1 / 20

Changing the Economics of Storage: Flash at the Price of Disk

Changing the Economics of Storage: Flash at the Price of Disk. Neeraj Matiyani Director Enterprise Storage Solutions. Data centers need to handle many types of diverse application workloads. Read-intensive. Write-intensive. Random. Sequential. Indexing Front End Web OLTP Database

lenci
Download Presentation

Changing the Economics of Storage: Flash at the Price of Disk

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Changing the Economics of Storage: Flash at the Price of Disk Neeraj Matiyani Director Enterprise Storage Solutions

  2. Data centers need to handle many types of diverse application workloads Read-intensive Write-intensive Random Sequential Indexing Front End Web OLTP Database Email/Messaging Data Warehousing Virtual Desktop HPC Streaming Media Web Applications Collaboration Infrastructure Other • Applications are getting more dependent on higher I/O performance • Traditional storage systems are optimized for the middle of the spectrum • Data stored, backed up and archived is growing exponentially • SLAs are becoming more rigorous • IT budgets are often staying flat

  3. What characterizes your workloads?Traditional “one size fits all” data storage strategy is inefficient Performance • Hot data that needs to be retrieved extremely fast without latency • IOPS vs. throughput Access patterns and frequency • Data accessed frequently vs. cold, rarely accessed data • Heavily read vs heavily written data • Data optimized for tiering vs caching I/O patterns • Read and write percentages • Transfer sizes • Sustained vs. bursty I/O patterns Business value • Data that is highly valuable to the business and can be used for deriving business intelligence, financial analysis, etc. Each type of data needs to be treated differently to align the application performance with capacity and cost requirements

  4. Reliance on I/O-intensive solutions for critical applications is growing Hot/ Active Data I/O intensive solutions growing at 57% CAGR as SSD prices are reduced Cost/capacity optimized 19% CAGR growth is driven by the explosion of unstructured data Cold/ Static Data I/O Intensive (Flash) Performance Optimized Cost Optimized IT Managers are looking for storage solutions that can span both, hot and cold data, optimizing for performance and value IDC,, Worldwide Enterprise Storage Systems 2013–2017 Forecast, May 2013; Doc #241033

  5. Flash adoption is acceleratingPent-up demand for improving application performance at the right price point Flash prices falling faster than HDDs but still considerably higher than 15K HDDs New form factors: PCIe cards, appliances, shared storage 15K HDD Flash SSD $/GB 7.2K HDD Multi-core processors drive up processor utilization and demand for more I/O per server Virtualization is increasing the demand for random I/O with the “mixer effect”

  6. Flash storage can be deployed at various layers to accelerate application performance Server-based cache (Tier-0) • Flash drives placed directly on high speed PCI bus in application servers • Caching software enables server to leverage flash storage as an extension of memory cache SSD tiering in shared storage • A few SSDs can be added to the array to dramatically increase performance for targeted workloads OR • An entire storage tier can be deployed on SSDs so that complete volumes and datasets can be fit in flash COMING SOON: Fluid Cache for SAN • Benefits of Fluid Cache extended to the entire SAN for maximum performance and reliability Client Network Dell Fluid Cache for DAS on PowerEdge Storage Network Dell Compellent All-Flash or Hybrid-Flash Array

  7. Best use of flash storage: transactional, IOPS-intensive workloads Data Warehousing Indexing OLTP 1 2 VDI 3 Random 4 Sequential Read-intensive Write-intensive OLTP • Manage transaction-oriented applications, i.e. retail, banking • Business can be hurt if data not accessible,slow • During peak usage, customer experience can be effected VDI • Hosting desktop OS within a VM on a centralized server • Facilitate a quick retrieval of gold images • Boot storms, write allocations and latency are issues Indexing • Maintains indexes to allow quicker access to data • Runs on databases to accelerate locating a block of data in queries Data Warehousing • Stored data used to create reports or derive Business Intelligence • Used for data mining, analysis, hypothesis testing, modeling

  8. SSDs have different characteristics Compellent is the first storage solution to tier sub-LUN data across SSDs Write-Intensive (WI) SLC SSDs Read-Intensive (RI) MLC SSDs WI SLC SSDs Tier 1 Workload Mainstream Applications Any usage Mostly Read 90/10 R/W Mix Capacity 400 GB 1.6 TB Write endurance Great Moderate RI MLC SSDs Tier 2 Full-drive writes/day 30 3 Random read perf. Great Great Write performance Great Moderate HDDs Tier 3 Relative costs 4x 1x Compellent Hybrid-Flash Array Tiering flash across multiple SSD types enables Compellent to disrupt the flash cost model with a superior $/GB ratio.

  9. All Dell Primary Storage Solutions offer Flash Workload-optimized solutions for any size enterprise EqualLogic Easy-to-use virtualized storage Converged infrastructures Dell Software Data management and protection PowerVault Affordable, entry level storage Compellent High-performance storage Cloud deployments I/O intensive applications Vertical file-intensive apps Maximize efficiency, drive results faster and boost resiliency

  10. Lets look at Dell Compellent first

  11. Secret sauce behind the flash optimized Compellent: tiering with Data progression Flash-optimized Data Progression leverages the endurance of write-intensive SSDs and the value of read-intensive SSDs. $$$ $$ $ Tier 1 Tier 2 RAID 10 Write Intensive Enterprise SAS SSD RAID 5 Tier 3 RAID 5 RAID 5 Read Optimized Enterprise SAS SSD RAID 6 RAID 6 7.2K SAS HDDs Advanced software tiering seamlessly manages data • Incoming writes are written to the write intensive SSDs for fast access • Read intensive data is automatically moved to less expensive read optimized SSDs keeping Tier 1 free for new incoming writes • Cold and non-performance data is migrated to lower tier, less expensive rotating disk By automatically tiering data across SLC and MLC SSDs, Dell can reduce costs up to 80% compared to other Flash-optimized solutions.

  12. New economies of flash with Compellent : 1/5th the price vs. competitive all-flash solutions Compellent With Flash Tiering Other Vendors Single SSD type (SLC or eMLC) only Write Intensive (SLC) Read Intensive (MLC) Hardware 80% Lower costs Higher costs Power / Cooling Software Support

  13. New economies of flash with Compellent : 1/2 the price vs. competitive hybrid solutions 50% Less Competitors Significant price advantage with Compellent hybrid solutions due to tiering flash across multiple SSD types. Solution Price $/GB Compellent WI or RO SSDs Aggregate $/GB of competitive hybrid systems Using flash tiering: Tier 1: WI SSDs Tier 2: RI SSDs Tier 3: 7k HDDs

  14. Compellent extends the economic value with a feature-rich storage platform Tiering Automatically move data between Flash tiers and HDD tiers Optimal $/GB & $/IO Lower TCO by balancing performance and capacity needs in a single platform Scalability Scale to meet storage requirements with either SSD or HDD  $ Unified Storage Management Manage multi tiered storage solution with a single pane of glass Replication Replicate business critical data between disparate storage systems Enterprise Class Support Leverage best of breed enterprise class support

  15. Slash costs with more efficient use of every disk Automated Tiered Storage Thin Provisioning Block-Level Intelligence Free Space Recovery Unified Block and File Thin Replication Fast Track Space efficient snapshots Compellent can reduce total storage software costs up to 96% and hardware costs up to 44% over a 10 year period. • No pre-allocation required with Thin provisioning • Buy fewer and less expensive drives with Automated tiered storage • Cost-effective Disaster Recovery with Thin Replication • Single solution for unified block and file

  16. Dell EqualLogic

  17. Dell EqualLogic PS6200S is All Flash. & PS6200XS is Hybrid Flash PS6100/6110 Business-responsive performance FS7600/7610 NAS unified storage PS-M4110 blade array One streamlined data center PS4100/4110 Entry-level advanced functionality PS6500/6510 Powerful, virtual. data-center ready • Ideal for general purpose small-medium consolidated loads • Converged storage servers and networking • Highly virtualized • Highly scalable • 10GbE • Ease of management • Ideal for smaller SAN deployments • Remote office • Pay as you grow scalability • Designed for virtualization • 10GbE capability • Ideal for VDI • deployments • Hybrid SAN • Highly virtualized • Scale capacity • & performance • 10GbE capability • Self-optimized • performance • Options for intelligent tiering • Ideal for heavy sequential workloads • 10GbE capability • 2.5x bandwidth per array • Scales up to 2.3PB • Ideal for file & block environments • Flexible pool of scale-out file & block • Massive share capacities for flexibility • Single interface to manage file and block data • 1GbE or 10GbE SAN and client networks • Includes file-level snapshots and replication

  18. Automatic load balancing EqualLogic SAN Storage Pool SSD SAS or NL-SAS Array 1 Array 2 Latency: 0.1s Latency: 0.4s Maximize performance by seamlessly assigning data across all available resources • Make the most of network ports and disk utilization • Automatically adjust to workloads as I/O changes • Dynamically detect and correct latency issues • Purchase storage for the predictable vs. peak workloads • Align hot and cold data within an array and across arrays (based on dynamic workloads) • Improve performance and minimize latency

  19. Storage tiering and pooling Application/Web Servers Back-up Servers Active Directory Servers File Servers Domain Servers SQL Servers Exchange Servers EQL Switched GbE Network EqualLogic EqualLogic Backups Backups Pool 1 Pool 2 AD/Dem eMail Shares Shares Shares SQL Optimize storage and prioritize placement for maximum application performance and lowest cost • Automatically classify and match application data with the most appropriate storage • Tier between multiple generations of arrays • Dedicate pools for unique security, compliance or performance demands • Improve performance at the right pricefor heavy workloads (VDI, transaction database) • Combine tiered storage with service catalogues to simplify management and significantly lower overall storage costs

  20. Thank You

More Related