4. Virtual Disk Pooling
Consolidate like or unlike disk resources
► Split pool into tiers of different
price/performance/capacity
► Create and assign virtual disks
of desired sizes
► Define access rights
► Explicitly assign virtual disks
to hosts or groups of hosts
► Expand capacity without downtime
► Eliminate stranded disk space
4
5. Remove Direct-Attached Drawbacks
DAS
Eliminate stranded
storage
W brand
6TB # P1 Make available capacity
accessible to most
needy applications
Free
W brand
8TB # G2 Choose disk
characteristics based on
program needs
Defer additional purchase
X brand until pool nears depletion
11TB # S3 Used
5
6. Manage Capacity as a Pool of Disk Space
Tier Cost RAS Speed Density Disk Space (TBs)
W brand W brand
P1 model P3 model
1 Hi Hi Hi Lo
6TB
Used
W brand W brand
G2 model G5 model
2 Med Hi Med Med
8TB
Used
SAN
X brand Z brand
S3 model Q7 model
3 Lo Med Lo Hi
11TB
Used
6
7. SAN-wide Centralized Management
Control / monitor all pooled resources from one console
► Intuitive to set up and operate
► Automates repetitive tasks
► Self-guided wizards for
key workflows
► Comprehensive diagnostics
& troubleshooting tips
► Configurable views of system
behavior and performance
► Role-based, administrative
permissions
7
8. Auto-Tiering
Intelligent trade-offs between cost and performance
► No special disk hardware
required
► You select which disks make
up each tier
► Adapts to provide most
demanding workloads with
speediest response
8
9. Traditional: Virtual Disk Assigned to Tier
Different pools satisfy different
performance/capacity needs
Best practice: Don't mix disks of different performance
characteristics and/or sizes within one pool
Tier 1
Fastest Performance
Tier 2
Balanced performance/costs
Tier 3
High capacity/Low Cost
ERP
Transactional workloads
Databases File
Email
User home directory Archive
Backup-to-disk
Scratch files
9
10. Making Good Decisions on Your Behalf
Monitor sustained I/O behavior across groups of disk
blocks (chunks)
Compare access frequency against average for each tier
Promote or demote chunks to most suitable tier
Tier 1 Tier 2 Tier 3
Chunks
Above average Below average
moves up a tier moves down a tier
11
11. Virtual Disk distribution within a pool
Each Virtual Disk has a Storage Profile assigned
Storage Profiles define tier preferences
DISK POOL
VD Tier 1 Tier 3 VD
Storage Profile Storage Profile
Tier 2
"High" "Low"
VD
Storage Profile
"Normal"
12
12. Sensing Hot Spots within a Virtual Disk
Heaviest hit blocks take advantage of
Tier 1 performance
Other blocks moved to Tier 2 & Tier 3
DISK POOL
Frequently
VIRTUAL
DISK
Accessed Tier 1 Tier 3
Moderately
Accessed
Infrequently
Tier 2
Accessed
13
13. High-Speed Caching
Speeds up performance
► Accelerates disk I/O response
from existing storage
► Uses x86-64 CPUs and
memory from DataCore
nodes as powerful,
inexpensive “mega caches”
► Anticipates next blocks to
be read, and groups writes
to avoid waiting on disks
14
14. Load Balancing
Improve response and throughput
Improve response and throughput
► Overcome typical storage-
related bottlenecks
► Spread load on physical
devices using different
channels for different
virtual disks
► Automatically bypasses
failed or offline channels
15
15. Thin Provisioning
Allocate just enough space, just-in-time
► Appears to computers as very
large drives (e.g. 2 TB disks)
► Takes up only space actually
being written to
► Dynamically allocates more
disk space when required
► Reduces need to resize LUNs
► Reclaims zeroed out
disk space
16
16. Guessing on Allocation Leads to Waste
Disk 1 Disk 2 Disk 3
Free
Few Blocks Actually
Free Disk 1 Written To
Free
Disk 3
Free
Free Disk 2
Disk 2
Capacity consumed
whether used or not!
17
17. Waste-Free Thin Provisioning
2TB
Free Virtual + 2.0TB
V Disk 3 Allocation: + 2.0TB
+ 2.0TB
0 2TB
6.0TB
Free
V Disk 2 Actual = 0.4TB
0 2TB
Free
V Disk 1
0
Physical Disk Pool Time to add more
60%
Free Expand as No Device
Needed Dependency
0 0.4TB 1TB
18
19. Thin Provisioning Space Reclamation
Recover space from deleted files in virtual volumes
2TB
Free Space previously
V Disk 1 allocated to deleted file
0 2TB
Free
V Disk 1 Fill with zeroes
0
Release zero-filled
space back to free pool
0 Physical Disk Pool
20
20. RAID Striping
Better protection & performance
► Circumvents drive failures
► Spreads I/O across
multiple spindles
► Offloads RAID 0 & 1
► Supports popular RAID
devices in pool
21
21. Synchronous Mirroring
Real-time I/O replication for High-Availability
► Eliminates storage as
a single point of failure
► Enhances survivability using
physically separate nodes
► Updates two distributed
copies simultaneously
► Mirrored virtual disks behave
as one, multi-ported shared drive
22
22. Synchronously Mirrored Writes
“Virtual Disk” • Separate Chassis
Client
• Different Location
• Independent Power
• Independent Cooling
1 • Separate UPS
4
2
Active Synchronous Partner
Mirroring
3
5 Cache Cache 6
Mirror (P) Mirror (S)
Node A Node B
23
23. Automatic Failover via Multipath I/O
“Virtual Disk”
Client
Preferred
MPIO Alternate
1 1
4 Retry
Offline Partner
3
Log
Updates 2
Cache Cache
Out-of-
Mirror (P)
Sync Mirror (S)
Node A Node B
24
24. Resynchronization of Mirrored Drives
“Virtual Disk”
Client
MPIO Alternate
Transmit changes
since outage
Rejoin Partner
Log
Updates
Cache Cache
Out-of-
Mirror (P)
Sync Mirror (S)
Node A Node B
25
25. Resume Normal Operations
“Virtual Disk”
Client
Preferred
MPIO Alternate
OFF
Synchronous
Active Mirroring Partner
Cache Cache
Mirror (P) Mirror (S)
Node A Node B
26
27. Virtual Disk Migration
Transparently move contents from one disk to another
► Allows non-disruptive
hardware disk upgrades
► Clears & reclaims space
occupied by original
► Provides pass-through
access to drives previously
used on other systems
28
28. Non-Disruptive Disk Upgrades
Virtual disks remain active
while contents are migrated to
MPIO
a different device
Physical Storage Pool
Old New
Tier 1
Tier 2
Retain or discard
Tier 3 older device
29
29. Online Snapshots
Capture point-in-time images without tying up much disk
space or make complete clones
► Recover quickly at disk
speeds to known good state
► Eliminate back-up window
► Provide “live” environment
for analysis, development
& testing
► Save snapshots in lower tier,
thin-provisioned disks
without taking up space on
premium storage devices
30
30. Low-impact, Point-in-time Snapshots
Copy Empty when
original contents
Snapshot st write
here on 1 enabled Active program
If unchanged,
read from Source
1
(Earlier point-in-time)
2
E p-i-t Source
3
Current
image
4
• Back-up
• Testing
• Analysis
• Recovery Only changed blocks Source disk
take up space blocks
31
31. Continuous Data Protection & Recovery
Return to any point-in-time without taking explicit backups
► Dial back to restore arbitrary
point-in-time within a 48 hour
time frame
► Logs and timestamps all I/Os
to the selected virtual disks
► No need to quiesce or
interrupt applications
► No host agents required
► Easy to turn on and
revert from
32
32. Revert to Earlier Image
Active CDP Time-
Stamped
Updates
Cache
Choose time before
problem occurred
11:09am
= 10:22am
9am 10am 11am
33
33. Remote Replication
Update distant copies without impacting local performance
► Perfect for disaster recovery
► Only needs a basic IP
connection to secondary site
► Bi-directional (sites can
replicate to each other)
► Asynchronous
► Compressed, multi-stream
transfers for fast performance
& optimum use of bandwidth
34
34. Keep Remote Disaster Recovery
Site Updated
DR Site
Recovery
Destination Image
Source
Replicated
Updates
Cache
Snapshot
Queued
Updates
Active
Cache
Central IT site Asynchronous
IP
Copy
35
35. Advanced Site Recovery
Expedite central site restoration
► Reverses direction of
replication from the disaster
recovery (DR) site to the
primary datacenter
► Same automated process
for virtual & physical systems
36