Skip navigation

Clustering Solutions Feature Summary

FEATURE Amdahl (LifeKeeper 1.0 on Microsoft (Wolfpack Beta 2) Vinca (Vinca StandbyServer Data General Octopus (Octopus SASO 2.0) Network Specialists Cubix (RemoteServ/IS)

EnVista Servers)on IBM PC Servers for NT) on IBM PC Servers (NT Cluster-in-a-Box) (Double-Take 1.3 Beta)

Supported hardware EnVista Servers Model FS and ES Any Wolfpack-certified NT server + Any NT-certified Intel-based serverAV3600R 1-4 processorAny NT-supported hardwareAny Intel CPUPentium and Pentium Pro (models/CPU types) with LVS 4500disk array

Maximum number of nodes3Wolfpack phase 1: 2 22UnlimitedUnlimited4094 nodes per cluster (designed), 50 to 100 nodes per cluster (practical); unlimited number of clusters

Mixed hardware (different models Different models: yes; different Yes; only if certified and not on an Yes; Intel-basedNoYesYesYes; load balancing accounts for and different manufacturers)manufacturers: noAlpha-Intel comboperformance differences

Proprietary hardware Amdahl systems only NoNoYesNoNoNo; processors conform to industry standard ISA and PCI architecture

Active/active or active/standbyEitherEitherActive/standbyActive/activeEitherActive/activeActive/active; all servers provide simultaneous user session support with session recovery in the event of server failure

Cluster aliasNoYes; cluster can have one network NoNo; next releaseNoNoYes; cluster names default to domain name, as can each virtual server and workgroup but can be changed running on the clusterwithout system restart

Remote administration and NoYes; graphical administrator's NoYes; Web-basedYes; remote installation and Windows GUIYes; cluster and hardware configuration (methods)console can be run from any administrationmanagement from local or remote NT server or workstation on (dial or LAN/WAN) Win95 or NT the networkworkstations

GUIYesYesYesYes; browserYesYesYes

Manual controls (failover/failback)Yes/yesYes/yesYesYes/yesYes/yesYes/yesYes/yes; individual servers can be reassigned to other clusters without affecting cluster operation; auto recovery from server failures

Reconfigure without rebootingYesYesYesYesYesYesYes; within current cluster

Based on loss of heartbeat YesYesYesYesYesYesN/A
(mirroring and network link)

Based on application failure Not directly supported by LifeKeeper, YesYes; with use of third-party productsYesNoNoN/A (task manager)but can coexist or be scripted

Database rollbackNot directly supported by LifeKeeper, YesNoYesNoNoN/A but can coexist or be scripted

Failover scriptsYesProvides wizards for setting up YesYesYesYesN/A applications for failover, which can trigger an application script

SNMP management (for use with NoYes; Microsoft is working with NoYesNoNoYes; provided through GlobalVision SMS, CA, etc.)vendors to include necessary management application SNMP MIBs in Wolfpack

Related tools (add-ons, support N/ASeveral tool vendors plan to Yes; specific scripts for popular Agent toolkit for developing custom N/ANoNot required; all tools are for other software)support Wolfpack, including applicationsapplication agentsintegrated in bundled package Computer Associates, NSI, Octopus, Veritas, and Vinca

Client agents (e.g., NT, UNIX, Not requiredNo special client software required;Supports all NT clientsAny IP-based (socket) clientNT, Win95NTNT, Win95, Win3.1x, and DOS NetWare)supports any client (e.g., DOS, Win3.1, Win95, NT Workstation, UNIX, and Apple Macintosh) that can connect to an IP network and use IP protocols to find server services

Automatic reconnect (user YesBecause Wolfpack does not YesYes; for Windows clientsYesYesYes; activated through server option transparency)require special client software, to automatically redial user in the this feature depends on the client event of a disconnected session component of the client/server application

Network protocolsTCP/IP, NetBEUIIP and all related protocols Supports NT standard protocols; TCP/IPAny NT-supported protocolsAny NT-supported protocolsTCP/IP, IPX/SPX, NetBEUI/NetBIOS, supported by NT Server mirroring traffic is proprietaryasynchronous connection (e.g., TCP, NBT, SMB, UDP)

Number of shared buses2 As many as the servers can Based on a shared-nothing 0; uses split busUnlimitedN/AN/A (Each LVS 4500 storage system support; no limits exist in architecture; disks and disk has two redundant controllers, each the Wolfpack softwarechannels are not shared attached to a separate SCSI bus that is attached to both nodes)

FEATURE Amdahl (LifeKeeper 1.0 on Microsoft (Wolfpack Beta 2) Vinca (Vinca StandbyServer Data General Octopus (Octopus SASO 2.0) Network Specialists Cubix (RemoteServ/IS)

EnVista Servers)on IBM PC Serversfor NT) on IBM PC Servers(NT Cluster-in-a-Box)(Double-Take 1.3 Beta)

RAID (levels supported)Levels 0, 1, 3, 5Works with built-in software RAID Supports all hardware-based RAID Levels 0, 1, 3, 5Any levelN/ALevels 1 through 5 (as supported of NT Server (RAID 0, 1, 5) and levelsin WinFrame/NT and NT 4.0 hardware RAID devices within implementation; Cubix processor validated cluster configurationsonboard SCSI)

Two-way simultaneous disk accessYes to same physical disk; no to NoN/ANoYesN/AN/A
same logical volume

Shared SCSI or disk mirroringShared diskShareable disk; only one server at a Disk mirroringShared SCSI (cabinet sharing, EitherN/AYes; as supported in WinFrame and time can own each disk; ownershipdrives logically partitioned NT 4.0 can be triggered automatically by between servers until failover) server or application failure, or transferred manually by the Administrator

Automatic failbackYesYesNoYesYesNoYes; recovered server automatically reintegrates in cluster without Administrator intervention

Server failover (name, YesYesYes; supports complete server LANS, disk volumesYesYesYes; suspended sessions can be resources, etc.)identity failoverresumed within any server in cluster

Disk failoverYesYesN/AYesNoYesYes; with RAID implementation

Connection failoverYes, on failure of secondary; YesYes NoYesYesYes; auto dialback in the event of otherwise, manualdisconnected session; suspended sessions can be restored within any server in cluster

IP failoverYes YesYesYesYesYesYes

Integrated application failoverNot directly supported by LifeKeeper, YesYesYes; SQL Server, Oracle, IIS, NoNoYes but can coexist or be scriptedNetscape

Cluster-aware application failoverNot directly supported by LifeKeeper, YesYesNoNoNoYes but can coexist or be scripted

APIs or application recovery kitsYesYes; but not necessary for failover YesYes; agentsYesNo; scripting possibleN/A operation

Microsoft Exchange and YesExchange: yes (Exchange 5.0 with Yes; both tested; installation scripts Exchange: yes; Notes: noYesNoN/A Lotus Notes failoverSP1); Notes: Iris Associates has developed for each been involved in the Wolfpack Open
Process and Beta Test, however the company hasn't publicly announced plans relative to Wolfpack

Other DBMS's (Oracle, Sybase, Yes; OracleSQL Server 6.5 and later: yes; Yes; SQL Server and DB2Oracle, SQL ServerYesYes; using application journalN/A etc.) failoverOracle has publicly discussed its plans for Oracle Failsafe on Wolfpack clusters; other major DBMS vendors are involved in the Wolfpack Open Process and Beta Test but have not publicly announced plans relative to Wolfpack

IIS and Netscape failoverNot directly supported by LifeKeeper, IIS: yes; Netscape is involved in the Yes; both tested; installation scripts YesYesYes; multiple IPN/A but can coexist or be scriptedWolfpack Open Process and Beta developed for each Test but has not publicly announced plans relative to Wolfpack

Hardware independence (support LifeKeeper available from other YesYesYes; with FirstWatchYesYesYes; supports multiplatform two different manufacturers) vendors clustering

Support for three or more different YesWolfpack phase 1: No; Wolfpack NoNoYesYesYes; 50 nodes to 100 nodes practical, nodes per cluster phase 2: Yes4094 nodes design limit

Dynamic load balancing NoWolfpack phase 1: No; Wolfpack NoNoNoNo; dynamic network performance Yes; user adjustable to meet (performance clusters)phase 2: Yesmonitoring and performance environment thresholds for replication traffic

Failover based on application Not directly supported by LifeKeeper, Yes; application must exploit APIs to NoNo; but can write agent for this NoNo No performancebut can coexist or be scriptedcreate separate diagnostics for "looks capability alive" and "is alive" heartbeats

LifeKeeper: $2000 per server, $750Will be finalized when closer to $3995 $110,000$1199 for Octopus Server; $1499 $1875 $14,495 per recovery kit; EnVista FS: $30,720; releasefor Octopus SASO; $499 for LVS 4500: $93,737WinStation

800-223-2215Microsoft 206-882-8080, 801-223-3100, 508-898-5000, 215-579-5600, or 800-919-1009, 201-656-2121, 800-829-0550

AVAILABILITYAvailable now Summer 1997 Available now Available now Available now Available now Available now

Corrections to this Article:

  • The "Clustering Solutions Feature Summary" contained two inadvertent mistakes regarding Data General's NT Cluster-in-a-Box. First, the hardware is not proprietary in nature; the product uses standard Intel processors. Second, the price of $110,000 was for the high-end, quad-processor model we tested in the "Windows NT Magazine" Lab; the entry price for NT Cluster-in-a-Box is $70,000, which includes all necessary hardware and software coomponents to implement an NT custer.
Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.