Learn about IC Manage Views, a FUSE-based user space file system that allows near Zero Time Sync (ZTS) for workspaces of any size and complexity on both Windows and Linux.
1. 1
Workspace Acceleration & Network
Storage Reduction
Shiv Sikand
VP of Engineering
Roger March
Chief Technology
Officer
2. 2
Driven by Semiconductor Design, Video game software, Enterprise software…
Perforce depots: multi-TB+, 100,000,000s+ files
Single workspaces: multi-GB, 10,000s+ files
Users: 1000’s across globe
Digital Assets Growing at Rapid Rate
3. 3
Source: 2012 LaunchM Survey, 524 respondents
Impact of Increasing Digital Assets:
Survey cites Top File System Problems
44%
36%
36%
33%
25%
18%
5%
Slow
WS
syncs,
apps
Bandwidth
Storage
capacity,
cost
5. 5
Network Storage
Network Storage
Workspace C File 1 File 2 File 3
Workspace B File 1 File 2 File 3
Workspace A File 1 File 2 File 3
User A
User B
User C
• Duplicates
file
storage
for
every
user
workspace
• De-‐dupe
too
slow
for
fast
changing
data
sets
• Saturates
precious
Tier
1
storage
space
• Under-‐uAlizes
local
caching
High
Demand
on
Network
Storage
Perforce
SCM Server
File Data
Metadata
File 1 File 2 File 3
6. 6
File Not Yet
Opened
User A
User B
User C
Local Cache
File 2
0 b
File 1
0 b
File 3
0 b
• Virtual
workspaces
projected
using
Perforce
db.have
table
informaAon
from
fstat
• 100%
dynamic,
user
controlled
workspaces
• Local
storage
agnosAc,
e.g.
disk,
tmpfs,
SSD....
Local Cache
File 2
0 b
File 1
0 b
File 3
0 b
Local Cache
File 2
0 b
File 1
0 b
File 3
0 b
Dynamic
Virtual
Workspaces
Instantly
displays
fully
populated
workspaces
7. 7
Local
Caching
Streams
actual
files
on
applicaAon
/
user
demand
User A
User B
User C
Local Cache
File 2
0 b
File 1
0 b
File 3
0 b
• Files
can
be
streamed
from
main
Perforce
server,
Perforce
proxy
or
Perforce
replica
• TCP
streaming
protocol
avoids
tradiAonal
NFS-‐delivery
latency
• 2x
faster
retrieval
of
cache-‐warm
files
due
to
on-‐board
access
File Not Yet
Opened
File Opened
Read-only
Local Cache
File 1
0 b
File 3
0 bFile 2
Local Cache
File 2
0 b
File 1
0 b
File 3
8. 8
User A
User B
Local Cache
File 2
Local Cache
File 3
0 b
File 2
0 b
User C
Local Cache
File 2
0 b
File 1
0 b
File 3
File 1
0 b
Automa0c
cache
management
stays
within
storage
quotas
using
LRU
algorithm
File 3
File
File 1
File Not Yet
Opened
File Opened
Read-only
Local
caching
-‐
manages
cache
quotas
9. 9
Intelligent
File
RedirecAon:
2x
faster
file
access
Reads
Writes
Intelligent
File
Filtering
approach
• Separates
reads
from
writes,
stores
reads
in
local
cache
• Modified
files
automaAcally
wriWen
to
NAS
(or
local
cache)
Advantages
• Gets
on-‐board
speeds
for
reads
instead
of
network
storage
• Modified
files
safely
stored
on
network
disks
• Automated
-‐
No
manual
management
Intelligent
File
Redirect
On-‐board
speeds
for
reads
Network
Storage
Local Cache
Read/
Write
Read
-only
Read
-only
Read
-only
Read/
Write
Read/
Write
Read
-only
Read
-only
Read/
Write
Read/
Write
10. 10
User A
User B
Local Cache
Local Cache
File 3
0 b
File 2
0 b
User C
Local Cache
File 2
0 b
File 1
0 b
File 3
File 1
0 b
Newly
created
files,
checked
out
files
saved
to
network
storage
volume
or
kept
local
File 3
File
File 2
0 b
Network Storage
TEMP 1
File Opened
Read/Write
File 1
File Not Yet
Opened
File Opened
Read-only
Intelligent
file
redirect
11. 11
User A
User B
Local Cache
Local Cache
File 3
0 b
File 2
0 b
User C
Local Cache
File 2
0 b
File 1
0 b
File 3
File 1
0 b
Checked
in
objects
automa0cally
purged
from
write
storage
Avoids
duplicated
files,
frees
up
space.
File 3
File
File 2
0 b
Network Storage
File 1 TEMP 1
Real-‐Ame
de-‐duplicaAon
-‐
frees
space
aer
Perforce
check-‐ins
File Opened
Read/Write
File Not Yet
Opened
File Opened
Read-only
13. 13
IC
Manage
Views
AddiAonal
Benefits
• NAS
agnosAc
-‐
100%
compaAble
with
exis0ng
storage
technologies
• Scalable
-‐
Savings
increase
with
number
of
users
and
the
size
of
databases.
• Flexibility
to
build
Workspaces
on
Demand.
SoVware
teams
can
build
workspaces
anywhere,
avoiding
problems
and
costs
associated
with
disk
space
alloca0on.
• Reliable.
Handles
cache
recovery
in
the
event
of
failures
or
errors.
• Designers
maintain
workspace
file
control
and
stability.
No
manual
management
of
network
cache
storage
and
different
versions
• Reduces
load
on
filers
and
network.