User:Sateesh Adusumilli/sandbox

CLARiiON Foundations                        ©  2006 EMC Corporation. All rights reserved. Welcome to CLARiiON Foundations. The AUDIO portion of this course is supplemental to the material and is not a replacement for the student notes accompanying this course. EMC recommends downloading the Student Resource Guide from the Supporting Materials tab, and reading the notes in their entirety. Copyright © 2006 EMC Corporation. All rights reserved. These materials may not be copied without EMC's written consent. Use, copying, and distribution of any EMC software described in this publication requires an applicable software license. THE INFORMATION IN THIS PUBLICATION IS PROVIDED “AS IS”. EMC CORPORATION MAKES NO REPRESENTATIONS OR WARRANTIES OF ANY KIND WITH RESPECT TO THE INFORMATION IN THIS PUBLICATION, AND SPECIFICALLY DISCLAIMS IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Celerra, CLARalert, CLARiiON, Connectrix, Dantz, Documentum, EMC, EMC2, HighRoad, Legato, Navisphere, PowerPath, ResourcePak, SnapView/IP, SRDF, Symmetrix, TimeFinder, VisualSAN, “where information lives” are registered trademarks. Access Logix, AutoAdvice, Automated Resource Manager, AutoSwap, AVALONidm, C-Clip, Celerra Replicator, Centera, CentraStar, CLARevent, CopyCross, CopyPoint, databaseXtender, Direct Matrix, Direct Matrix Architecture, EDM, E-Lab, EMC Automated Networked Storage, EMC ControlCenter, EMC Developers Program, EMC OnCourse, EMC Proven, EMC Snap, Enginuity, FarPoint, FLARE, GeoSpan, InfoMover, MirrorView, NetWin, OnAlert, OpenScale, Powerlink, PowerVolume, RepliCare, SafeLine, SAN Architect, SAN Copy, SAN Manager, SDMS, SnapSure, SnapView, StorageScope, SupportMate, SymmAPI, SymmEnabler, Symmetrix DMX, Universal data Tone, VisualSRM are trademarks of EMC Corporation. All other trademarks used herein are the property of their respective owners. CLARiiON Foundations Upon completion of this course, you will be able to: y Draw and describe the basic architecture of a CLARiiON Disk Array y Differentiate among the architectures of the various CLARiiON models y List the data protection options available on CLARiiON y Using a diagram, illustrate the relationship between CLARiiON physical disk drives and LUNs y Illustrate some of the high availability features of the CLARiiON and how this potentially impacts data availability   ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 2  These are the learning objectives for this training. This course is reinforced by lab exercises using the Navisphere Manager Simulator. Please refer to the supporting materials section to access the labs and download the Simulation tool. CLARiiON Foundations              CLARIION RANGE AND COMPONENTS             ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 3  First, we’ll take a look at the range and components of CLARiiON. CLARiiON Timeline                FC4500  2000    FC4700  2001    CX200  CX400  CX600  2002     CX300  CX500  CX700  2003     CX300i  CX500i  2005     CX3-20  Cx3-40  Cx3-80  2006      ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 4  The CLARiiON storage array is the EMC mid-tier storage offering, which delivers enterprise quality features and functionality. This does not mean that the CLARiiON range cannot meet with the most demanding of storage environments, but rather the CLARiiON range will meet with a different set of criteria than the traditional "high end" storage solution. The CX3 series is the latest generation of the CLARiiON storage systems. Each generation has added enhancements to performance, availability, and scalability, while the high level architecture has remained constant. This allows interoperability between generations and helps preserve the customer’s investment in hardware, software, knowledge, and skills. CLARiiON was the first full Fibre Channel storage system and the industries first 4Gb end to end storage solution. High-End Storage: The New Definition   High-End Then y Simple redundancy – Automated Fail-over y Benchmark performance (IOPs and MB/s)  – Single and/or simple workloads y Basic local and remote data replication  – Backup windows, testing, and disaster recovery y Scalability – Capacity y Manage the storage system  – Easy configuration, simple operation, minimal tuning   ©  2006 EMC Corporation. All rights reserved. High-End Today y Non-disruptive everything – Upgrades, operation, and service y Predictable performance… unpredictable world – Complex, dynamic workloads y Replicate any amount, any time, anywhere – Replicate any amount data, across any distance, without impact to service levels y Flexibility  – Capacity, performance, multi protocol connectivity, workloads, etc. y Manage service levels  – Centralized management of the storage environmCeLARniiOtN Foundations -  5    The decision of which storage product to purchase should always be driven by the application requirements. EMC offers multiple products that meet varied requirements for performance, availability, and functionality. The requirement for economics is the same for both mid-tier and high-end. Performance: Mid-tier storage architectures are ideally suited for applications with predictable workloads that need fast response times and high sustained throughput. Mid-tier architectures can also be manually tuned and balanced to optimize performance for the applications they support. Availability: For applications that require 99.999% availability, mid-tier architectures are designed with redundant components, support (multiple levels of) RAID, and should have the capability to continuously monitor the entire system looking for health status and latent errors. Functionality: Mid-tier storage should also have functional capabilities to make point-in-time local replicas and do remote replication for business continuity. The management tools should be intuitive and seamlessly plug into other management frameworks. Economics: Lowest possible acquisition price, affordable upgrades, and investment protection. Flexible, High Availability Design   y Fully redundant architecture –  Power, cooling, data paths, SPS –  Non-stop operation –  Online software upgrades –  Online hardware changes y Continuous diagnostics –  Data and system integrity –  CLARalert Phone Home y Dual I/O paths with non- disruptive failover y Leader in data integrity –  Mirrored write cache –  Destage write cache to DISK upon power failure –  SNiiFF Verify –  Background Verify-Per RAID Group   y SnapView and MirrorView replication software y SAN Copy y No single points of failure, modular architecture y Fibre Channel, SATA, and ATA disk drives y From 5 to 480 disks y Flexibility –  Individual Disk –  RAID levels 0, 1, 1/0, 3, 5 –  Mix drive types –  Mix RAID levels y Up to 16 GB of memory –  8 GB per Storage Processor –  Configurable read and write cache size   ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 6  Taking a high-level look at the architecture you will see why CLARiiON is so successful. Today, there are over 165,000 that have been sold worldwide. Superior Availability: CLARiiON's design has no single point of failure—all the way down to the fans and power cords. It’s designed for non-stop operation—software upgrades and hardware reconfigurations can be done while online. It's always checking the integrity of its hardware and software. Potential problems are escalated proactively to EMC through its ClarAlert phone home capability. Scalability: A single CLARiiON can go from a complete 3/4 TB storage system in about 30 centimeters of rack space, all the way up to a 480 drives configuration using Fibre Channel, ATA or SATA spindles).  Performance:  CLARiiON was the first full Fibre Channel storage system. That leadership translates today into superior performance. CLARiiON CX3 series provide the industries first 4 GB end to end storage solution.  Flexibility:  Customers can mix and match performance, capacity, cost, and data protection schemes, all in the same unit. Unlike storage systems that come from server vendors, CLARiiON is unique at supporting multiple operating systems and multi protocols while offering full interoperability between different servers, all at the same time. The ability to choose between Fibre Channel and SATA drives is only one example of CLARiiON’s flexibility.  Replication: CLARiiON also supports storage-based local and remote data replication for backup and disaster recovery through its SnapView and MirrorView software. CLARiiON CX3 Models 20, 40, and 80 y  Latest generation, full Fibre Channel networked storage running FLARE Operating Environment y  Flexible connectivity and bandwidth –   Up to 8 FC-AL or FC-SW host ports –   1, 2 or 4 Gb Fibre Channel host connections y  Scalable processing power –   Dual or Quad-processors supporting advanced storage- based functionality y  Industry-leading performance and availability –   2 GB, 4 GB, or 8 GB memory per SP and dual or quad redundant 2/4 Gb back-end storage connections y  Cross-generational software support y  Non-disruptive hardware replacement and software upgrades    ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 7  The market for mid range storage is growing year to year and CLARiiON is leading the way! With the CLARiiON CX3 series supporting both FC and iSCSI protocols, CLARiiON will continue as a major player in the mid tier storage space. EMC CLARiiON series offer entry level arrays using ATA or SATA disk drives to Enterprise arrays that scale to 480 drives. Modular Building Blocks y DAE3P Disk Enclosure – Supports CX3 platform –  Supports up to 15 low-profile, 2/4Gb Fibre Channel (FC), ATA or SATA disk drives –  Can operate at either 4Gb or 2Gb speeds –  Drive fillers must be installed in empty slots y Uses same Chassis and Power Supply/Cooling module as DAE2P Enclosure y Uses same cables as DAE2P Enclosure –  8 Meter maximum between DAE3Ps –  5 Meter maximum between SPEs and DAEs y Two 4G Link Control Cards (LCCs)     ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 8  The CX3 Series platform supports the DAE3P Disk Array Enclosure. The DAE3P can house up to 15 low-profile, 2/4Gb Fibre Channel (FC), Advanced Technology Attached (ATA), or Serial Advanced Technology Attached (SATA) disk drives, and can operate at either 2Gb or 4Gb speeds. The DAE3P uses the same 4 Chassis, Power Supply/Cooling module as the older DAE2P Enclosures used with the CX series. The same cables used for the DAE2P support the DAE3P. DAE3P uses 4Gb LCCs. LCCs connect the DAE3Ps together in the same cabinet. CLARiiON SATA (Serial Advanced Technology Attachment)   y  Lower $/MB for backup or bulk storage y  Alternative to Fibre Channel HDAs y  Same Software capability as FC y  Uses FC interconnect –  Mix FC, ATA, and SATA enclosures –  First shelf must be Fibre Channel   y Full HA features – Dual-ported access – Redundant power and LCCs – Hot swap capability                  ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 9  SATA disks are low-cost disk drives typically found in desktop PCs  SATA disk drives are ideal for applications requiring high capacity and lower performance levels than FC drives. IDE/SATA drives represent the vast majority of all drives manufactured worldwide, and therefore can be offered at a lower price point. They are also significantly slower than Fibre Channel disks in both rotational and seek speeds. CLARiiON supports mixing and matching of SATA DAEs and Fibre Channel DAEs in the same storage system. ( Note different drive types cannot be mixed within the same DAE3P) All CLARiiON software and functionality is supported with SATA drives. CLARiiON with SATA is targeted for backup applications with backup-to-disk functionality. CLARiiON with SATA also lowers the hardware implementation cost for SnapView BCVs (Clones) and MirrorView target sites. CLARiiON CX3 Series Architecture 1/2/4Gb Fibre Channel Front End       Storage Processor / Blade   CLARiiON Messaging Interface (CMI) Multi-Lane PCI-Express bridge link     Storage Processor / Blade     4G LCC    2/4 Gb Fibre Channel Back End      4G LCC      4G LCC   4G LCC       4G LCC    4G LCC      4G LCC    4G LCC         ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 10  The CX3 Series module architecture allows the customer to add drives as needed to meet capacity requirements. CLARiiON Architecture is based on intelligent Storage Processors/Blades ( SPs) that manage physical drives on the back end and service host requests on the front end, be it Fibre Channel (FC) or iSCSI protocols. Depending on the model, each Storage Processor or Blade includes either one or two CPUs. Storage Processors communicate to each other over the CLARiiON Messaging Interface (CMI) however, unlike previous CX series arrays, the CX3 Series will be the first CLARiiON platform to use PCI-Express as the high-speed CMI path. PCI Express architecture delivers advance I/O technology to new levels of performance and scalability. 4 Gb Logical link Controllers connect the DAE3Ps in the same cabinet. CLARiiON Foundations           THEORY OF OPERATION               ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 11  Next, we will look at the theory of operations. Storage Processor Introduction   y  Storage processors are configured in pairs for maximum availability y  One or two processors per Storage Processor board y  Two or four Fibre Channel front-end ports for host connectivity –  Small Form-Factor Pluggable (SFP) –  1Gb 2Gb or 4Gb speeds –  Arbitrated loop or switched fabric –  iSCSI y  Dual-ported Fibre Channel Disk drives at the back-end –  Two or Four Arbitrated Loop connections –  2 Gb or 4Gb speeds y  Maximum of 8 GB of memory per SP –  Write Cache is mirrored between Storage Processors for availability using the CMI (CLARiiON Messaging Interface) –  Write Caching accelerates host writes           4 Fibre Channel Ports  Mirrored Cache CPU	CPU FC-AL	FC-AL LCC    LCC            CMI   y  Ethernet connection for management 	Storage Processor   ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 12  The main component in all CLARiiON series arrays is the Storage Processor. SPs are configured in pairs for maximum availability and are Field Replaceable Units (FRUs). SPs provide both front-end connectivity to the hosts and back-end connectivity to the physical disks. Each Storage Processor includes up to 8 GB of memory depending on the model, most of which is used for cache. Cache memory is segmented into read cache and write cache. Read cache memory stages and prefetches read requests from the host. Write cache helps to accelerate host writes to the storage system. With write cache enabled, host writes are mirrored to the write cache memory on the other SP over the CLARiiON Messaging Interface (CMI). Each storage processor also includes a TCP/IP connection used for configuration and managing the storage system. Each storage system ships with a complete copy of FLARE software installed on the first four disks on back-end loop 0. Disks 0_0 and 0_2 store mirrored copies of the software for SP A, and disks 0_1 and 0_3 store mirrored copies of the software for SP B. When you power up the storage system, each SP boots and executes FLARE software. Cache Memory y Cache memory on an SP performs two tasks – Staging: Temporary buffering of current read and write data ¾ Always performed on each I/O – Storage: Repository for frequently accessed data ¾ Maintaining copies of read and write data (uses least-recently used algorithm to manage storage cache) ¾ User must explicitly enable this (for both read and write) y Benefits of caching – Burst Smoothing - Absorb bursts of writes without becoming “disk bound” ¾ Write cache optimization – Locality - Merge several writes to the same area into a single operation ¾ Increases write performance – Immediacy - Satisfy user requests without going to the disks ¾ Read cache optimization prefetching of data for sequential reads    ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 13  The CLARiiON CX3 series arrays use cache in a manner similar to traditional workstations and servers. Application response times are better if application code can buffer its data locally as opposed to having to read from CLARiiON cache. This is a primary design goal of smart applications and is used with both file system and raw partitions. Data is periodically flushed to the storage device. This “lazy write” approach would result in large bursts of large I/Os to the storage systems - a perfect fit with the “burst smoothing” benefit of CLARiiON’s caching. Cache page size is perhaps the most influential parameter on cache performance. Matching the host I/O size to the CLARiiON cache page size will help in aligning I/Os and help overall performance. Under certain I/O loads it may be beneficial to bypass cache and write directly to disk. Mirrored Write Caching     SP-A	SP-B         CMI   y Write cache size is user configurable and is allocated in pages y How much write cache is used by each SP is dynamically adjusted based on workload y All write requests to a given SP are copied to the other SP y data integrity ensured through hardware failure events y CLARiiON Messaging Interface (CMI) used to communicate between SPs      Storage System   ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 14  The amount of cache memory allocated to write operations is dynamically adjusted, based on workload. Navisphere Analyzer is used to baseline actual workloads as a aid to understanding the efficiency of cache utilization. When write cache is enabled, a write operation received by one storage processor is sent to the other storage processor and a confirmation that it was successfully stored is returned before write complete is sent to the host. The mirroring insures data protection if a single SP should fail. The CMI, carries data and status information between SPs over the high speed PCI Express bus. On FC-series arrays, the CMI is a single connection, running at 100 MB/s. On CX-series arrays, the CMI is a dual connection, with each connection running at 200 MB/s. Managing Write Cache y To maintain free space in write cache for I/O bursts, pages are flushed from cache to the drives – Only the least-recently used pages are flushed y Three levels of flushing: – Idle - Low I/Os to the LUN; user I/Os continue – Watermark - Priority depends on cache fullness; user I/Os continue – Forced - Cache has no free space; user I/Os queue y For maximum performance: – Provide a “cushion” of unused cache for I/O bursts – Minimize/avoid forced flushes      ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 15  The flushing algorithm is usually referred to as the Least Recently Used (or “LRU”) algorithm. The LRU approach focuses on write cache immediacy (as the most recent used data remains available), and less on locality (as the location of the data is considered only to produce larger flushes of data). Idle flushing keeps some free space in write cache when I/O activity is relatively low. If data immediacy were most important, idle flushing would be sufficient. If idle flushing cannot maintain free space, though, forced flushes create space for new I/Os. Forced flushes maintain the desired free space, though it will impact overall performance as all read and write operations are halted to clear space in the write cache. Write Cache Protected by “Vault” y The “vault” is a reserved area found on specific protected disks y At the first sign of an event which could potentially compromise the integrity of the data in write cache, cache data is dumped to the vault area y After the data is dumped to the vault, it will be migrated to the LUNs where it belongs – When power is restored data is migrated from the vault back to cache (if an SPS has a charged battery) y Other failures such as a SP or Cache failure will disable write cache    ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 16  The vault is a reserved area found on the first 9 disks of the DPE on the FC series and the first 5 drives on a CX series system. For this reason, only the DPE needs to be kept alive in the event of a power failure. At the first sign of an event which could potentially compromise the integrity of the write cache data, that data is dumped to the vault area. It is protected there by the non- volatile nature of disk storage, as well as the RAID level (RAID-3) implemented for the vault. After the data is dumped to the vault, it will be migrated to the LUNs where it belongs. Until all data has been migrated, I/Os check the vault because the most recent copy of data may be located there. Once the migration process is complete, the vault is marked as empty, and I/Os need no longer check it. Persistent Storage Manager (PSM) y PSM is a hidden LUN that records configuration information y Both SPs access a single PSM so environmental records are in sync – If one SP receives new configuration info, that info is written to the PSM and the other SP instantaneously updates itself y If one SP needs to be replaced, the new one can easily find the unique environmental information on the PSM y Enables SPs to be completely field-replaceable       ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 17  The PSM is created during initialization of the array. The PSM resides on the first loop of drives. The default size is 1GB on CX systems, 512 MB on the FC4700. The table below details the information contained in the PSM. Host Information	Array Information	Other Drive Mapping	SP IP Addresses	Polling Rates Privileged Users (Host)	Privileged Users (Array)	Core OS & FW Revisions Host Information	ALPAs (Arbitrated Loop Physical Address)	 Initiator ID information	Pre-upgrade Packages (Snap, Core OS, etc.)	 	Pre-upgrade Packages (Snap, Core OS, etc.)	      Data Units on a Disk  User Dat a (5 12 Bytes )     Se c   Sec   Se c   Se c   Se c   Se c   Se c   Se c   Sec   Se c   Se c   Se c   e.0   e.1   e.2   e.3   e.4   e.5   e.122 e.123 e.124 e.125 e.126 e.127      Eleme nt s.0	Eleme nt s.1	Eleme nt s.2	Parity s. Eleme nt s.3	Eleme nt s.4	Eleme nt s.5   y Sector – 520 Bytes ¾ 512 bytes of user data ¾ 8 bytes of administrative data y Element – Element size is the number of blocks written to single disk in a RAID group before moving to the next disk y Stripe – A group of sequential elements starting with the first disk in a RAID group and ending with the last, including any parity or mirrored elements. © 2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 18  CLARiiON disks sectors contain 520 bytes each. 512 bytes are dedicated to host data, while the other 8 bytes are used for housekeeping purposes. One or more sectors make up an element size. The element size is the largest amount of data written to a disk before moving to the next disk. Elements make up a stripe size. A RAID 5 consisting of 4 data disks and 1 parity disk, use an element size of 64KB and a stripe size of 256KB. Element Tips y Element size is defined for each LUN – Large elements are more efficient for reading from disk ¾ Good for sequential data – Smaller elements scatter the data across the RAID group ¾ Good for random data – Element size cannot be changed without unbinding the LUN ¾ Unbinding a LUN destroys the data – Default element size of 64K is optimal for most applications y Read and write operations that cross an element boundary degrade performance – Element size set as a multiple of the average unit of data requested ¾ For 2k I/O size - element should be 2k, 4k, 8k,…etc – Monitor LUNs using Navisphere Analyzer  ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 19  The element size is configurable on a per LUN basis and must be set up when the LUN is created. To a large degree, the element size will determine the performance of the LUN in the environment. For RAID level 5, where the disks can operate independently, small element sizes are most efficient when I/Os are random. The data is then scattered across all the disks, and the chances are good that multiple disks can be used simultaneously to fetch data. In the case of data which is largely sequential, large element sizes are written to contiguous areas of the disk and therefore result in little or no head movement. This increases performance – one of the major factors that influences performance is the seek time of the disks. If the average I/O is not an even multiple of the element size, each read or write operation will involve 2 or more disks, reducing performance, particularly write performance. Ideally, the element size should always be an even multiple of the average I/O size. Navisphere Analyzer tracks these events, and calls them “disk crossings” or “stripe crossings”. If the percentage of disk crossings is high for a LUN, it may be an indication that the element size needs to be increased. Bear in mind that this process is destructive. These parameters should only be changed by experienced and trained individuals, if advised to do so by CLARiiON Performance Engineering. CLARiiON RAID Options y Disk – No protection – JBOD y RAID-0: Stripe – No protection – Performance JBOD y RAID-1: Mirroring       Step 0 - Planning   – Some performance gain by splitting read operations – Protection against single disk failure – Minimum performance hit during failure      ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 20  RAID-0 spreads the data across the drives in the array, one segment at a time. Characteristics include data spreading across the drives in the array by system block size. data is striped across all of the drives in the RAID Group without generating any redundant data. A RAID-0 Group consists of three to sixteen disk modules. The number of drives and groups is scaleable. In RAID-0, there is no error recovery over or above what is normally on a disk drive. Advantages include high I/O rate (small block size) and transparent to host system software. Disadvantages include a somewhat slower transfer rate than RAID 3 and no parity disk or redundant disk. RAID Level 1 is data mirroring. Any time data is written to the LUN, a write occurs to a mirror disk. This is done transparently to the attached host. Characteristics include a traditional approach for improving reliability of disk storage as this technology is well understood. It is an expensive option because all disk drives are duplicated. Every write to a data disk is also a write to the mirror disk. A RAID-1 group consists of 2 disk modules. data on a replaced drive is rebuilt from the mirror drive. In terms of Error Recovery, in RAID Level 1, if a disk fails, the disk array controller uses the mirror drive for data recovery and continuous operation. Advantages include high data availability and high I/O rate (small block size). Disadvantages include a total number of disks in the array equaling 2 times the data (useable) disks. The overhead cost equals 100%, while usable storage capacity is 50%. Applications that write sequentially in small I/Os like Microsoft Exchange; transaction logs are a good choice for RAID-1 disks. CLARiiON RAID Options y RAID-1/0: Striped Mirrors      Step 0 - Planning   – Performance of stripes combined with split read operations – Protection against single disk failure – Minimum performance hit in failure mode y RAID-3: Striped Elements – Each data element striped across disks - parity kept on the last disk in the RAID group – Extremely fast read access from the disk – Used for streaming media – Parity protection against single disk failure – Performance penalty during failure     ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 21  RAID Level 1/0 is essentially a striped RAID-1 group. It combines the speed advantage of RAID-0 with the redundancy advantage of RAID-1. This is a more expensive option compared to other parity protection schemes, because all disk drives are duplicated. Every write to a data disk is also a write to the mirror disk. RAID-1/0 Groups consists of 4, 6, 8, 12, 14 or 16 disk modules. In RAID Level 1/0, if a disk fails, the disk array controller uses either its mirror drive or the data drive for data recovery and continuous operation. If there is a hot spare, data is rebuilt onto the hot spare from the mirrored copy. When the failed disk is replaced, data from the other mirror (copy) is used to rebuild the replaced disk. Advantages include high data availability, high I/O rate (small block size), and the ability to withstand multiple drive failures as long as they occur on different mirrors. Disadvantages include that the total number of disks equals two times the data disks, with overhead cost equaling 100%. RAID-1/0 is a good solution for sites that normally have RAID-1 but need the added performance boost that RAID-1/0 provides. RAID Level 3 is an array of disk drives transferring data in parallel. RAID-3 combines 5 or 9 disk drives to act as a large virtual drive with high throughput (two or more times that of a conventional drive). A parity drive is added to the Group to store parity information. Information on the parity drive allows data to be reconstructed, should a drive fail. Parity is the Exclusive_OR (XOR) of data on drives 1, 2, 3, and 4. Parallel data paths are provided to the drives. With Error Recovery, if a disk fails, the controller dynamically recovers the drive's data, using data on the other data drives and the parity drive. The controller uses data from the good drives in the normal positions within the data block. The controller combines the first byte of each sector in the good drives and the parity drive to regenerate the first byte of the sector on the failed drive. This process is repeated to generate each byte of the failed drive. Advantages include total number of disks equals 1.25 times the data disks for group of 5, overhead cost is 20% for group of 5, and a high bandwidth (large block size). There is good throughput on large data transfer. Disadvantages include poor media efficiency of small blocks. RAID Level 3 is not good for transaction processing systems. data is lost if multiple drives fail within the same RAID-3 Group. RAID- 3 is a good choice for applications that require moderate to high I/O that is sequential, such as MirrorView targets, Clones, and on-line backups. CLARiiON RAID Options y RAID-5: Striping with Parity – Performance of striping  – Protection from single disk failure       Step 0 - Planning    – Parity distributed across member drives within the RAID Group  – Write performance penalty  – Performance impact if a disk fails in RAID Group  y Hot Spare  – Takes the place of failed disk within a RAID group  – Must have equal or greater capacity than the disk it replaces  – Can be located anywhere except on Vault disks  – When failing disk is replaced, the hot spare restores the data to the replacement disk and returns to the hot spare pool  ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 22  RAID-5 is an independent disk array, meaning it does not read and write data to parallel disks like RAID-3, but it performs independent or parallel read and write operations. There is no dedicated parity drive. The result is a transfer rate equal to that of a single drive, but with a high overall I/O rate. Characteristics: Independent data paths to the drives. A RAID-5 Group usually consists of 5 disk modules. The number of drives and groups is scaleable. data is striped by system block size and distributes the data and parity information per sector across all disks (no parity disk). In RAID Level 5, if a disk fails, the controller rebuilds its data from the data and parity from the remaining drives. For example, to rebuild sector 0 on drive 5, the controller uses the parity information from drive 1 and data block information from drives 2, 3, and 4. The controller combines the information of each sector in the remaining drives and the parity information to regenerate the first block of the sector on the failed drive. The process is repeated to generate each block of the failed drive. Advantages include it being good for parallel processing (multi-tasking) applications or environments. Also, the total number of disks equals 1.25 times the number of data disks for a group of five, or 1.1 times the number of data disks for a group of 10, while the overhead cost is 20% for groups of 5 and 10% for groups of 10. It has a high I/O rate (small block size). Disadvantages include somewhat slower transfer rate than RAID-3. There is degradation in performance in recovery and reconstruction modes and data loss if multiple drives within the same group are lost. RAID-5 offers great price/performance. Which RAID Level Is Right y RAID 0 – data striping – No parity protection, least-expensive storage     Step 0 - Planning   – Applications using read-only data that require quick access, such as data down-loading y RAID 1 – Mirroring between two disks – Excellent availability, but expensive storage – Transaction, logging or record keeping applications y RAID 1/0 – data striping with mirroring – Excellent availability, but expensive storage – Provides the best balance of performance and availability y RAID 3 – data striping with dedicated parity disk y RAID 5 – data striping/parity spread across all drives – 	Very good availability and inexpensive storage y Support mixed types of RAID in the same chassis  ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 23  Use a RAID 0 group where the best overall performance is important. A RAID 0 group is less available than an individual unit. A RAID 0 group is useful for applications using short-term data to which you need quick access. Use a RAID 1 group for logging or record keeping applications because it requires fewer modules than a RAID 0 group and provides high availability and fast write access. Or you could use it to store daily updates to a database that resides on a RAID 5 group, and then, during off- peak hours, copy the updates to the database on the RAID 5 group. A RAID 1/0 group provides the best balance of performance and availability. You can use it very effectively for any of the RAID 3 or RAID 5 applications. Use a RAID 3 group with a single-task application that uses large I/O transfers (more the 64 Kb). Use a RAID 5 group for a database repository or database server that uses a normal or less-than- normal percentage of write operations (writes are 33% or less of all I/O operations). Use a RAID 5 group where multi-tasking applications perform I/O transfers of different sizes. Write caching can significantly enhance the write performance of a RAID 5 group. Storage Configuration and Provisioning   y Understanding application and server requirements and planning configuration is critical! y RAID Group is a collection of physical disks – RAID Protection level is assigned to all disks within the RAID group y Binding LUNs is the creation of Logical Units from space within a RAID Group y Storage groups are collections of LUNs that a host or group of hosts     Step 0 - Planning   Step 1 – Create RAID Groups   Step 2 – Bind LUNs   Step 3 – Create Storage Groups   Step 4 – Add LUNs to Storage Groups   have access to 	Step 5 – Connect Hosts with Storage Groups  ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 24  The first step in configuring a CLARiiON Storage System is to understand application and server requirements as they pertain to availability, capacity, access control, and other features such as clustering and local and remote replication. While this step is only mentioned here, it will take more time and effort than all the other steps combined. If this step is skipped, reconfiguration will most likely be required. In Step 0, you will have identified the availability and performance goals and, from that, should have determined the appropriate RAID level. A RAID Group is a collection of physical disk that share the same RAID protection scheme or RAID level. The actual protection scheme does not get set until you “bind” the first LUN. In step 0, you will also have determined the number and size of Logical Units (LUNs) that are required. Each RAID group can contain one or more LUNs. Access control is implemented in the CLARiiON using the Access Logix code that runs in the Storage Processor. Access Logix uses the concept of Storage Groups. LUNs are assigned to a Storage Group and hosts are associated with the Storage Group to allow access. Again, planning is critical here as well. In Step 0, you should have identified which hosts need access to which LUNs and any special requirements such as shared access that would be required if the hosts where part of a cluster. Creating RAID Groups y RAID protection levels are set through a RAID group      Step 1 – Create RAID Groups   y Physical disks part of one RAID group only  – Drive types cannot be mixed in the RAID Group y May include disks from any enclosure y RAID types may be mixed in an array y RAID groups may be expanded y Users do not access RAID groups directly 5 disk RAID-5 group 	4 disk RAID-1/0 group        ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 25  The physical disks on the CLARiiON are configured into RAID groups. Any RAID Group should consist of all FC disks or all SATA disks, but NOT a mix of both FC and SATA. Binding a LUN y Binding is the process of building LUNs onto RAID Groups       Step 2 – Bind LUNs    y May be up to 128 LUNs in a RAID Group y May be up to 2048 LUNs per CLARiiON array y LUNs are assigned to one SP at a time – The SP owns the LUN – The SP manages the RAID protection of the LUN – The SP manages access to the LUN y The LUN uses part of each disk in the RAID Group – Same sectors on each disk    ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 26  A logical unit (LUN) is a grouping of one or more disks into one span of disk storage space. A LUN looks like an individual disk to the server’s OS. It has a RAID type and properties that define it. The process of building a LUN onto a RAID Group is called binding. When you bind a LUN on a RAID Group, you specify how much of the Group’s user space (contiguous free space) you want the LUN to use. The LUN is distributed equally across all the disks in the RAID Group. The characteristics of a LUN are: y 128 LUNs per RAID Group is the maximum for all RAID types y Up to 2048 LUNs per array depending on the model y An SP owns a LUN, and different SPs can own different LUNs in the same RAID Group y All LUNs in the RAID Group have the same RAID type y Each LUN in a RAID Group can have a different element size where applicable      Bind Operation - Setting Parameters     y Fixed Bind parameters   Step 2 – Bind LUNs   – Disk numbers, RAID type, LUN #, element size – Can’t be changed without unbinding and rebinding y Variable parameters – Cache enable, rebuild time, verify time, auto assignment – Can change without unbinding y Bind Operation – Fastbind is the almost instantaneous bind achieved on a factory system or implemented in the latest code       ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 27  Once bound a LUNs characteristics can be viewed from the properties screen. From there, you may change the variable parameters such as; cache enable, rebuild time, verify time, and auto assign without unbinding the LUN. Should any of the fixed parameters need to be changed, such as Disk numbers, RAID type, LUN #, or element size, the LUN must be first unbound and then bound again, which will cause a loss of data. Fast Bind allows the user to immediately access the LU upon issuing a bind command to the array, while having the initialization of the LU occur in parallel with user IO traffic. Fast Bind is automatically used with all newly bound LU’s of all RAID types except Hot Spares. Storage Groups y Storage Groups are a feature of Access Logix and used to implement LUN Masking      Step 3 – Create Storage Groups   – Storage Groups define the LUNs each host can access – A Storage Group contains a subset of LUNs grouped for access by one or more hosts and inaccessible to other hosts ¾ Without Storage Groups, all host can access all LUNs – Storage groups can be viewed as a “bucket” that has dedicated and/or shared LUNs accessible by a server or servers – Access Logix controls which hosts have access to a storage group – Host access the array and provide information through the Initiator Registration Records process y Storage Group planning is required   ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 28  Storage groups are another key feature. Without Storage Groups, all hosts can access all LUNs. These Storage Groups provide data Access Control by defining which LUNs each host can access. Storage Groups are a combination of LUNs assigned to a particular host. An initiator can only be connected to a single Storage Group on any storage system. The exception would be in a clustered environment where 2 or more hosts share the same storage group. There is also a new LUN mapping table applied to the LUN numbers in Standard SCSI commands. Each entry includes the LUN alias, or the number presented to the host and the FLARE LUN number which is created during binding. These pieces are key to Access Logix and shared environments. Example: Designing Storage Groups Step 3 – Create Storage Groups       Storage Group 1   Storage Group 2   Storage Group 3             y Storage Groups are logical grouping of LUNs for access control y One or more hosts may be connected to a Storage Group to allow access to the LUNs ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 29  Here is a picture of what we’re talking about with storage groups and how they can be set up. Unless all hosts require access to the same LUNs, it is easier to create a storage group per server. It will allow the flexibility of adding and removing LUNs on the fly. In the graphic above, Storage Group 2 is being accessed by two servers. While this is an acceptable configuration, it is usually only recommended when the servers are in a clustered environment. An example of this is when a NAS head is front-ending a CLARiiON array. Additionally, the LUNs in each of the Storage Groups need not be unique. For example, LUNs 1, 2, and 3 could be present in both SG 1 and in SG 3. Care should be exercised when doing so since access to the same volumes, at the same time, may corrupt the data. metaLUNs y A metaLUN is created by combining LUNs – Dynamically increase LUN capacity – Can be done on-line while host I/O is in progress – A LUN can be expanded to create a metaLUN and a metaLUN can be further expanded by adding additional LUNs – Striped or concatenated ¾ data is restriped when a striped metaLUN is created y Appears to host as a single LUN – Added to storage group like any other LUN – Can be used with MirrorView, SnapView, or SAN Copy y Supported only on CX family with Navisphere 6.5+ metaLUN  + 	+ 	=    ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 30  MetaLUNs are supported only on CX-Series storage systems. The metaLUN feature lets you dynamically expand the capacity of a single LUN (base LUN) into a larger unit called a metaLUN. You do this by adding LUNs to the base LUN. You can also add LUNs to a metaLUN to further increase its capacity. Like a LUN, a metaLUN can belong to a Storage Group, and can participate in SnapView, MirrorView and SAN Copy sessions. During the expansion process, the host is able to process I/O to the LUN or metaLUN and access any existing data on the Base LUN. It does not, however, have access to any added capacity until the expansion is complete. Depending on the operating system, a reboot of the host or the use of a disk expansion utility, such as diskpar in Windows, may be used to claim the additional space. Each set of striped LUNs is called a component. All metaLUNs contain at least one component which includes the base LUN and one or more LUNs. Any data that gets written to a metaLUN component is striped across all the LUNs in the component. You can expand a LUN or metaLUN in two ways — stripe expansion or concatenate expansion. A stripe expansion takes the existing data on the LUN or metaLUN, and restripes (redistributes) it across the existing LUNs and the new LUNs you are adding. The stripe expansion may take a long time to complete and will affect performance while the expansion is in process. A concatenate expansion creates a new metaLUN component that includes the new LUNs and appends this component to the end of the existing LUN or metaLUN. There is no restriping of data between the original storage and the new LUNs. The concatenate operation completes immediately. Advanced Availability: LUN Ownership Model y Only one Storage Processor “owns” a LUN at any point in time – Assigned when LUN is created but can also be changed using Navisphere Manager or CLI y If the Storage Processor fails, ownership of the LUN can be moved to the surviving SP – Process is called LUN Trespassing – CLARiiON EMC PowerPath provides this function y For maximum availability, careful design of I/O path for no Single-Point-Of-Failure is required      ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 31  CLARiiON uses the single ownership model. At any time, only 1 SP owns a LUN. Ownership may be changed with Navisphere Manager or the Navisphere CLI. When optional failover software is available on the host, such as PowerPath, temporary changes are automatically made when certain failures occur. An SP failure causes all LUNs to be failed over to the surviving SP. Host Connectivity Redundancy PowerPath – Failover Software y Host resident program for   automatic detection and management of failed paths y Host will typically be configured with multiple paths to LUN y If HBA, cable or Switch fails, PowerPath will redirect I/O over surviving path y If Storage Processor fails, PowerPath will “Trespass” LUN to    Application  PowerPath  Request 	Request sd HBA   surviving Storage Processor and redirect I/O y Dynamic load balancing across HBA and Fabric – Not Storage   FC Switch   FC Switch   Processors   0	1 SP A   0	1 SP B       ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 32  To support the redundancy of the Hardware and Software of the CLARiiON arrays, host connectivity is also protected with PowerPath. PowerPath provides automatic load balancing of Host I/O to the array and is customer configurable. The CLARiiON Trespass policy will enable the PowerPath software to interact with the Access Logix software on the array to change to the surviving path, and temporarily change the SP ownership of a LUN in the event of a SP failure. Administrators can improve the server’s ability to manage heavy storage loads through continuous and intelligent I/O load balancing. CLARiiON Foundations              SOFTWARE AND MANAGEMENT ENVIRONMENT            ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 33  This section will look at the management environment of CLARiiON. FLARE Operating Environment         EMC ControlCenter CLARiiON Based Applications   Navisphere    FLARE Operating Environment    CLARiiON Hardware   y FLARE Operating Environment is the “Base Software“ that runs in the CLARiiON Storage Processor –  I/O handling, RAID algorithms –  End-to-end data protection –  Cache implementation y Access Logix provides LUN masking that allows sharing of storage system y Navisphere middleware provides common interface for managing CLARiiON y CLARiiON replication software including –  MirrorView, SnapView, SAN Copy y EMC ControlCenter provides end-to- end management of a CLARiiON       ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 34  Let’s briefly introduce the software components, as most functionality is based in software and supported by the hardware. FLARE software manages all functions of the CLARiiON storage system. Each storage system ships with a complete copy of FLARE software installed. When you power up the storage system, each SP boots and executes FLARE software. y FLARE performs provisioning and resource allocation y Memory budgets for caching and for snap sessions, mirrors, clones, copies y Process Scheduling y Boot Management Access Logix software is optional software that runs within the FLARE operating environment on each storage processor (SP). Access Logix provides access control and allows multiple hosts to share the storage system. This “LUN Masking” functionality is implemented using Storage Groups. A Storage Group is one or more LUNs within a storage system that are reserved for one or more hosts and are inaccessible to other hosts. When you power up the storage system, each SP boots and executes its Access Logix software. Navisphere Management software is a suite of tools that allows centralized management of CLARiiON storage systems. Navisphere provides a centralized tool to monitor, configure, and analyze performance. CLARiiON can also be managed as part of EMC ControlCenter, allowing full end-to-end management. CLARiiON Management Options y There are two CLARiiON management interfaces – CLI (Command Line Interface) ¾ naviseccli commands can be entered from the command line and can perform all management functions – GUI (Graphical User Interface) ¾ Navisphere Manager is the graphical interface for all management functions to the CLARiiON array               ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 35  Navisphere Secure CLI allows a command line interface for storage management. Provided the host has the agent file configured, a user can perform all the functions needed to manage the array. Navisphere GUI is an intuitive, easy to use graphical user interface for all management functions on the array. EMC Navisphere Management Software y Centralized management for CLARiiON storage throughout the enterprise – Centralized management means more effective staff y Allows user to quickly adapt to business changes y Keeps business-critical applications available y Key features – Java based interface has familiar look and feel   – Multiple server support – EMC ControlCenter integration – Management framework integration   y Navisphere Software Suite – Navisphere Manager – Navisphere Analyzer – Navisphere CLI    ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 36  Navisphere Management software is a suite of tools that allows centralized management of CLARiiON storage systems. Navisphere provides a centralized tool to monitor, configure, and analyze performance. The Navisphere Suite includes: ƒ Navisphere Manager: Allows graphical user interface (GUI) management and configuration of a single or multiple system, and is also the center for management and configuration of system-based access and protection software including Access Logix, SnapView, and MirrorView applications ƒ Analyzer: A performance analysis tool for CLARiiON hardware components ƒ Agent: Provides the management communication path to the system. Enables CLI access   Navisphere can be launched from EMC ControlCenter. In the event that you already have a Symmetrix and ControlCenter, ControlCenter will allow you to manage all your information from one central location. Navisphere stores initiator and host information on the array. It is used to create the relationships used for access control, as well as information presented in Navisphere. It also can automatically set any unique features required to support an operating system, enabling it to support a heterogeneous environment on a single port. Navisphere Manager y Discover – Discovers all managed CLARiiON systems y Monitor –  Show status of storage systems, Storage Processors, disks, snapshots, remote mirrors, and other components –  Centralized alerting y Apply and provision –  Configure volumes and assign storage to hosts –  Configure snapshots and remote mirrors –  Set system parameters –  Customize views via Navisphere Organizer y Report –  Provide extensive performance statistics via Navisphere Analyzer  ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 37  Navisphere Manager provides speed and flexibility using the familiar Microsoft windows interface. It lowers cost of management/ownership, including training and administrative costs, and has a proactive focus on addressing potential problems. It reduces personnel requirements by its ability to increase the productivity of staff by managing larger amounts of storage with fewer resources. Creating a RAID Group- GUI Step 1 – Create RAID Groups                       ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 38  Before a LUN can be created, you must create the RAID group on which you will bind the LUN. A RAID Group is one or more Physical Drives that share the same RAID type. Each RAID Group has a unique number to identify it. By default, the next available RAID Group ID is automatically assigned, however, you have the capability of selecting from the list of available identifiers. The actual physical drives that make up the RAID group can be assigned automatically by simply specifying the number of disks. The next available drives in sequence will be used or, if you select Manual, you will be allowed to choose the disks that make up the RAID Group. Under RAID Group Parameters, in the Expansion/Defragmentation Priority list, click the priority for the new RAID Group. Select the Automatically Destroy check box to enable automatic dissolution of the RAID Group when the last LUN is unbound, or clear the check box to disable automatic dissolution. When you create a RAID Group and specify the number of drives, the supported RAID type will be displayed. However, the actual RAID type for a RAID Group is not set until the first LUN is created within the group. Binding a LUN  Step 2 – Bind LUNs                        ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 39  This shows how a LUN is bound in a RAID Group. The following steps are used to bind a LUN: y For each LUN you create, select the type of RAID protection you would like from the RAID Type box y For the RAID Group, select a number for the LUN y Under LUN properties, select a LUN ID, Element Size, Rebuild Priority, and Verify Priority y Enable the Read or Write Cache by checking the appropriate box y Specify the LUN Size, Number of LUNs to Bind, and which SP will own the LUN (Default Owner)      LUN Properties - General   Step 2 – Bind LUNs                        ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 40  This shows the LUN Properties screen with the General tab selected. The characteristics of the LUN are shown. To change the Unique ID, Element, or RAID Type, you must unbind and rebind the LUN with the new characteristics. Unbinding is disruptive and will result in all data being lost on that LUN. The Rebuild Priority, Verify Priority, and Auto Assign can be changed on this screen without affecting the data stored on the LUN. The default owner is the SP that assumes ownership of a LUN after power is turned off, then on again. You may want to transfer ownership of a LUN to balance LUNs between SPs. Ownership is changed on this screen by selecting SP A or SP B under Default Owner. Creating a Storage Group  Step 3 – Create Storage Groups                      ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 41  This is an example of creating a Storage Group in Navisphere Manager. When a Storage Group is created, a unique name is assigned. The Storage Group Properties – General Tab will display this name, as well as the Storage Group Fibre Channel World Wide Name that is used internally to control access to LUNs within the Storage Group. Storage Group Properties - LUNs   Step 4 – Add LUNs to Storage Groups                      ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 42  The Storage Group Properties – LUNs tab in Navisphere Manager is used to assign LUNS to a storage group. Hosts that are “connected” to a Storage Group will have full read-write access to all LUNs that are selected as part of the Storage Group. Storage Group Properties - Hosts Step 5 – Connect Hosts to Storage Groups                       ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 43  The Storage Group Properties – Host tab in Navisphere Manager is used to connect hosts to a storage group. When a host is “connected” to a Storage Group, it will have full read-write access to all LUNs in the group. Non-connected hosts will have no access to these LUNs. Course Summary Key points covered in this course: y The basic architecture of a CLARiiON Disk Array y The architectures of the various CLARiiON models y The data protection options available on CLARiiON y The relationship between CLARiiON physical disk drives and LUNs y High availability features of the CLARiiON and how this potentially impacts data availability       ©  2006 EMC Corporation. All rights reserved. CLARiiON Foundations - 44  These are the key points covered in this training. Please take a moment to review them. This concludes the training. Please proceed to the Course Completion slide to take the assessment.