94e2859571bcbb37bda4bd1a134ede76.ppt
- Количество слайдов: 34
Marathon ever. Run™ Next Generation Availability Martin Kyprianides EMEA Systems Engineer Marathon Technologies
Agenda • Market • About Marathon • ever. Run Solutions – ever. Run FT – ever. Run HA – ever. Run Split. Site • Product Demo / Q&A 2
The Growing Need for Availability • Application availability – No longer a luxury for any business • Cost of downtime – Has risen dramatically in recent years • Regulatory Compliance – New and changing laws governing availability • Server consolidation – Will drive high availability in the enterprise 3
Lim m ire qu Re us tin uo Ap ite Da pli d. R ta cat eq uir A v ion em ai s en la b a n t ili d ty Co n en t The Availability Pyramid AL 4: Continuous Availability “Real-time” 24 x 7 Operations, No User Interruption AL 3: High Availability Clustering/failover User Interruption. AL 2: Recovery Data Back-up & Data Replication User interruption. AL 1: Reliability Hot swappable components, ECC memory All service stops. AL 0: Unprotected Servers No redundant system components, no protection for data, application, or OS. All service stops. 4 Source: IDC
Comprehensive Solutions Automated Application Availability • What we do? – Marathon delivers Next Generation Availability for Windows applications in a simple, comprehensive & automated solution – Superior application availability on standard Windows servers – Fault Tolerance and High Availability for ANY Windows application – No scripting or modifications required 5
Comprehensive Solutions Automated Application Availability • How do we do it? – A patented virtualization architecture that creates a single virtual server environment – Compute. Thru® failures vs. traditional failover approaches – Provide Intelligent Policy Management – automatic vs. scripted 6
Marathon’s Technology Vision Next Generation Availability - A suite of software products designed to meet a broad range of availability requirements 7
Marathon Momentum • Core technology & IP development began in 1993 – A history of invention - 15 US Patents • Launched software only solution in January 2004 – Further expanded product line in April 2006 • Market traction – Over 1, 400 Companies are now using Marathon – Shipped over 2, 500 software licenses to-date • Global – Offices in US, Europe, APAC – Worldwide Channel Partner Network 8
Availability Market Verticals (Line of Business) Horizontals (Infrastructure) • MARATHON’S “Sweet Spot” • 60% of current revenue Large Enterprise Medium Enterprise • Rapidly Growing • 30% of current revenue • Key Verticals: • Manufacturing • Financial Services • Media/Broadcasting • Gaming • Healthcare • Key Applications: • Exchange • SQL Server • Oracle • Notes • ERP/CRM • Strong ISV support Small Enterprise 9 • Emerging • “Availability for the Masses” • Far easier, “Automated” and less expensive
Marathon Solutions Proven in Mission Critical Environments Financial Media Airports Government Manufacturing Resorts/Casinos Energy $1 Billion Hedge Fund Logan Pharmaceutical 10 Technology Telecom
ever. Run™ Solutions
ever. Run FT
Marathon ever. Run FT • Flagship continuous availability offering • Complete hardware redundancy – Compute-through all failures – Ensures zero downtime through minor and major faults – Ensures zero data loss • Unique innovation in lockstepping – 15 patents, others pending • Mainframe-class availability (99. 999%+ uptime) – Equal to ~5 minutes of unplanned downtime per annum • Non-Intrusive – Supports any Windows application – No need for cluster-aware versions that cost more 13
Marathon ever. Run FT How It Works • Lockstep processing, requires 100% identical hardware – Both servers process transactions in parallel • Both systems completely synchronized – Memory – I/O – Storage Redundant Gigabit Ethernet Links Co. Server 1 LAN 14 Fault Tolerant Virtual Server Co. Server 2
ever. Run FT Architecture Virtual Server Hosted on Both Co. Servers Lockstep ever. Run FT Environment VS OS Public Ethernet 15 VS Processor Co. Server Processor PCI-X Gig-e Main Memory Storage VS OS PCI-X Co. Server OS Co. Server 1 Co. Server 2 Storage Public Ethernet
Marathon ever. Run FT Compute. Thru® Server & Component Failure Marathon ever. Run FT Application Environment Both Servers Operating Redundantly 16 Marathon ever. Run FT Application Environment Operational Server Failed Server
Device Virtualization Building Redundancy • Creates a virtual Windows server – Single IP address – Single hostname – Single MAC address • Like physical devices are redirected Virtual NIC – Disks – Network Interfaces – CD-ROM – Tape Drives • Paired to create a single virtual device • Failed devices do not affect application availability 17 ever. Run Virtual Server Application Environment Virtual Disk LAN
RAID 1 Disk Mirroring Logical View Virtual Server Application Environment • Single write by the application • Simultaneous writes to each physical disk Virtual Disk Mirrored Physical Disks 18
System Management Marathon Manager Interface n n 19 Monitor Health and Status Check Failure Detail Enable/Disable Devices Make Operational Settings w Mirror Copy Bias w Auto Recovery w Co. S Link Optimization w Delta/Full Copy
ever. Run FT Hardware/Software Requirements • Intel-based P 4/Xeon or AMD servers – 2 Processors (Physical / Logical) – 1 Processor with Hyper. Threading or Dual-Core • 512 MB RAM (1 GB, recommended, max 4 GB) • 2 NICs (3 or 4 recommended) • 1 NTFS disk drive (2 or more recommended) • OS options for both Co. Server and FT (one for each Co. Server, one for the FT) – Windows 2000, Server or Advanced – Windows 2003, Standard or Enterprise • Single license for deployed application(s) 20
ever. Run HA
Marathon ever. Run HA The Next Generation Availability Technology • Minimizes downtime – Applications Compute. Thru® most device failures – Downtime only as a result of a complete server failure • Scalability & Flexibility – Highly scalable – 2 x, 4 x, 8 x (physical) SMP – Up to 32 GB RAM – No need for identical server pairs • Utilize existing hardware • Broad Platform Compatibility – Intel / AMD multi-core CPU designs – SAN / DASD storage configurations 22
ever. Run HA Architecture Virtual Server Hosted on Active Co. Server Virtual Server OS Public Ethernet 23 Ready VS Processors Co. Server Processors PCI-X Gig-e Main Memory Storage Active VS Processors PCI-X Active Co. Server OS Active Co. Server 1 Ready Co. Server 2 Storage Public Ethernet
RAID 1 Disk Mirroring Logical View Virtual Server Application Environment Identical to ever. Run FT • Single write by the application • Simultaneous writes to each physical disk Virtual Disk Mirrored Physical Disks 24
Easy to Operate & Maintain Compute. Thru® Component Failures Storage Mirror Storage Disk Failure Occurs Storage I/O is redirected to redundant component (identical to ever. Run FT) 25 Storage
Automated Recovery Complete Server Failure Recovery Server Failure Occurs Active Co. Server operating normally 26 Primary server fails Ready Co. Server OS and app restart
Online Migration Reduce Scheduled Downtime • Requires 100% identical hardware App Online Migration Application running in Active Co. Server Application running in Ready Co. Server Application Operations Never Stop Clients Never Disconnect 27 App
ever. Run HA Minimum Configuration • Processors – Intel-based P 4/Xeon or AMD – One single-core HT processor – Two single-core processors – One dual-core • Physical Memory: 1 GB (Co. Server defaults to 256 MB) • Storage: 1 drive (NTFS for all boot partitions) • Networks – (1) GB Ethernet for Co. Server Link – (1) 10/1000 for client (redirected) network • Storage Controllers – SCSI Host Bus Adapter (FC, Copper), IDE Controller, Serial ATA, RAID • Windows Server 2003 SP 1, Standard Edition 28
ever. Run HA Maximum Configuration • Processors: 4 dual-core HT (16 -way) – Standard Edition supports up to 4 -way – Enterprise Edition supports up to 8 -way • Physical Memory: Maximum allowed by Windows Server 2003 – Standard Edition supports up to 4 GB – Enterprise Edition supports up to 32 GB • Networks (7 total) – (2) Co. Server – (1) Co. Server Management – (4) Redirected • Windows Server 2003 SP 1 Enterprise Edition 29
ever. Run Split. Site
ever. Run Split. Site® Survive A Disaster Rather than Recover From One • Permits geographical separation of Co. Servers to support disaster tolerant configurations – Different rooms in a building – Different buildings on a campus – Different buildings in different towns/cities • Provides necessary integrity for failures where Co. Servers are physically separated 31
ever. Run Split. Site - Networking Flexible Co. Server Connectivity • Co. Server links can be: Dedicated fiber – Private switched network Management Network – Public internet • Bandwidth requirements (per link), for fault handling & performance WAN Public Network Switched – Minimum bandwidth = 51 Mbps – Recommended bandwidth > 155 Mbps Management Network WAN • Latency requirement: – Maximum latency = 5 msec – Recommended latency < 2 msec – Co. Servers on same IP subnet or v. LAN 32 Public Network Management Network WAN Public Network
ever. Run Split. Site - Quorum Service Protects Against Network Failures Co. Server Links Management Network Public Network Quorum Server 33 • Service runs on any networkaccessible server • Scenario: Simultaneous break in all Co. Server communications (e. g. , an errant backhoe) • Co. Server that can communicate with quorum server stays online • Co. Server that can’t communicate with quorum server takes itself offline • Split-brain operation is avoided while maintaining application availability • Redundant Quorum Servers – If one quorum server fails, Co. Servers elect a second quorum server – Maximum protection from site disasters
Thank You. Questions?
94e2859571bcbb37bda4bd1a134ede76.ppt