Clustering

1
Clustering Fibre Channel

• Next wave in PC Computing

2
Trends in Clustering

• Today clusters are a niche Unix market. But,
  Microsoft will bring clusters to the masses
• Microsoft has announced NT clusters
• SCO has announced UnixWare clusters
• Sun has announced Solaris / Intel clusters
• Novell has announced Wolf Mountain clusters
• In 1998, 2M Intel servers will ship
• 100K in clusters
• In 2001, 3M Intel servers will ship
• 1M in clusters (IDCs forecast)
• Clusters will be a huge market and RAID is
  essential to clusters
• Fibre Channel and storage area networks are ideal
  for use in clusters.

3
Why is the IT Market Moving To Clustering

• Growth In On-Line Applications
• E-Commence
• On-Line Transaction Processing
• Web Servers
• Real Time Manufacturing
• Server Must Be On Line
• 24 Hours a Day, 7 Days A Week, 365 Days a Year
• New Levels Of Fault Tolerance
• Performance Must Scale
• Without Our Data, Our Business Is Dead

4
What Are Clusters?

• Group of independent servers which
• Function as a single system
• Appear to users as a single system
• And are managed as a single system
• Clusters are virtual servers

5
Why Are Clusters Important?

• Clusters Improve System Availability
• Clusters Enable Application Scaling
• Clusters Simplify System Management
• Clusters Are A Superior Server Solution

6
Clusters Improve System Availability

• When a networked server fails, the service it
  provided is down.
• When a clustered server fails, the service it
  provided failover and downtime is avoided.

Mail Server
Internet Server
Mail Internet
Networked Servers
Clustered Servers
7
Clusters Enable Application Scaling

• With networked SMP servers, application scaling
  is limited to a single server
• With clusters, applications scale across multiple
  SMP servers (typically up to 16 servers)

Oracle
Voice mail Oracle E-Mail
E-Mail
Voice mail
8
Clusters Simplify System Management

• Clusters present a Single System Image the
  cluster looks like a single server to management
  applications
• Hence, clusters reduce system management costs

9
An Analogy to RAID

• RAID Makes Disks Fault Tolerant
• Clusters make servers fault tolerant
• RAID Increases I/O Performance
• Clusters increase compute performance
• RAID (GAM) Makes Disks Easier to Manage
• Clusters make servers easier to manage

10
Two Flavors of Clusters

• High Availability Clusters
• Microsoft's Wolfpack 1
• Compaqs Recovery Server
• Load Balancing Clusters
• (a.k.a. Parallel Application Clusters)
• Microsofts Wolfpack 2
• Digitals VAXClusters

Note Load balancing clusters are a superset of
high availability clusters.
11
Available Clustering Software

• Failover Clusters
• Microsoft Clustered Servers (Enterprise Server
  4.0)
• Compaqs Recovery Server
• NCR Lifekeeper
• Digital NT Clusters
• UnixWare 7 Reliant
• NSI DoubleTake failover in Wide Area Networks
• Novells Vinca Corp. Standby Server
• Load Balancing Clusters
• Microsofts Clustered Servers (NT 5.0)
• Digitals VAXClusters
• IBMs HACMP
• Oracles Parallel Server
• Novells Orion

12
Failover Example

• Two node clusters (node server)
• During normal operations, both servers do useful
  work
• Failover
• When a node fails, applications failover to the
  surviving node and it assumes the workload of
  both nodes

Mail
Web
Mail Web
13
Load Balancing Example

• Failover (contd)
• Multi-node clusters (two or more nodes)
• Load balancing clusters typically runs a single
  application, e.g. database, distributed across
  all nodes
• Cluster capacity is increased by adding nodes
  (but like SMP servers, scaling is less than
  linear)

14
Load Balancing Example (contd)

• Cluster rebalance the workload when a node dies
• If different apps are running on each server,
  they failover to the least busy server or as
  directed by preset failover policies

15
Sharing Storage In Clusters
Two Models For Storage

• 1. Shared Nothing Model
• Microsofts Wolfpack Cluster
• 2. Shared Disk Model
• VAXClusters

16
What is Fibre Channel?

• Fibre channel is a high performance multiple
  protocol data transfer technology. Fibre
  channels primary task is to transport data
  extremely fast with the least possible delay.
  Fibre channel is a serial interconnect standard
  that allows servers, storage devices, and
  workstation users to share large amounts of data
  quickly. Fibre channel gives networks superior
  data transfer speeds, flexible topology, and
  flexible upper-level protocols. Fibre channel
  easily handles both networking and peripheral I/O
  communication over a single channel.

17
Key Features of Fibre Channel

• Data Transfer Rates up to 100 MB/sec per
  connection
• Three Fibre channel connection schemes
• Point-to-Point
• Provides a single connection between two servers
  or a server and its RAID storage.
• Switched Fabric
• Using a Fibre channel switch this connection
  allows each server or RAID storage to be
  connected point-to-point to a switch. This
  method allows for the construction of massive
  data storage and server networks.
• Loop
• Connects up to 126 servers, RAID systems, or
  other storage devices in a loop topology.

18
Key Features of Fibre Channel

• Long Distance Connections
• Up to 10 kilometers
• High reliability connections providing high data
  integrity
• Low Overhead
• Multiple media types support
• Copper
• lower cost
• not as reliable as fibre optic cable
• distances up to 20 meters between nodes
• Fibre optic
• higher cost
• more reliable than copper
• distance up to 10,000 meters between nodes
• Allows for creation of Storage Area Networks

19
Benfits of Fibre Channel

• Runs multiple protocols over a single network or
  loop
• Ideal solution for clustering of servers
• Enables multiple servers to share storage and
  RAID systems
• Scales from small numbers of peripherals attached
  short distances apart, to large numbers attached
  many kilometers apart.

20
Benfits of Fibre Channel (contd)

• Delivers speeds that are 2.5 to 250 times faster
  than existing communication and I/O interfaces
• Overcomes today's network performance limitations
  regarding error detection and recovery
• Provided low-cost and reliable performance for
  distance connections
• Offers flexible protocols and topologies to best
  leverage existing technology investments

21
Cluster Interconnect

• This is about how servers are tied together and
  how disks are physically connected to the cluster.

22
Cluster Interconnect

• Clustered servers always have a client network
  interconnect, typically Ethernet, to talk to
  users, and at least on cluster interconnect to
  talk to other nodes and to disks.

23
Cluster Interconnect

• Or they can have Two Cluster Interconnects
• One for nodes to talk to each other
• Heartbeat Interconnect typically Ethernet
• And one for nodes to talk to disks
• Shared Disk Interconnect typically SCSI or Fibre
  channel

Cluster Interconnect
Shared Disk Interconnect
24
NT Cluster with Host-Based RAID Array

• Each node has
• Ethernet NIC -- Heartbeat
• Private system disks (HBA)
• PCI-based RAID controller - SCSI or Fibre, such
  as Mylexs eXtremeRAID
• Nodes share access to data disks but do not share
  data

25
NT Cluster w/ SCSI External RAID Array

• Each node has
• Ethernet NIC -- Heartbeat
• Multi-channel HBAs connect boot disk and
  external array
• Shared external RAID controller on the SCSI
  Shared Disk Interconnect

26
NT Cluster w/ Fibre External RAID Array

• Fibre Channel Host to Ultra2 SCSI LVD RAID system
• Fibre Channel Host to Fibre Channel RAID system

Heartbeat Interconnect
Shared Disk Interconnect
27
Internal vs External RAID in Clustering

• Internal RAID
• Lower cost solution
• Higher performance in read-intensive applications
• Proven TPC-C performance enhances cluster
  performance
• External RAID
• Higher performance in write-intensive application
• Write-back cache is turned-off in PCI-RAID
  controllers
• Higher connectivity
• Attach more disk drives
• Greater footprint flexibility
• Until PCI-RAID implements fibre

28
Active / Active

• Fibre Channel Host - Ultra2 LVD RAID Controller
  in a Clustered Loop

29
SF (or FL) Active / Active Duplex

• Fibre Channel Host - Ultra2 LVD RAID Controller
  in a Clustered Environment

FC Disk Interconnect
Dual FC Array Interconnect
30
Fibre Host
Active / Active Duplex
Single FC Array Interconnect
31
FC - FC in Cluster
Dual FC Array Interconnect
32
Mylex - The RAID Clustering Experts

• eXtremeRAID 1100
• AcceleRAID 250
• DAC SF
• DAC FL
• DAC FF