Storage Area Network (SAN)
Storage Area Network (SAN)
Storage Area Network (SAN)
Outline
Shared Storage Architecture Direct Access Storage (DAS) SCSI RAID Network Attached Storage (NAS) Storage Area Network (SAN) Fiber Channel and Fiber Channel Switch
2
Storage domain
Database (dbms)
Host Network
Block aggregation
Device
Block layer
4
2010+
Quick Overview
DAS
Storage Type sectors
NAS
shared files
SAN
blocks
Data Transmission
Access Mode Capacity (bytes) Complexity
Management Cost (per GB)
IDE/SCSI
clients or servers 109 Easy High
Fibre Channel
servers
1012
Difficult Low
8
DAS
clients
NAS
FC-SAN
servers
FC Switch
storage
9
Small Server
SCSI Channel
Used Used
clients
Large Server
10
Client (Initiator)
response
12
13
14
TCP IP
Physical interface
Fibre Channel
Ethernet
15
SCSI Domain
SCSI Service Delivery Subsystem
16
Provide fault tolerance for shared data and applications Different implementations: Level 0-5 Characteristics: Storage Capacity Speed: Fast Read and/or Fast Write Resilience in the face of device failure
17
RAID Functions
Striping
Write consecutive logical byte/blocks on consecutive physical disks
Mirroring
Write the same block on two or more physical disks
Parity Calculation
Given N disks, N-1 consecutive blocks are data blocks, Nth block is for parity When any of the N-1 data blocks is altered, N-2 XOR calculations are performed on these N-1 blocks The Data Block(s) and Parity Block are written Destroy one of these N blocks, and that block can be reconstructed using N-2 XOR calculations on the remaining N-1 blocks Destroy two or more blocks reconstruction is not possible
18
1 1 1
disk 1: odd bits
001
disk 2: even bits
110
parity bits (even parity)
Example 2: 1 0 1 0 1 1
1 0 3k+1 bits
01 3k+2 bits
11 3k bits
RAID Types
RAID 0
Stripe with no parity (see next slide for figure)
RAID 1
Mirror two or more disks
RAID 3
Synchronous, Subdivided Block Access; Dedicated Parity Drive
RAID 5
Like RAID 4, but parity striped across multiple drives
20
RAID 0
RAID 1
Disk Mirror
21
22
RAID 3
RAID 5
23
SMB
NetBIOS
TCP IP 802.3
26
Clients
Servers
Storage Devices
28
Case Study
Product: MicroNet ProtinumNAS Storage: 1TB and more Price: < $1,000 Protocol: CIFS/SMB, RAID
29
Discussion
Need: a lot more storage (hundreds of GB) and a scalable solution (~2 TB) for home users
USB (USB 2.0) to a server, up to 480M bps Firewire (IEEE 1394) to a server, up to 3.2G bps SCSI to a server: up to 320MB (3208 bps) NAS: no need for a server
SAN Benefits
Storage consolidation Data sharing Non-disruptive scalability for growth Improved backup and recovery Tape pooling LAN-free and server-free data movement High performance High availability server clustering Data integrity Disaster tolerance Ease of data migration Cost-effectives (total cost of ownership)
32
The lines are becoming blurred between the two technologies now, and while the SANversus-NAS debate continues, the fact is that both technologies complement each another.
33
Fibre Channel
Fiber Channel is well established in the open systems environment as the underlining architecture of the SAN. Fibre Channel is structured with independent layers, as are other networking protocols. There are five layers, where 0 is the lowest layer. The physical layers are 0 to 2. These layers carry the physical attributes of the network and transport the data created by the higher level protocols, such as SCSI, TCP/IP, or FICON.
34
Designed to transport multiple protocols, such as HIPPI, IPI, SCSI, IP, Ethernet, etc. Full duplex medium Channels are established between the originator and the responder. Transfer rate from 100MB/s to Gigabits/s Distance >10 km (single mode fiber) Multi-layer stack functions (not mapped to the OSI model)
35
FC Protocol Layers
Gbaud
Gbaud
IPI: Intelligent Peripheral Interface HIPPI: High Performance Parallel Interface SCSI SBCCS: Single Byte Command Code Set
36
FC Layers: 0 & 1
37
FC Layer 2
Port_ID
Port_ID
38
FC Address
FC node a node has many ports FC port the end point of a link (either transmission or reception). Port ID: a unique 24-bit address for a port In Frame Header (see Slide-49), there are two fields: Source address (transmission port) and Destination address (reception port)
39
N_Port ID: 24-bit port address An N_Port has a point-to-point connection with another N_Port. An N-Port may be attached to a fabric port, F_port. Connection between fabric switches is via expansion ports, E_ports. A switch port, if configured for either one, is a generic port, G_Port.
40
FC Port Naming
Fiber Channel Host Fiber Channel Switch Fiber Channel Switch Fiber Channel Stores
N-Port
F-Port
E-Port
E-Port
F-Port
N-Port
41
FC Layers 3 & 4
(one)
42
SAN Topologies
Fibre Channel based networks support three types of topologies:
Point-to-point Loop (arbitrated) shared media Switched
43
FC - Point-to-Point
The point-to-point topology is the easiest Fibre Channel configuration to implement, and it is also the easiest to administer. The distance between nodes can be up to 10 km
44
Data SCSI FC
Data SCSI FC
45
Arbitrated Loop
Shared Media Transport
Similar in concept to shared Ethernet
Not common for FC-based SAN Commonly used for JBOD (Just a Bunch of Disks) An arbitration protocol determines who can access the media.
ARB primitive
46
Rx
Tx Rx Tx Rx
47
48
49
Switched FC SAN
Fibre Channel-switches function in a manner similar to traditional network switches to provide increased bandwidth, scalable performance, an increased number of devices, and, in some cases, increased redundancy. Fibre Channel-switches vary in the number of ports and media types they support. Multiple switches can be connected to form a switch fabric capable of supporting a large number of host servers and storage subsystems
50
FC Switched SAN
Servers
Clients
51
Data SCSI FC
52
Clients
53
Repeat Overview
DAS
Storage Type sectors
NAS
shared files
SAN
blocks
Data Transmission
Access Mode Capacity (bytes) Complexity
Management Cost (per GB)
IDE/SCSI
clients or servers 109 Easy High
Fibre Channel
servers
1012
Difficult Low
54
55
Course Outline
IP over FC (RFC 2625) IP-SAN
iSCCI (RFC 3720)
Storage Virtualization
56
SAN FC/IP Gateway Data IP L2 IP L2 IP RFC 2625 FC-based Storage Device Data IP RFC 2625
PHY
PHY
FC
FC SAN
FC
58
SAN FC/IP Gateway Data IP L2 PHY IP L2 PHY IP RFC 2625 FC FC IP RFC 2625 FC IP L2 PHY FC/IP Gateway NAS-based Storage Device Data IP L2 PHY
SAN
59
IP-SAN
60
61
IP Network Capabilities
62
IP-SAN Protocols
63
IP - SAN
IP
IP FC -SAN
Storage Devices
64
IP storage networking carrying storage traffic over IP Uses TCP, a reliable transport for delivery Can be used for local data center and long haul applications Two primary IETF protocols/standards:
iSCSI Internet SCSI allows block storage to be accessed over a TCP/IP network as though it were locally attached
IP TCP iSCSI SCSI Data
FCIP Fibre-Channel-over-IP used to tunnel Fibre Channel frames over TCP/IP connections
IP
TCP FCIP FC SCSI Data
65
66
iSCSI Benefit
67
68
iSCSI IP FC -SAN IP
iSCSI
iSCSI
69
iSCSI
NFS
70
Conclusion: 1. Comparable performance in character read/write and block write 2. Significant advantage of iSCSI in block read (20-25%)
71
72
73
Both iSCSI and SAN use Block I/O to transport data, whereas NAS uses File I/O. SAN offers better performance (c.f. NAS), but is more expensive and requires a higher skill set to implement. iSCSI and NAS offer better pricing and skills may already be in place to implement them. Both SAN and iSCSI offer the performance benefit of Block I/O.
74
FC over IP
SAN IP SAN
Data
an IP tunnel for FC-based SAN Application: interconnect SAN over IP-WAN. FCIP FC 0-2 TCP IP IP TCP IP FCIP FC 0-2
Data
SCSI
FCP FC 0-2
SCSI
FCP FC 0-2
L2
PHY
L2
PHY
L2
PHY
75
Storage Virtualization
PHYSICAL Virtualization LOGICAL
FC -SAN
IP
Storage Virtualization
Definition: storage virtualization hides the physical storage from applications on host systems, and presents a simplified (logical) view of storage resources to the applications. Virtualization allows the application to reference the storage resource by its common name where the actual storage could be on a complex, multilayered, multipath storage networks. RAID is an early example of storage virtualization.
77
Virtualization Intelligence
Host-Based: storage virtualization could be implemented on the host through Logical Volume Management (LVM) which provides the logical view of the storage to the host operating system. Switch-based: intelligence of storage virtualization could be implemented on the SAN switches. Each server is assigned a Logical Unit Number (LUN) to access the storage resources.
Switch-based virtualization could be in dual configuration for high availability. Pros: ease of configuration and management ; redundancy/high availability Cons: potential bottleneck on the switch; higher cost
78
Storage Virtualization
LVM LVM SAN Switch
LUN
LUN
RAID
RAID
JBOD
RAID
RAID
JBOD
79
SAN Challenges
Standards
ANSI T10 (SCSI) ANSI T11 (FC), IETF (IP-SAN), Ethernet (IEEE 802.3), SNIA, etc.
Interoperability High availability and data synchronization between remote locations Convergence
DAS, NAS, FC-SAN => IP-SAN
Management Security
80
Summary
Needs for large storage continual growth
109 (G) => 1012 (T) => 1015 (P) => 1018 (E) .
81