Larry Chiang

1
THE ART OF STORAGE NETWORKING

• Larry Chiang
• Director - Sales Mkting
• FalconStor Asia Pacific
• larry_at_falconstor.com.tw

2
Company Highlight

• Headquartered in Melville, New York
• Over 100 (US/Asia/Europe) people and counting
• Seasoned Veterans of Cheyenne Software
• Successfully made ARCserve the 1 backup solution
  for the enterprise
• Experienced in Storage, Operating System and
  Networking
• Provide Network Storage Infrastructure Software
  to
• aggregate and virtualize storage resources (SCSI,
  FC, iSCSI, SSA, InfiniBand..)
• provision the storage services via SAN (IP,
  iSCSI, FC) or NAS (NFS, CIFS)
• Target Corporate and Internet Data Centers (xSP)
• Version 2 Released Sept, 2001
• Strong partner endorsement and OEM partnership
• Over 50 signed contracts world wide
• Became the first emerging Virtualization
  Software company to achieve public trading
  status (NASDAQFALC)

3
Our Vision
Leverage Off-the-Shelf, Industry-Standard
Components to Virtualize, Provision and Manage
Storage Resources in a Distributed Environment.
4
We help IT Manger accommodate storage growth and
contain TCO
5
Break Storage apart from Servers
CIFS Folder

• I/O Interface Agnostic SCSI, FC, iSCSI,
  Infiniband
• Device/Vendor Neutral Compaq, EMC, Fujitsu,
  Hitachi, IBM, NEC, SUN..
• Transport Independent SAN/IPtm, iSCSI, FC

IPStor NAS Clients
IPStor iSCSI Client
IPStor SAN/IP Client
IPStor FC Client
IP SAN (LAN MAN WAN)
Fibre Channel SAN
Java Management Console
SCSI
iSCSI Disks
FC Disks
SCSIDisks
Storage Farm
TapesLibrary
6
Remote Disaster Protection
7
Costs of Data Unavailability

• Revenue
• Billing losses
• Investment losses
• Loss of future revenue
• Additional expenses, e.g., shipping
• Productivity
• Working time
• Production delays
• Damaged Reputation
• Customers, suppliers
• Financial markets, banks

Source Gartner Research
8
Key Factors to Be Considered in Disaster Recovery
Planning

• How long can the institution operate without an
  online system or access to data?
• What are the expected losses per hour of down
  time?
• Up to what time should the data be complete?
• To the last transaction
• Small number of modifications can be lost
• For example, last night or last week
• What are the estimated costs to recover lost
  data?

9

• How likely is a disaster? Against which type of
  disaster should one plan?
• Existing corporation infrastructure - for
  example, more than one computer
• center, remote vault or distances
• Electrical power supply, transmission line
  bandwidth and utilization
• What budget can be allocated to the
  disaster-recovery project?

10
Synchronous vs. Asynchronous

• Synchronous Mirror
• Most hardware vendors provide this solution
• Suitable in high speed WAN/MAN link
• Data must write on both sites at same time
• Asynchronous Replication
• Suitable for WAN/MAN link, especially in Asia
  countries
• Data is synchronize based on the time schedule or
  water marks
• Safer and more realistic in most enterprise

11
Why SAN ?

• Consolidated backups and archives
• Disk mirroring, backups to disaster
• recovery sites
• High availability mission critical databases
• Distributed (logical) server clustering
• Disk virtualisation

• Enhance applications performance by freeing up
  enterprise network
• Permits more desktop use of RAID technology

12
Is SAN in the WAN possible ?

• Yes, and SAN traffic loads are typically less
  than many people think
• Very few disks or RAID systems can stream at 10
  Mbytes/sec, although peaks of 30 Mbyte/sec are
  common
• Even high performance UNIX servers can rarely
  exceed 20 Mbyte/sec
• NT servers are much worse, typically /sec

13
Is SAN in the WAN possible ?

• Tape Subsystems are quite slow
• 4 to 10 Mbyte/sec streaming is normal. Peak data
  of 20 Mbyte/sec maximum per interface and drive
• Disk mirroring depends upon application
• Transaction or database system often below 1
  Mbyte/sec
• Backups may be faster, but are limited by system
  (controller / drive) performance

14
IPStor enables the Asymmetric model for remote
replication over IP!
15
Zero-Impact Backup (Off-Site) at the Disaster
Recovery Center
MULTIPLE Customer sites can connect to a single
DRC!!
LAN
Backup Data Flow
App. Servers
IP-SAN (LAN,MAN,WAN)
IPStor Server
IPStor Server
Linux Backup App
R
Replication over IP
R
Tape Library/Drive attached to IPStor Server
Corporate Data Center
Disaster Recovery Center
1) The system administrator can replicate the
data to the Disaster Recovery Center (or Vault).
2) The administrator in the DR Center can
utilize off-the-shelf Linux Backup Applications
(e.g. NetVault, NetBackup/Linux..) on the IPStor
Server to perform high-speed Snapshot
(Point-In-Time Image) backup/restore of the
virtual disks assigned to the application/file
servers.
16
Replication/Migration Path for High-end Users
Only delta (changes) are replicated to remote
site. T3 Line (45 Mb/s) can sustain 10 GB/h of
data transfer.
IPNetwork LAN, MAN, WAN
IPStor Server
IPStor SAN Client
IPStor SAN Client
disk arrays
EMC or others
Local Mirroring (RAID 1 or LVM)
Local Mirroring (RAID 1 or Dynamic Disk)
Corporate Data Center
Application level mirror will take up CPU time.
But only when data is being migrated.
Remote Replication Site
17
Inter-Campus Disaster Recovery Synchronous
(realtime) Mirroring
LAN
App. Servers
IP-SAN (LAN,MAN,WAN)
IPStor Server
IPStor Server
Realtime Mirroring
R
Fibre Channel SAN
R
This Connection is done Via DWDM Connection (
good to 120 KM)
Data Center
Disaster Recovery Center
Storage in the Data Center and the Disaster
Recovery Center can be configured (RAID, JBOD)
differently
18
Inter-Campus Disaster Recovery Asynchronous
Replication
LAN
App. Servers
IP-SAN (LAN,MAN,WAN)
IPStor Server
IPStor Server
Replicate Data over IP
R
R
Disaster Recovery Center
Data Center
Storage Device in the Data Center and the
Disaster Recovery Center can have different
interface (SCSI, SSA, FC), configuration (RAID-X,
JBOD)
19
Impacts

• Lowered Total Cost of Ownership
• Enable the SSP/DRC Business Model
• No addition networking hardware needed
• Win-win for bandwidth, SSP, service, and
  equipment providers
• Only solution that provides asymmetric model for
  data replication

20
IPStor T1 WAN Replication
Percentage of T1 speed
Throughput Mb/s
Latency
Theoretical maximum of T1 is 1.544 Mb/sec
21
(No Transcript)
22
Bell Micro WAN Architecture
23
Summary

• Allows companies to create a less costly Disaster
  Recovery site
• Utilize existing infrastructure for HOT SITE
• Enhance Companies DR Plans

• Allow companies to manage remote location with
  limited number of staffs
• Allow companies to create remote location
  failover sites

24
QA