
• Apply the concept for file server migration
– Allow intermediate states storing partial data and metadata
– Retrieve un-copied data when stub is accessed
12 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Dummy Partial Normal
File and directory status
• Define intermediate states storing partial (meta)data
– Defer (meta)data copy until the file is accessed
– Dummy object can be created by only parent directory entry
– Partial object has full metadata and parts of data blocks
• Regards hole blocks as un-copied blocks
– Normal object has full metadata and data
• Migration is completed and not produce any source server accesses
Stub info.
Metadata Metadata
Data
inodeAccess
whole
data
Access
metadata
Access
parent
directory
Data
Intermediate states
inode
Stub info.
inode Internal
data
structure
Migration completedMost attributes are empty
13 / 23
© Hitachi, Ltd. 2018. All rights reserved.
• Additional metadata to manage intermediate states
• Stores full path name of corresponding objects in
source server
• Maintains the relationship even after move operations
Stub info.
Stub information
root
X
move(“/B”,”/A/C”)
Source Target
A B
Y
root
X
A B
/B
root
X
A
C
Y
/B/Y
/B
read(“/A/C/Y”)
status: dummy
source path: /A/X
Generate path name
from parent stub info.
14 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Software stack of target server
Implementation
• Implemented on Linux-based file server.
– Stub management is implemented in kernel VFS layer to
intercept file accesses effectively
– Retriever and crawling programs run in user space
– Used in our production server
Linux Kernel File System
Stub management
Retriever program
libsmbclient Crawling
program
Retrieve data/metadata
kernel
nfsd
Samba
kernel
NFS client
Command queue on
unix domain socket
File accesses
File accesses
Clients
Source
15 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Protocol dependent issues
• NFS hardlinks
– Reproduce hardlink
using representative
file whose name is
based on inode number
• NFS JUKEBOX return code
– Return JUKEBOX if NFS server needs long time to respond to
suppress excessive client retry
– NFS client waits a while and retries later
– Copy process keeps going in background
Our implementation
X
hidden
10Y
Representative
file
NG
X Y X Y
Copied
separately
Hard-linked
(inode# = 10)
16 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Performance evaluation
• Access files under two situations:
A) Pre-copied files
B) Un-copied files that induces on-demand migration
• Evaluation items:
1. Access downtime for client changing connection
2. File access performance overhead after migration
B) on-demand migrationA) pre-copied file
ReadRead
Copy
17 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Test Environment
• Same machine for file servers and client:
– CPU: Xeon E5-2603×2
– DRAM: 64GB
– NIC: 10GbE Copper × 1
– Disk: NVMe SSD
• OS for client and source server:
– NFSv3: Ubuntu Server 16.04LTS
– SMB3: Windows Server 2016 10GbE
SW
Source Target
Client
18 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Downtime (1)
• Read 1GiB file halfway, change connection, and
resume reading
– Total amount of time of remount, open, and reading first byte
is downtime for client
• Change directory depth of test file and width of each
directory
• Remount does not depend on situations
– Less than 0.3 seconds
10MB Read Close Unmount Mount Open 10MB Read 10MB Read
Connect to source Connect to target
10MB Read
remount open read
19 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Downtime (2)
• Open depends on directory structure due to migration
of directories on the pathname
– NFS results shorter for small directory due to less metadata
than SMB (ex. ACL)
– NFS takes much longer for large directory because Linux NFS
client insert 5 seconds wait for NFS JUKEBOX return code
• Client does not cause timeout error due to the code
0.001
0.005
0.001
0.005
0.023
0.023
0.023
0.023
0.052
0.277
5.684
90.075
0.11
0.623
1.62
26.217
0.001
0.01
0.1
1
10
100
1 2 3 4
Timetoopen(seconds)
系列1 系列2 系列3 系列4
Directory parameters (depth, width)
20 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Downtime (3)
• Reading first byte increase response time a little due
to on-demand data migration
– Short time even for 1GB file because target server copies
only accesses blocks before response
– SMB results longer due to more packet exchanges to open
• In total, the downtime is less than 30s in most cases.
– Other cases are supported by NFS JUKEBOX return code.
0.017 0.022
0.058
0.217
1 2
Time to read first byte (seconds)
系列1 系列2
21 / 23
© Hitachi, Ltd. 2018. All rights reserved.
Overhead after migration
• Using filebench for practical case
– Client reads 10,000 files randomly so files are copied little
by little
– As migration progresses, the overhead decreases
0
500
1000
1500
2000
2500
1 21 41 61 81 101
系列1 系列2
(Readperseconds)
(sec)
0
500
1000
1500
2000
1 21 41 61 81 101
系列1 系列2
(Readperseconds)
(sec)
22 / 23
© Hitachi, Ltd. 2018. All rights reserved.
6. Conclusion
• Quick migration scheme of heterogeneous file servers
– Post-copy of file data and metadata
– Combine on-demand and background migration built on stub
management
• Achieve requirements
– Independent and no modification of source server
– Short downtime (<30 seconds in usual case)
– No performance overhead remained after migration
23 / 23

More Related Content

What's hot

Ibm spectrum scale fundamentals workshop for americas part 5 ess gnr-usecases...
Ibm spectrum scale fundamentals workshop for americas part 5 ess gnr-usecases...Ibm spectrum scale fundamentals workshop for americas part 5 ess gnr-usecases...
Ibm spectrum scale fundamentals workshop for americas part 5 ess gnr-usecases...xKinAnx
 
IBM Spectrum Scale Overview november 2015
IBM Spectrum Scale Overview november 2015IBM Spectrum Scale Overview november 2015
IBM Spectrum Scale Overview november 2015Doug O'Flaherty
 
Ibm spectrum scale_backup_n_archive_v03_ash
Ibm spectrum scale_backup_n_archive_v03_ashIbm spectrum scale_backup_n_archive_v03_ash
Ibm spectrum scale_backup_n_archive_v03_ashAshutosh Mate
 
IBM #Softlayer infographic 2016
IBM #Softlayer infographic 2016IBM #Softlayer infographic 2016
IBM #Softlayer infographic 2016Patrick Bouillaud
 
IBM Spectrum Scale for File and Object Storage
IBM Spectrum Scale for File and Object StorageIBM Spectrum Scale for File and Object Storage
IBM Spectrum Scale for File and Object StorageTony Pearson
 
Accelerate with ibm storage ibm spectrum virtualize hyper swap deep dive
Accelerate with ibm storage  ibm spectrum virtualize hyper swap deep diveAccelerate with ibm storage  ibm spectrum virtualize hyper swap deep dive
Accelerate with ibm storage ibm spectrum virtualize hyper swap deep divexKinAnx
 
Components of System Unit
Components of System UnitComponents of System Unit
Components of System UnitAfaq Siddiqui
 
Ibm spectrum scale fundamentals workshop for americas part 1 components archi...
Ibm spectrum scale fundamentals workshop for americas part 1 components archi...Ibm spectrum scale fundamentals workshop for americas part 1 components archi...
Ibm spectrum scale fundamentals workshop for americas part 1 components archi...xKinAnx
 
Ibm spectrum scale fundamentals workshop for americas part 4 Replication, Str...
Ibm spectrum scale fundamentals workshop for americas part 4 Replication, Str...Ibm spectrum scale fundamentals workshop for americas part 4 Replication, Str...
Ibm spectrum scale fundamentals workshop for americas part 4 Replication, Str...xKinAnx
 
NGENSTOR_ODA_P2V_V5
NGENSTOR_ODA_P2V_V5NGENSTOR_ODA_P2V_V5
NGENSTOR_ODA_P2V_V5UniFabric
 
Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...
Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...
Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...Patrick McGarry
 
Spectrum Scale Best Practices by Olaf Weiser
Spectrum Scale Best Practices by Olaf WeiserSpectrum Scale Best Practices by Olaf Weiser
Spectrum Scale Best Practices by Olaf WeiserSandeep Patil
 
Windows Server 2012 R2 Software-Defined Storage
Windows Server 2012 R2 Software-Defined StorageWindows Server 2012 R2 Software-Defined Storage
Windows Server 2012 R2 Software-Defined StorageAidan Finn
 
IMCSummit 2015 - Day 2 Developer Track - The NVM Revolution
IMCSummit 2015 - Day 2 Developer Track - The NVM RevolutionIMCSummit 2015 - Day 2 Developer Track - The NVM Revolution
IMCSummit 2015 - Day 2 Developer Track - The NVM RevolutionIn-Memory Computing Summit
 
Ibm spectrum scale fundamentals workshop for americas part 2 IBM Spectrum Sca...
Ibm spectrum scale fundamentals workshop for americas part 2 IBM Spectrum Sca...Ibm spectrum scale fundamentals workshop for americas part 2 IBM Spectrum Sca...
Ibm spectrum scale fundamentals workshop for americas part 2 IBM Spectrum Sca...xKinAnx
 
Engage for success ibm spectrum accelerate 2
Engage for success   ibm spectrum accelerate 2Engage for success   ibm spectrum accelerate 2
Engage for success ibm spectrum accelerate 2xKinAnx
 
Ibm spectrum scale fundamentals workshop for americas part 5 spectrum scale_c...
Ibm spectrum scale fundamentals workshop for americas part 5 spectrum scale_c...Ibm spectrum scale fundamentals workshop for americas part 5 spectrum scale_c...
Ibm spectrum scale fundamentals workshop for americas part 5 spectrum scale_c...xKinAnx
 
Xiv svc best practices - march 2013
Xiv   svc best practices - march 2013Xiv   svc best practices - march 2013
Xiv svc best practices - march 2013Jinesh Shah
 
IBM SAN Volume Controller Performance Analysis
IBM SAN Volume Controller Performance AnalysisIBM SAN Volume Controller Performance Analysis
IBM SAN Volume Controller Performance Analysisbrettallison
 
Red Hat Storage Day Atlanta - Designing Ceph Clusters Using Intel-Based Hardw...
Red Hat Storage Day Atlanta - Designing Ceph Clusters Using Intel-Based Hardw...Red Hat Storage Day Atlanta - Designing Ceph Clusters Using Intel-Based Hardw...
Red Hat Storage Day Atlanta - Designing Ceph Clusters Using Intel-Based Hardw...Red_Hat_Storage
 

What's hot (20)

Ibm spectrum scale fundamentals workshop for americas part 5 ess gnr-usecases...
Ibm spectrum scale fundamentals workshop for americas part 5 ess gnr-usecases...Ibm spectrum scale fundamentals workshop for americas part 5 ess gnr-usecases...
Ibm spectrum scale fundamentals workshop for americas part 5 ess gnr-usecases...
 
IBM Spectrum Scale Overview november 2015
IBM Spectrum Scale Overview november 2015IBM Spectrum Scale Overview november 2015
IBM Spectrum Scale Overview november 2015
 
Ibm spectrum scale_backup_n_archive_v03_ash
Ibm spectrum scale_backup_n_archive_v03_ashIbm spectrum scale_backup_n_archive_v03_ash
Ibm spectrum scale_backup_n_archive_v03_ash
 
IBM #Softlayer infographic 2016
IBM #Softlayer infographic 2016IBM #Softlayer infographic 2016
IBM #Softlayer infographic 2016
 
IBM Spectrum Scale for File and Object Storage
IBM Spectrum Scale for File and Object StorageIBM Spectrum Scale for File and Object Storage
IBM Spectrum Scale for File and Object Storage
 
Accelerate with ibm storage ibm spectrum virtualize hyper swap deep dive
Accelerate with ibm storage  ibm spectrum virtualize hyper swap deep diveAccelerate with ibm storage  ibm spectrum virtualize hyper swap deep dive
Accelerate with ibm storage ibm spectrum virtualize hyper swap deep dive
 
Components of System Unit
Components of System UnitComponents of System Unit
Components of System Unit
 
Ibm spectrum scale fundamentals workshop for americas part 1 components archi...
Ibm spectrum scale fundamentals workshop for americas part 1 components archi...Ibm spectrum scale fundamentals workshop for americas part 1 components archi...
Ibm spectrum scale fundamentals workshop for americas part 1 components archi...
 
Ibm spectrum scale fundamentals workshop for americas part 4 Replication, Str...
Ibm spectrum scale fundamentals workshop for americas part 4 Replication, Str...Ibm spectrum scale fundamentals workshop for americas part 4 Replication, Str...
Ibm spectrum scale fundamentals workshop for americas part 4 Replication, Str...
 
NGENSTOR_ODA_P2V_V5
NGENSTOR_ODA_P2V_V5NGENSTOR_ODA_P2V_V5
NGENSTOR_ODA_P2V_V5
 
Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...
Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...
Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...
 
Spectrum Scale Best Practices by Olaf Weiser
Spectrum Scale Best Practices by Olaf WeiserSpectrum Scale Best Practices by Olaf Weiser
Spectrum Scale Best Practices by Olaf Weiser
 
Windows Server 2012 R2 Software-Defined Storage
Windows Server 2012 R2 Software-Defined StorageWindows Server 2012 R2 Software-Defined Storage
Windows Server 2012 R2 Software-Defined Storage
 
IMCSummit 2015 - Day 2 Developer Track - The NVM Revolution
IMCSummit 2015 - Day 2 Developer Track - The NVM RevolutionIMCSummit 2015 - Day 2 Developer Track - The NVM Revolution
IMCSummit 2015 - Day 2 Developer Track - The NVM Revolution
 
Ibm spectrum scale fundamentals workshop for americas part 2 IBM Spectrum Sca...
Ibm spectrum scale fundamentals workshop for americas part 2 IBM Spectrum Sca...Ibm spectrum scale fundamentals workshop for americas part 2 IBM Spectrum Sca...
Ibm spectrum scale fundamentals workshop for americas part 2 IBM Spectrum Sca...
 
Engage for success ibm spectrum accelerate 2
Engage for success   ibm spectrum accelerate 2Engage for success   ibm spectrum accelerate 2
Engage for success ibm spectrum accelerate 2
 
Ibm spectrum scale fundamentals workshop for americas part 5 spectrum scale_c...
Ibm spectrum scale fundamentals workshop for americas part 5 spectrum scale_c...Ibm spectrum scale fundamentals workshop for americas part 5 spectrum scale_c...
Ibm spectrum scale fundamentals workshop for americas part 5 spectrum scale_c...
 
Xiv svc best practices - march 2013
Xiv   svc best practices - march 2013Xiv   svc best practices - march 2013
Xiv svc best practices - march 2013
 
IBM SAN Volume Controller Performance Analysis
IBM SAN Volume Controller Performance AnalysisIBM SAN Volume Controller Performance Analysis
IBM SAN Volume Controller Performance Analysis
 
Red Hat Storage Day Atlanta - Designing Ceph Clusters Using Intel-Based Hardw...
Red Hat Storage Day Atlanta - Designing Ceph Clusters Using Intel-Based Hardw...Red Hat Storage Day Atlanta - Designing Ceph Clusters Using Intel-Based Hardw...
Red Hat Storage Day Atlanta - Designing Ceph Clusters Using Intel-Based Hardw...
 

Similar to The Quick Migration of File Servers

Software Defined Analytics with File and Object Access Plus Geographically Di...
Software Defined Analytics with File and Object Access Plus Geographically Di...Software Defined Analytics with File and Object Access Plus Geographically Di...
Software Defined Analytics with File and Object Access Plus Geographically Di...Trishali Nayar
 
FileCatalyst v3.3 preview - multi-file transfers and auto-zip
FileCatalyst v3.3 preview - multi-file transfers and auto-zipFileCatalyst v3.3 preview - multi-file transfers and auto-zip
FileCatalyst v3.3 preview - multi-file transfers and auto-zipFileCatalyst
 
Cloud computing UNIT 2.1 presentation in
Cloud computing UNIT 2.1 presentation inCloud computing UNIT 2.1 presentation in
Cloud computing UNIT 2.1 presentation inRahulBhole12
 
File service architecture and network file system
File service architecture and network file systemFile service architecture and network file system
File service architecture and network file systemSukhman Kaur
 
Directory Write Leases in MagFS
Directory Write Leases in MagFSDirectory Write Leases in MagFS
Directory Write Leases in MagFSMaginatics
 
HDFS- What is New and Future
HDFS- What is New and FutureHDFS- What is New and Future
HDFS- What is New and FutureDataWorks Summit
 
The basics of remote data replication
The basics of remote data replicationThe basics of remote data replication
The basics of remote data replicationFileCatalyst
 
Backups-khtn document 2023 tai lieu hay.pdf
Backups-khtn document 2023 tai lieu hay.pdfBackups-khtn document 2023 tai lieu hay.pdf
Backups-khtn document 2023 tai lieu hay.pdftrihang02122018
 
Scalable and High available Distributed File System Metadata Service Using gR...
Scalable and High available Distributed File System Metadata Service Using gR...Scalable and High available Distributed File System Metadata Service Using gR...
Scalable and High available Distributed File System Metadata Service Using gR...Alluxio, Inc.
 
Building a Distributed File System for the Cloud-Native Era
Building a Distributed File System for the Cloud-Native EraBuilding a Distributed File System for the Cloud-Native Era
Building a Distributed File System for the Cloud-Native EraAlluxio, Inc.
 
[India Merge World Tour] IC Manage
[India Merge World Tour] IC Manage[India Merge World Tour] IC Manage
[India Merge World Tour] IC ManagePerforce
 
exploring-the-oracle-database-architecture.ppt
exploring-the-oracle-database-architecture.pptexploring-the-oracle-database-architecture.ppt
exploring-the-oracle-database-architecture.pptAmitavaRoy49
 
Exploring the Oracle Database Architecture.ppt
Exploring the Oracle Database Architecture.pptExploring the Oracle Database Architecture.ppt
Exploring the Oracle Database Architecture.pptMohammedHdi1
 
FILEgrain: Transport-Agnostic, Fine-Grained Content-Addressable Container Ima...
FILEgrain: Transport-Agnostic, Fine-Grained Content-Addressable Container Ima...FILEgrain: Transport-Agnostic, Fine-Grained Content-Addressable Container Ima...
FILEgrain: Transport-Agnostic, Fine-Grained Content-Addressable Container Ima...Akihiro Suda
 
S014066 scale-ess-orlando-v1705a
S014066 scale-ess-orlando-v1705aS014066 scale-ess-orlando-v1705a
S014066 scale-ess-orlando-v1705aTony Pearson
 
BIOIT14: Deploying very low cost cloud storage technology in a traditional re...
BIOIT14: Deploying very low cost cloud storage technology in a traditional re...BIOIT14: Deploying very low cost cloud storage technology in a traditional re...
BIOIT14: Deploying very low cost cloud storage technology in a traditional re...Dirk Petersen
 
Apache Hadoop 3.0 Community Update
Apache Hadoop 3.0 Community UpdateApache Hadoop 3.0 Community Update
Apache Hadoop 3.0 Community UpdateDataWorks Summit
 

Similar to The Quick Migration of File Servers (20)

Software Defined Analytics with File and Object Access Plus Geographically Di...
Software Defined Analytics with File and Object Access Plus Geographically Di...Software Defined Analytics with File and Object Access Plus Geographically Di...
Software Defined Analytics with File and Object Access Plus Geographically Di...
 
FileCatalyst v3.3 preview - multi-file transfers and auto-zip
FileCatalyst v3.3 preview - multi-file transfers and auto-zipFileCatalyst v3.3 preview - multi-file transfers and auto-zip
FileCatalyst v3.3 preview - multi-file transfers and auto-zip
 
Cloud computing UNIT 2.1 presentation in
Cloud computing UNIT 2.1 presentation inCloud computing UNIT 2.1 presentation in
Cloud computing UNIT 2.1 presentation in
 
File service architecture and network file system
File service architecture and network file systemFile service architecture and network file system
File service architecture and network file system
 
Directory Write Leases in MagFS
Directory Write Leases in MagFSDirectory Write Leases in MagFS
Directory Write Leases in MagFS
 
HDFS- What is New and Future
HDFS- What is New and FutureHDFS- What is New and Future
HDFS- What is New and Future
 
The basics of remote data replication
The basics of remote data replicationThe basics of remote data replication
The basics of remote data replication
 
Backups-khtn document 2023 tai lieu hay.pdf
Backups-khtn document 2023 tai lieu hay.pdfBackups-khtn document 2023 tai lieu hay.pdf
Backups-khtn document 2023 tai lieu hay.pdf
 
Big data- HDFS(2nd presentation)
Big data- HDFS(2nd presentation)Big data- HDFS(2nd presentation)
Big data- HDFS(2nd presentation)
 
Scalable and High available Distributed File System Metadata Service Using gR...
Scalable and High available Distributed File System Metadata Service Using gR...Scalable and High available Distributed File System Metadata Service Using gR...
Scalable and High available Distributed File System Metadata Service Using gR...
 
Building a Distributed File System for the Cloud-Native Era
Building a Distributed File System for the Cloud-Native EraBuilding a Distributed File System for the Cloud-Native Era
Building a Distributed File System for the Cloud-Native Era
 
[India Merge World Tour] IC Manage
[India Merge World Tour] IC Manage[India Merge World Tour] IC Manage
[India Merge World Tour] IC Manage
 
HDFCloud Workshop: HDF5 in the Cloud
HDFCloud Workshop: HDF5 in the CloudHDFCloud Workshop: HDF5 in the Cloud
HDFCloud Workshop: HDF5 in the Cloud
 
exploring-the-oracle-database-architecture.ppt
exploring-the-oracle-database-architecture.pptexploring-the-oracle-database-architecture.ppt
exploring-the-oracle-database-architecture.ppt
 
Exploring the Oracle Database Architecture.ppt
Exploring the Oracle Database Architecture.pptExploring the Oracle Database Architecture.ppt
Exploring the Oracle Database Architecture.ppt
 
Hadoop introduction
Hadoop introductionHadoop introduction
Hadoop introduction
 
FILEgrain: Transport-Agnostic, Fine-Grained Content-Addressable Container Ima...
FILEgrain: Transport-Agnostic, Fine-Grained Content-Addressable Container Ima...FILEgrain: Transport-Agnostic, Fine-Grained Content-Addressable Container Ima...
FILEgrain: Transport-Agnostic, Fine-Grained Content-Addressable Container Ima...
 
S014066 scale-ess-orlando-v1705a
S014066 scale-ess-orlando-v1705aS014066 scale-ess-orlando-v1705a
S014066 scale-ess-orlando-v1705a
 
BIOIT14: Deploying very low cost cloud storage technology in a traditional re...
BIOIT14: Deploying very low cost cloud storage technology in a traditional re...BIOIT14: Deploying very low cost cloud storage technology in a traditional re...
BIOIT14: Deploying very low cost cloud storage technology in a traditional re...
 
Apache Hadoop 3.0 Community Update
Apache Hadoop 3.0 Community UpdateApache Hadoop 3.0 Community Update
Apache Hadoop 3.0 Community Update
 

More from Shinagawa Laboratory, The University of Tokyo

More from Shinagawa Laboratory, The University of Tokyo (8)

Towards Isolated Execution at the Machine Level
Towards Isolated Execution at the Machine LevelTowards Isolated Execution at the Machine Level
Towards Isolated Execution at the Machine Level
 
DMAFV: Testing Device Drivers against DMA Faults
DMAFV: Testing Device Drivers against DMA FaultsDMAFV: Testing Device Drivers against DMA Faults
DMAFV: Testing Device Drivers against DMA Faults
 
Deriving Optimal Deep Learning Models for Image-based Malware Classification
Deriving Optimal Deep Learning Models for Image-based Malware ClassificationDeriving Optimal Deep Learning Models for Image-based Malware Classification
Deriving Optimal Deep Learning Models for Image-based Malware Classification
 
遅延レイヤ取得による高互換コンテナ起動高速化手法
遅延レイヤ取得による高互換コンテナ起動高速化手法遅延レイヤ取得による高互換コンテナ起動高速化手法
遅延レイヤ取得による高互換コンテナ起動高速化手法
 
ライブマイグレーションにおけるサブページ書き込み保護の評価
ライブマイグレーションにおけるサブページ書き込み保護の評価ライブマイグレーションにおけるサブページ書き込み保護の評価
ライブマイグレーションにおけるサブページ書き込み保護の評価
 
A Robust and Flexible Operating System Compatibility Architecture
A Robust and Flexible Operating System Compatibility ArchitectureA Robust and Flexible Operating System Compatibility Architecture
A Robust and Flexible Operating System Compatibility Architecture
 
FaultVisor2: Testing Hypervisor Device Drivers against Real Hardware Failures
FaultVisor2: Testing Hypervisor Device Drivers against Real Hardware FailuresFaultVisor2: Testing Hypervisor Device Drivers against Real Hardware Failures
FaultVisor2: Testing Hypervisor Device Drivers against Real Hardware Failures
 
Distributed Denial of Service Attack Prevention at Source Machines
Distributed Denial of Service Attack Prevention at Source MachinesDistributed Denial of Service Attack Prevention at Source Machines
Distributed Denial of Service Attack Prevention at Source Machines
 

Recently uploaded

WSO2CON2024 - Why Should You Consider Ballerina for Your Next Integration
WSO2CON2024 - Why Should You Consider Ballerina for Your Next IntegrationWSO2CON2024 - Why Should You Consider Ballerina for Your Next Integration
WSO2CON2024 - Why Should You Consider Ballerina for Your Next IntegrationWSO2
 
WSO2Con2024 - WSO2's IAM Vision: Identity-Led Digital Transformation
WSO2Con2024 - WSO2's IAM Vision: Identity-Led Digital TransformationWSO2Con2024 - WSO2's IAM Vision: Identity-Led Digital Transformation
WSO2Con2024 - WSO2's IAM Vision: Identity-Led Digital TransformationWSO2
 
AzureNativeQumulo_HPC_Cloud_Native_Benchmarks.pdf
AzureNativeQumulo_HPC_Cloud_Native_Benchmarks.pdfAzureNativeQumulo_HPC_Cloud_Native_Benchmarks.pdf
AzureNativeQumulo_HPC_Cloud_Native_Benchmarks.pdfryanfarris8
 
Announcing Codolex 2.0 from GDK Software
Announcing Codolex 2.0 from GDK SoftwareAnnouncing Codolex 2.0 from GDK Software
Announcing Codolex 2.0 from GDK SoftwareJim McKeeth
 
WSO2CON 2024 - How CSI Piemonte Is Apifying the Public Administration
WSO2CON 2024 - How CSI Piemonte Is Apifying the Public AdministrationWSO2CON 2024 - How CSI Piemonte Is Apifying the Public Administration
WSO2CON 2024 - How CSI Piemonte Is Apifying the Public AdministrationWSO2
 
WSO2CON 2024 - Architecting AI in the Enterprise: APIs and Applications
WSO2CON 2024 - Architecting AI in the Enterprise: APIs and ApplicationsWSO2CON 2024 - Architecting AI in the Enterprise: APIs and Applications
WSO2CON 2024 - Architecting AI in the Enterprise: APIs and ApplicationsWSO2
 
WSO2CON 2024 Slides - Open Source to SaaS
WSO2CON 2024 Slides - Open Source to SaaSWSO2CON 2024 Slides - Open Source to SaaS
WSO2CON 2024 Slides - Open Source to SaaSWSO2
 
WSO2Con2024 - Facilitating Broadband Switching Services for UK Telecoms Provi...
WSO2Con2024 - Facilitating Broadband Switching Services for UK Telecoms Provi...WSO2Con2024 - Facilitating Broadband Switching Services for UK Telecoms Provi...
WSO2Con2024 - Facilitating Broadband Switching Services for UK Telecoms Provi...WSO2
 
WSO2CON 2024 - Does Open Source Still Matter?
WSO2CON 2024 - Does Open Source Still Matter?WSO2CON 2024 - Does Open Source Still Matter?
WSO2CON 2024 - Does Open Source Still Matter?WSO2
 
WSO2Con2024 - GitOps in Action: Navigating Application Deployment in the Plat...
WSO2Con2024 - GitOps in Action: Navigating Application Deployment in the Plat...WSO2Con2024 - GitOps in Action: Navigating Application Deployment in the Plat...
WSO2Con2024 - GitOps in Action: Navigating Application Deployment in the Plat...WSO2
 
Driving Innovation: Scania's API Revolution with WSO2
Driving Innovation: Scania's API Revolution with WSO2Driving Innovation: Scania's API Revolution with WSO2
Driving Innovation: Scania's API Revolution with WSO2WSO2
 
WSO2Con2024 - Navigating the Digital Landscape: Transforming Healthcare with ...
WSO2Con2024 - Navigating the Digital Landscape: Transforming Healthcare with ...WSO2Con2024 - Navigating the Digital Landscape: Transforming Healthcare with ...
WSO2Con2024 - Navigating the Digital Landscape: Transforming Healthcare with ...WSO2
 
WSO2CON 2024 - Building a Digital Government in Uganda
WSO2CON 2024 - Building a Digital Government in UgandaWSO2CON 2024 - Building a Digital Government in Uganda
WSO2CON 2024 - Building a Digital Government in UgandaWSO2
 
WSO2CON 2024 Slides - Unlocking Value with AI
WSO2CON 2024 Slides - Unlocking Value with AIWSO2CON 2024 Slides - Unlocking Value with AI
WSO2CON 2024 Slides - Unlocking Value with AIWSO2
 
WSO2CON 2024 - Lessons from the Field: Legacy Platforms – It's Time to Let Go...
WSO2CON 2024 - Lessons from the Field: Legacy Platforms – It's Time to Let Go...WSO2CON 2024 - Lessons from the Field: Legacy Platforms – It's Time to Let Go...
WSO2CON 2024 - Lessons from the Field: Legacy Platforms – It's Time to Let Go...WSO2
 
WSO2Con2024 - Simplified Integration: Unveiling the Latest Features in WSO2 L...
WSO2Con2024 - Simplified Integration: Unveiling the Latest Features in WSO2 L...WSO2Con2024 - Simplified Integration: Unveiling the Latest Features in WSO2 L...
WSO2Con2024 - Simplified Integration: Unveiling the Latest Features in WSO2 L...WSO2
 
WSO2CON 2024 - Navigating API Complexity: REST, GraphQL, gRPC, Websocket, Web...
WSO2CON 2024 - Navigating API Complexity: REST, GraphQL, gRPC, Websocket, Web...WSO2CON 2024 - Navigating API Complexity: REST, GraphQL, gRPC, Websocket, Web...
WSO2CON 2024 - Navigating API Complexity: REST, GraphQL, gRPC, Websocket, Web...WSO2
 
WSO2CON 2024 - Building the API First Enterprise – Running an API Program, fr...
WSO2CON 2024 - Building the API First Enterprise – Running an API Program, fr...WSO2CON 2024 - Building the API First Enterprise – Running an API Program, fr...
WSO2CON 2024 - Building the API First Enterprise – Running an API Program, fr...WSO2
 
%in Soweto+277-882-255-28 abortion pills for sale in soweto
%in Soweto+277-882-255-28 abortion pills for sale in soweto%in Soweto+277-882-255-28 abortion pills for sale in soweto
%in Soweto+277-882-255-28 abortion pills for sale in sowetomasabamasaba
 

Recently uploaded (20)

WSO2CON2024 - Why Should You Consider Ballerina for Your Next Integration
WSO2CON2024 - Why Should You Consider Ballerina for Your Next IntegrationWSO2CON2024 - Why Should You Consider Ballerina for Your Next Integration
WSO2CON2024 - Why Should You Consider Ballerina for Your Next Integration
 
WSO2Con2024 - WSO2's IAM Vision: Identity-Led Digital Transformation
WSO2Con2024 - WSO2's IAM Vision: Identity-Led Digital TransformationWSO2Con2024 - WSO2's IAM Vision: Identity-Led Digital Transformation
WSO2Con2024 - WSO2's IAM Vision: Identity-Led Digital Transformation
 
AzureNativeQumulo_HPC_Cloud_Native_Benchmarks.pdf
AzureNativeQumulo_HPC_Cloud_Native_Benchmarks.pdfAzureNativeQumulo_HPC_Cloud_Native_Benchmarks.pdf
AzureNativeQumulo_HPC_Cloud_Native_Benchmarks.pdf
 
Announcing Codolex 2.0 from GDK Software
Announcing Codolex 2.0 from GDK SoftwareAnnouncing Codolex 2.0 from GDK Software
Announcing Codolex 2.0 from GDK Software
 
WSO2CON 2024 - How CSI Piemonte Is Apifying the Public Administration
WSO2CON 2024 - How CSI Piemonte Is Apifying the Public AdministrationWSO2CON 2024 - How CSI Piemonte Is Apifying the Public Administration
WSO2CON 2024 - How CSI Piemonte Is Apifying the Public Administration
 
WSO2CON 2024 - Architecting AI in the Enterprise: APIs and Applications
WSO2CON 2024 - Architecting AI in the Enterprise: APIs and ApplicationsWSO2CON 2024 - Architecting AI in the Enterprise: APIs and Applications
WSO2CON 2024 - Architecting AI in the Enterprise: APIs and Applications
 
WSO2CON 2024 Slides - Open Source to SaaS
WSO2CON 2024 Slides - Open Source to SaaSWSO2CON 2024 Slides - Open Source to SaaS
WSO2CON 2024 Slides - Open Source to SaaS
 
WSO2Con2024 - Facilitating Broadband Switching Services for UK Telecoms Provi...
WSO2Con2024 - Facilitating Broadband Switching Services for UK Telecoms Provi...WSO2Con2024 - Facilitating Broadband Switching Services for UK Telecoms Provi...
WSO2Con2024 - Facilitating Broadband Switching Services for UK Telecoms Provi...
 
WSO2CON 2024 - Does Open Source Still Matter?
WSO2CON 2024 - Does Open Source Still Matter?WSO2CON 2024 - Does Open Source Still Matter?
WSO2CON 2024 - Does Open Source Still Matter?
 
WSO2Con2024 - GitOps in Action: Navigating Application Deployment in the Plat...
WSO2Con2024 - GitOps in Action: Navigating Application Deployment in the Plat...WSO2Con2024 - GitOps in Action: Navigating Application Deployment in the Plat...
WSO2Con2024 - GitOps in Action: Navigating Application Deployment in the Plat...
 
Driving Innovation: Scania's API Revolution with WSO2
Driving Innovation: Scania's API Revolution with WSO2Driving Innovation: Scania's API Revolution with WSO2
Driving Innovation: Scania's API Revolution with WSO2
 
WSO2Con2024 - Navigating the Digital Landscape: Transforming Healthcare with ...
WSO2Con2024 - Navigating the Digital Landscape: Transforming Healthcare with ...WSO2Con2024 - Navigating the Digital Landscape: Transforming Healthcare with ...
WSO2Con2024 - Navigating the Digital Landscape: Transforming Healthcare with ...
 
Abortion Pill Prices Tembisa [(+27832195400*)] 🏥 Women's Abortion Clinic in T...
Abortion Pill Prices Tembisa [(+27832195400*)] 🏥 Women's Abortion Clinic in T...Abortion Pill Prices Tembisa [(+27832195400*)] 🏥 Women's Abortion Clinic in T...
Abortion Pill Prices Tembisa [(+27832195400*)] 🏥 Women's Abortion Clinic in T...
 
WSO2CON 2024 - Building a Digital Government in Uganda
WSO2CON 2024 - Building a Digital Government in UgandaWSO2CON 2024 - Building a Digital Government in Uganda
WSO2CON 2024 - Building a Digital Government in Uganda
 
WSO2CON 2024 Slides - Unlocking Value with AI
WSO2CON 2024 Slides - Unlocking Value with AIWSO2CON 2024 Slides - Unlocking Value with AI
WSO2CON 2024 Slides - Unlocking Value with AI
 
WSO2CON 2024 - Lessons from the Field: Legacy Platforms – It's Time to Let Go...
WSO2CON 2024 - Lessons from the Field: Legacy Platforms – It's Time to Let Go...WSO2CON 2024 - Lessons from the Field: Legacy Platforms – It's Time to Let Go...
WSO2CON 2024 - Lessons from the Field: Legacy Platforms – It's Time to Let Go...
 
WSO2Con2024 - Simplified Integration: Unveiling the Latest Features in WSO2 L...
WSO2Con2024 - Simplified Integration: Unveiling the Latest Features in WSO2 L...WSO2Con2024 - Simplified Integration: Unveiling the Latest Features in WSO2 L...
WSO2Con2024 - Simplified Integration: Unveiling the Latest Features in WSO2 L...
 
WSO2CON 2024 - Navigating API Complexity: REST, GraphQL, gRPC, Websocket, Web...
WSO2CON 2024 - Navigating API Complexity: REST, GraphQL, gRPC, Websocket, Web...WSO2CON 2024 - Navigating API Complexity: REST, GraphQL, gRPC, Websocket, Web...
WSO2CON 2024 - Navigating API Complexity: REST, GraphQL, gRPC, Websocket, Web...
 
WSO2CON 2024 - Building the API First Enterprise – Running an API Program, fr...
WSO2CON 2024 - Building the API First Enterprise – Running an API Program, fr...WSO2CON 2024 - Building the API First Enterprise – Running an API Program, fr...
WSO2CON 2024 - Building the API First Enterprise – Running an API Program, fr...
 
%in Soweto+277-882-255-28 abortion pills for sale in soweto
%in Soweto+277-882-255-28 abortion pills for sale in soweto%in Soweto+277-882-255-28 abortion pills for sale in soweto
%in Soweto+277-882-255-28 abortion pills for sale in soweto
 

The Quick Migration of File Servers

  • 1. © Hitachi, Ltd. 2018. All rights reserved. 06/05/2018 Keiichi Matsuzawa*† Mitsuo Hayasaka* Takahiro Shinagawa† The Quick Migration of File Servers * Hitachi, Ltd. † University of Tokyo
  • 2. © Hitachi, Ltd. 2018. All rights reserved. Background • Most cases of installing new file servers require server and data migration across file servers Source: Storage Magazine, Vol.16, No.2, 2017 20% 24% 32% 39% 56% 1 2 3 4 5 What are the top drivers for your new NAS system? Improve performance for existing apps Replace old hardware Deploy new application Support virtual server better Use as a backup target Migrate from existing old server 1 / 23
  • 3. © Hitachi, Ltd. 2018. All rights reserved. Challenges for migration scheme 1. Independent of the source server – Heterogenous file server migration is needed to: • Change to a more attractive vendor and product • Dispose of End-of-Life products • Aggregate small file servers to a enterprise file server – Should not modify the source server 2. No long downtime for clients (<30 seconds) – Common OSs define network file access timeout about tens of seconds 3. No performance overhead after migration – Improving performance is one of top drivers 2 / 23
  • 4. © Hitachi, Ltd. 2018. All rights reserved. • Pre-copy data per file [rsync][robocopy] – Tracking the last updates takes long time to traverse tree • Copy disk volume [SnapMirror] – Source and target must share the same disk layout Existing solutions Source Target Cannot mount volume >1day downtime to track the last update rsync Source Target 3 / 23
  • 5. © Hitachi, Ltd. 2018. All rights reserved. Our approach • Post-Copy per file – Clients change connection before copying data – Target server copies metadata and data on demand • Transfer data only via common interface (ex. NFS) X 5: return 3: read(“/A/X”) from target Source Target A root A X root B Y 4: copy(“/A/X”) Client 1: read(“/A/X”) from source 2: remount 4 / 23
  • 6. © Hitachi, Ltd. 2018. All rights reserved. Data migration method • Migrate whole file namespace by: 1. On-demand migration – Copies objects accessed by clients on-demand 2. Background migration – Copies un-accessed objects in background 3. Stub-based file management: – Enables intermediate migration state and supports above migrations 5 / 23
  • 7. © Hitachi, Ltd. 2018. All rights reserved. On-demand migration • Copy directories and files recursively when accessed • After preparing root directory, any files are accessible – Accessing directory only generates intermediate state objects to reduce time to copy root A B C root root A B C X Y Z root A B C X Y Z lookup “/” lookup “/A/” lookup “/A/Y” read( “/A/Y” ) Create objects without data 6 / 23
  • 8. © Hitachi, Ltd. 2018. All rights reserved. Background migration (1) • Copy un-accessed files by clients • Crawling program traverses file namespace and induces on-demand migration – Should run when access load is low (ex. midnight or weekend) root A B C X Y Z root A B C X Y Z root A B C X Y Z Run crawling program 7 / 23
  • 9. © Hitachi, Ltd. 2018. All rights reserved. Background migration (2) • Crawling program may run intermittently – Copy huge file share across several nights – Need to handle updated files in the daytime • Restart from the root directory each time – Count intermediate objects until reaching zero root A B C X Y Z root A B C X Y Z root A B C X Y Z count: 4 count: 2 count: 0 8 / 23
  • 10. © Hitachi, Ltd. 2018. All rights reserved. • Should consider characteristics of file system: • Data structure – Hierarchical structure – Protocol dependent metadata (ex. ACL, uid) • Operations – Rename/move objects – Hardlinks – Truncate • Client behaviors (ex. Return code handling, thumbnail preview) • Product dependent features (ex. Compression, quota) Issues of post-copy per file (1) Scope of this presentation 9 / 23
  • 11. © Hitachi, Ltd. 2018. All rights reserved. Issues of post-copy per file (2) • Cause timeout to create inodes under huge directory – Retrieving metadata requires many packet exchanges and takes long time A X read(“/A/Y”) Source Target Y Z … A Target A X Y Z … Create all at once? 10 / 23
  • 12. © Hitachi, Ltd. 2018. All rights reserved. Issues of post-copy per file (3) • Keep consistency of source and target objects even after move operation root X move(“/B”,”/A/C”) Source Target A B Y root X A B Target root X A C Different path for source and target Y C 11 / 23
  • 13. © Hitachi, Ltd. 2018. All rights reserved. Stub management • Concept used in Hierarchical Storage Management – Stub is a virtual object does not have data contents but stores actual data location – File system intercepts stub accesses and inserts additional operation • Apply the concept for file server migration – Allow intermediate states storing partial data and metadata – Retrieve un-copied data when stub is accessed 12 / 23
  • 14. © Hitachi, Ltd. 2018. All rights reserved. Dummy Partial Normal File and directory status • Define intermediate states storing partial (meta)data – Defer (meta)data copy until the file is accessed – Dummy object can be created by only parent directory entry – Partial object has full metadata and parts of data blocks • Regards hole blocks as un-copied blocks – Normal object has full metadata and data • Migration is completed and not produce any source server accesses Stub info. Metadata Metadata Data inodeAccess whole data Access metadata Access parent directory Data Intermediate states inode Stub info. inode Internal data structure Migration completedMost attributes are empty 13 / 23
  • 15. © Hitachi, Ltd. 2018. All rights reserved. • Additional metadata to manage intermediate states • Stores full path name of corresponding objects in source server • Maintains the relationship even after move operations Stub info. Stub information root X move(“/B”,”/A/C”) Source Target A B Y root X A B /B root X A C Y /B/Y /B read(“/A/C/Y”) status: dummy source path: /A/X Generate path name from parent stub info. 14 / 23
  • 16. © Hitachi, Ltd. 2018. All rights reserved. Software stack of target server Implementation • Implemented on Linux-based file server. – Stub management is implemented in kernel VFS layer to intercept file accesses effectively – Retriever and crawling programs run in user space – Used in our production server Linux Kernel File System Stub management Retriever program libsmbclient Crawling program Retrieve data/metadata kernel nfsd Samba kernel NFS client Command queue on unix domain socket File accesses File accesses Clients Source 15 / 23
  • 17. © Hitachi, Ltd. 2018. All rights reserved. Protocol dependent issues • NFS hardlinks – Reproduce hardlink using representative file whose name is based on inode number • NFS JUKEBOX return code – Return JUKEBOX if NFS server needs long time to respond to suppress excessive client retry – NFS client waits a while and retries later – Copy process keeps going in background Our implementation X hidden 10Y Representative file NG X Y X Y Copied separately Hard-linked (inode# = 10) 16 / 23
  • 18. © Hitachi, Ltd. 2018. All rights reserved. Performance evaluation • Access files under two situations: A) Pre-copied files B) Un-copied files that induces on-demand migration • Evaluation items: 1. Access downtime for client changing connection 2. File access performance overhead after migration B) on-demand migrationA) pre-copied file ReadRead Copy 17 / 23
  • 19. © Hitachi, Ltd. 2018. All rights reserved. Test Environment • Same machine for file servers and client: – CPU: Xeon E5-2603×2 – DRAM: 64GB – NIC: 10GbE Copper × 1 – Disk: NVMe SSD • OS for client and source server: – NFSv3: Ubuntu Server 16.04LTS – SMB3: Windows Server 2016 10GbE SW Source Target Client 18 / 23
  • 20. © Hitachi, Ltd. 2018. All rights reserved. Downtime (1) • Read 1GiB file halfway, change connection, and resume reading – Total amount of time of remount, open, and reading first byte is downtime for client • Change directory depth of test file and width of each directory • Remount does not depend on situations – Less than 0.3 seconds 10MB Read Close Unmount Mount Open 10MB Read 10MB Read Connect to source Connect to target 10MB Read remount open read 19 / 23
  • 21. © Hitachi, Ltd. 2018. All rights reserved. Downtime (2) • Open depends on directory structure due to migration of directories on the pathname – NFS results shorter for small directory due to less metadata than SMB (ex. ACL) – NFS takes much longer for large directory because Linux NFS client insert 5 seconds wait for NFS JUKEBOX return code • Client does not cause timeout error due to the code 0.001 0.005 0.001 0.005 0.023 0.023 0.023 0.023 0.052 0.277 5.684 90.075 0.11 0.623 1.62 26.217 0.001 0.01 0.1 1 10 100 1 2 3 4 Timetoopen(seconds) 系列1 系列2 系列3 系列4 Directory parameters (depth, width) 20 / 23
  • 22. © Hitachi, Ltd. 2018. All rights reserved. Downtime (3) • Reading first byte increase response time a little due to on-demand data migration – Short time even for 1GB file because target server copies only accesses blocks before response – SMB results longer due to more packet exchanges to open • In total, the downtime is less than 30s in most cases. – Other cases are supported by NFS JUKEBOX return code. 0.017 0.022 0.058 0.217 1 2 Time to read first byte (seconds) 系列1 系列2 21 / 23
  • 23. © Hitachi, Ltd. 2018. All rights reserved. Overhead after migration • Using filebench for practical case – Client reads 10,000 files randomly so files are copied little by little – As migration progresses, the overhead decreases 0 500 1000 1500 2000 2500 1 21 41 61 81 101 系列1 系列2 (Readperseconds) (sec) 0 500 1000 1500 2000 1 21 41 61 81 101 系列1 系列2 (Readperseconds) (sec) 22 / 23
  • 24. © Hitachi, Ltd. 2018. All rights reserved. 6. Conclusion • Quick migration scheme of heterogeneous file servers – Post-copy of file data and metadata – Combine on-demand and background migration built on stub management • Achieve requirements – Independent and no modification of source server – Short downtime (<30 seconds in usual case) – No performance overhead remained after migration 23 / 23