Pure Storage Overview

Pure Storage Overview

Course Overview

  • Founders of Pure Storage: Understanding who founded the company.
  • Founding Vision: Understanding why Pure Storage was started despite the presence of established storage companies like HP, Dell EMC, and NetApp. This involves understanding what Pure Storage did differently in building products and solutions.
  • Products and Solutions: Identifying the products and solutions Pure Storage offers.
  • Core Principles: Defining the core principles of Pure Storage's product architecture.
  • Right Size Guarantee: Understanding the guarantee offered.
  • Evergreen Architecture/Subscription: Briefly discussing, with more detail in a later module.

Storage Timeline & Fundamentals

  • Storage Timeline: How storage technology has evolved.
  • What is Storage and Why do we need it?
    • Data needs to be in a centralized location for global access (e.g., social media, payment gateways, banking sites).
    • Applications are hosted on servers (hosts).
    • Servers are high-end computers that operate round the clock.
    • Servers have processors (CPU), RAM (DRAM), and hard drives for the OS and data.
    • Servers are used for hosting applications, not for data storage, processing, protection, recovery, or restoration.
    • Storage hardware sits behind servers.
      • It receives data from servers.
      • Processes the data.
      • Stores the data on disks.
    • Storage Hardware components:
      • Controllers: Contain processors, DRAM, OS, network cards, and HBA card slots.
      • Disk Enclosures: Connected to controllers, contain numerous disks for data storage.
      • Two identical controllers for redundancy.

Data Storage Product Building Blocks

  • Controllers.
  • Disks.
  • Expansion Disk Enclosures.
  • Types of Disks Used in Storage:
    • HDD (Hard Disk Drive)
      • First type of disk.
      • Invented in 1956.
      • The first HDD had a 5MB capacity and was the size of a refrigerator.
      • Cost: $50,000.
      • Contained a magnetic platter, spindle, actuator, and actuator arm.
      • HDD is a spinning/rotating media.
      • HDDs had limited performance and reliability.
      • Capacity improvements occurred over time (1GB, 10GB, 100GB, 500GB, 30TB).
      • Physical size reduced (3.5 inch, 2.5 inch).
      • Limitations: Less reliable, limited performance and capacity.
    • SCSI Protocol:
      • Developed for HDD communication between controller and hard disk drive.
      • Protocol defined as a "language" for communication.
      • The SCSI protocol is required for communication between your controller and your hard disk drive.
      • For communication that's SCSI protocol.
      • The SCSI protocol is required for communication between your controller and your hard disk drive.
    • SSD (Solid State Drive)
      • Aman said, the other type of drive behind is SSD.
      • Built on flash technology.
      • Also called flash drives.
      • Contains NAND gates (cells). Data is flashed onto these NAND gates.
      • No magnetic platter, spindle, actuator, or actuator arm.
      • Better performance and highly reliable.
      • Can build larger capacity drives.
      • Different types of SSDs:
        • SLC (Single Layer Cell): One bit of data per cell.
        • MLC (Multi-Layer Cell): Multiple layers, two bits of data per cell.
        • TLC (Triple Layer Cell): Three layers, three bits of data per cell.
        • QLC (Quad Layer Cell): Four layers, four bits of data per cell.
      • Example:
        • Hypothetical example of different SSD with 100 cells:
          • SLC: 100 bits of data.
          • MLC: 200 bits of data.
          • TLC: 300 bits of data.
          • QLC: 400 bits of data.
      • SSD gives larger capacity, low failure rate and better performance.
      • Physical size of SSD is still the same.
    • Cost Considerations:
      • Storage manufacturers were initially hesitant to use SSDs due to high costs.

Pure Storage Founding

  • Pure Storage founded in 2009.
  • Founders: John Kalogridis (COS) and John Hayes.
  • Founders analyzed limitations of spinning media disks.
  • Decided not to use HDDs in data storage products.
  • Founded by two people, John Kalgo and John Hayes.
  • John Kalgo and John Hayes started Pure Storage company.
  • John is still with Pure Storage.
  • He is the chief visionary officer for our company today.
  • John Hayes, he left Pure Storage.
  • Headquartered in Santa Clara, California, USA.
  • Develops all-flash data storage hardware and storage software products.

Founding Vision

  • Build storage with high performance, efficiency, and density.
  • Efficiency: Store more data in less raw space.
    • Achieved through data reduction:
      • Deduplication.
      • Compression.
    • Example:
      • Repeating data (A, B, C, D).
        • Instead of storing it multiple times, store it once and point to its pointer.
      • Compression removes sub-blocks of zeros and known patterns.
  • Data reduction technologies data can be packed to occupy less space by removing repeating patterns and compressing data.
    • Data can be packed to occupy less space by removing repeating patterns and compressing data.
  • High Density: More raw capacity.
    • Use flash drives.

Flash Drives Benefits

  • Better performance.
  • Implement deduplication and compression.
  • Larger capacity.

Strategy to Reduce Costs

  • Manufacturing drives instead of buying them from vendors.

Cloud Considerations

  • Building hardware good for structured and unstructured data from day one.
  • Products: FlashArray (structured data) and FlashBlade (unstructured data).

Simplicity

  • Data storage should be simple to deploy, manage, and troubleshoot.
  • Enable cloud-based management with Evergreen.
  • Pure One for cloud-based monitoring.
  • Today you buy one model of a storage. And tomorrow when you want to buy another model of a storage, You should not be buying the whole thing what you had already.
  • Evergreen Model: Ongoing value, avoid complete replacements during upgrades.
    • When upgrading, replace only the different components (e.g., controllers).

Hotel Scenario: Analogy with phone upgrades

Today you buy one model of a storage. And tomorrow when you want to buy another model of a storage, You should not be buying the whole thing what you had already.

  • Phone Example
    The difference between my phone and your phone will be within in with the in terms of, you know, the processor that is used, the chipset, or the amount of memory.
    Mine is, like, maybe, I don't know, maybe a eight GB RAM and one two fifty six GB RAM. Yours may be a 16 GB RAM and five twelve GB RAM.
    But Story, that should not be the case.

FlashArray Example: Evergreen model

    Today, when we make this flash array, we are making different models, like x 20, x 50, x 70, x 90. We are building a different model. From model to model, what is the difference?
    Difference is in terms with terms of what type of processor we are using, what chipset we are using. The difference is with regards to the amount of DRAM that the controller contains

    You will just have to remove that x 20 controllers and put x 50 controllers. Your system become x 50. So which means that whatever chassis that you use with x 20, it can be used with x 50. Whatever drives you use with x 20 can be used with x 50.

Proactive Support

  • Constant monitoring via Pure One.
  • Early problem identification and failure prediction.
  • Enabling call home feature enable to identify the problems of your body well in advance, you can take the preventive measures and avoid any major diseases. You can avoid the failure of any organs.

Differentiation from Competitors

  • Flash as the Medium of Change : Never used spinning-media drives.
  • Ten times faster than storage with hard disk drives.
  • Smaller: Each drive gives larger capacity. Less number of drives are used, which decreases the number of disk enclosures.
  • There is a decreased power consumption due to fewer components (drives, disk enclosures, power supplies).

Business Model

  • Evergreen Subscription Model:
    • Evergreen One.
    • Evergreen Flex.
    • Evergreen Forever: Pay the full amount and you buy the product, and you deploy it in your data center.
      • The product is owned by you irrespective of how much you use the product.
        No investment.
    • Entire hardware, the chassis, controllers, power supply drives, everything will be on a rental basis. Where you get the hardware, you deploy it in your data center, but the product will still be owned by the pure storage.

Technology Leadership

  • Using SSD, the flash drive.
  • Using NVMe protocol, rather than the 60 year old technology of hard disk drives.
  • Direct Flash Module (DFM): The processor directly communicates with all the NAND gates of an SSD/Flash drive.
  • Magic Quadrant Leader as per Gartner Report.
  • 82% of customers say that, hey. They will go out in the market and promote our product.
  • 80 two percent of our customers say that, hey. They will go out in the market and promote our product.
  • Net Promoter Score (NPS) : Very happy is because our NPS score, the Net Promoter Score, stands at 82.80 two percent of our customers say that, hey. They will go out in the market and promote our product. Right? Well, the industry average is, like, 21, 20 six.

Products and Solutions

  • FlashArray:
    • Provides both block and file services.
    • Different models: X series, XL series, C series, and E series.
  • FlashBlade: Provides file and object services.
  • FlashStack: A converged infrastructure solution where you get the server, the networking switches, and the storage, everything bundled together. Right.
  • Pure Cloud Block Store: The CloudBlock store, it provides that purity features on your AWS platform. You can also work on the Azure platform

Core Principles

  • To make it effortless, we have simplified a lot of stuff.
  • Effortless.
  • Efficient.
  • Evergreen. (Effortless, efficient and evergreen.)
Effortless Detailed
  • Simplified RAID.
    • Only one type of RAID.
    • RAID or this RAID is nothing but, like, your RAID six where it provides the fault tolerance of two.
      • Fault tolerance of two: Even if two blocks of data on different drives fail simultaneously, data can be reconstructed, because always our array is a nondestructive operation.
        Whenever you do a hardware upgrade or a software upgrade, it is nondestructive.
        We don't need any downtime.
  • No host to array block alignment issues. All the volumes will be having the same block size. Always do a variable block engine so that take it can take the data or. So to variable logs
  • Easy troubleshooting with pure one.
  • The performance has been made easy because if your array is connected to the pure one, the pure one could have done the log analysis, and it could have identified where is the performance bottleneck whether Because of the issues with the controller is it because of a slower disk?. So what is causing the performance to degrade?
  • No more tuning or tweaking, and the data will be rebalanced automatically data, the data that is there on the existing 20 drives will have to be distributed to that 20 more drives that you add later (Data will be rebalanced automatically)
  • Automatic rebalancing.
  • No Day long Raid rebuilds. (Distributed Sparing) * Spare blocks are there on every drive. * The data of the drive will be reconstructed, and it will be returned to the spare drive Okay. So unlike other storages, this storage, the flash array does not do not have a single drive as a spare drive. Right? It does not have a single drive as a spare drive. So it has distribute it has spare blocks are there on every drive spare box has been there on it.
    • This distributed sparing means.So, I mean, like, if I added them two or three drives fails also, it will be not an issue. Right? You can tolerate failure of up to two drives within each RAID segment.
  • No multiple data tiers: As for the multiple tiers is available. Like, you have SSDs and you have HDDs also. Right Which will give a different types of performance
    • Don't have the multiple tiers of storage storage anymore because you don't even have the multiple tiers of storage.
  • Less cabling.
    * See, if you have seen the other storages, you have the storages with two controllers. In each controller, you have, like, two ports to connect to the expansion disk enclosure And they say how many expansion disk enclosure can be connected? You can connect up to nine expansion disk enclosure.
  • Simplified cabling.
  • Noisy Neighbour.
    * identifying that noisy neighbor had always been a problem earlier. Four VMs are accessing one single volume. VM one, VM some few VMs are out to wait in queue .So which is a VM that is doing too many number of IO
    .*vCenter will share the logs with your flash array. Flash array, you enable the call home feature so it gets connected to the Pure one.. You wanna go and see.So identifying that which is a VM that is doing too many number of IOs, Right?
  • No More Painful Upgrades.
    * Because, we don't need any downtime for software upgrades ,Only one controller will reboot at a time, (Non-destructive operation)
Efficient Detailed
  • Six nines of high availability.
  • Run multiple different workloads (Multiple workloads can be run on the same storage).
  • Global flash management, garbage collection)Because it is,It reduces the space consumption.
Evergreen Details
  • You can also log in to the Pure one and generate the reports. You can see the logs. You can monitor array. What is a right size guarantee? As I told you, our array supports a deduplication and compression, so it reduces a space a consumption is because of this, That the company, the founding vision, and this is how the product has been built from the ground up.
  • Predictive analytics, forecasting.
Right Size Guarantee
  • Because of deduplication compression, the space will be reduced.
  • We understand what is your data requirement and then we apply to the workload intellegence to provide the right size guarantee. So if we do not deliver the guaranteed capacity, we will make it right by adding flash drives which is for free of cost is just like in case of the customer receives power consumption reduction and that fails, the power consumption decrease. It is just to say it that we are giving money in compensation to power consumption failing

Deep Dive into FlashArray Hardware

*Understanding what is this NVMe protocol. Because in the flash array, what we are using, what kind of disc are we using?
*WE are using the NVMe protocol based SSDs. Right? NVMe protocol based SSDs, we call it as a direct flash module. So first, we just need to understand why we require this NVMe protocol, how is it different than the SCSI, and why are we using this direct flash module, NVMe based SSD instead of a SaaS based SSD that the competitors use. And we'll understand what are the models we have in the FlashArray family.*Right?
*Different models - will learn in deep dive. Optionally, a direct memory module to launch. What the benefits will occur*

  • Module Objectives
    • Understanding of NVMe protocol.
    • Overview of FlashArray models.
    • Understanding of Direct Memory Module.
    • Interactive App for Hardware Visualization - RightSize
Product Recap- Overview.

flash array, and we have a FlashBlade ,These are the two different hardwares we have FlashArray(purity FA slash) runs with an operating system called as a purit slash slash FA. FlashBlade product runs on an operating system called as purity slash slash b .hough the operating system for both the hardwares has the name purity, so they are different. Right

flash Array: Purity FA Detailed

*Some of the features on both the products will be same.Right?There are some common features on both the OS, but they are different .So, for for example, purity slash slash fb cannot be installed on a flash array but whereas purity slash slash f a cannot be installed on your Flash play

Flash Array: Functions.

Flash array , a unified story that gives both of block and file services. Right ?So if you see the type of storages that were available couple of years back or maybe two decades back, about twenty years back or fifteen years back, you had a block storage . You had a file storage .Right

  • NAS and your SAN, NAS storage and your block storage. But now what we are what we have built is a flash array, which is a unified storage that can be used as a block storage and the same storage can also be used as a file storage* There may be three servers, three hosts connected to your flash array But one of of the If it is exported to only one host then it is volume only to him,not to Two and 3 ,The host operating system will put the file system on this one and then it will be in. Then the host can start writing the data You pick up your phone, you dial one number. In your phone, there are hundred phone numbers that is stored.
    You have hundred friends. Hundred phone numbers are stored But from your phone, when you are calling a particular person, only that person can listen to you, and others cannot listen.. Right but on a storage will put the file system within and the forders will be given name. Everyone needs access to the folder to broad cast and can also acccess it as what a file system is.. Right
    Right?Some organizations, they will go for the block storage. Some organizations, they will require a file storage. Some organizations, they need both, but they will use a block storage for a different purpose. They will will use a file storage for different purpose. So when one storage once if I make one storage product, which gives a block service, another storage product which gives a files file service. Right?
    What are the available series? We have different series. Right? We have XL, X, C, and a E ,With the XL series, we have two models, one thirty and one seventy with the X series now currently , we have the revision four in the XR four series. We have different models like x twenty, fifty, 70, and 90, with c, r four r four means revision four. That is what we currently have
    *The r one, r two, and r three are end of life and r fourwhat are models currently we are selling? Twenty, fifty, 70, and nine with the e series. There's only e series there's no different models.
    *This series to series, what is the difference?*
    X series is designed to provide high performance. Excel series is designed to provide higher performance with capacity. C series is for capacity optimized. It gives lesser performance compared to x, but it provides a higher capacity. So this is our performance model. This is our performance plus capacity balanced. Right?This one is capacity optimized ,Right and the Excel series, if you see, it gives high performance plus high capacity. Both performance at scale, we say .Right? ,Right*
    *About the C Series:
    *Not as high as the x or the Excel, but little bit lower performance But, yes, definitely higher capacity* Excel series So it will have more number of what do you say more number of p PCI slots So more number of HBA cards.
    So many number of host ports so many number of host can be connected ,More number of host can be connected to the Excel series compared to your x series. Right? , With the x series, if you see, what are the models we have? So when the x series came up, we came up with that revision one. In the revision one, we had x ten fifty, 10, 20, 50, 70 nine .

The x series performance optimized, which is aimed to provide to as low as two fifty microsecond latency to your hosts IOIt it can be used for mission critical applications for for workload consolidation or in a virtualized environment..Excel series- it gives performance as well higher capacityBecause in the 20 discs are available at base level in a five, u rack. Which yields for number of disc and thus yielding more capacity . Right. So then we came up with r two So r one was end up Right and r two again ten, twenty, 50, 70 nine .Then we came up with r three, right, in 2023 At ten, twenty, 50, 70, 90, In the r four, we came up with twenty, twenty four last year, So we bought we and we just ex excluded We just made this '20, '15, '70, and 90 (New Models:X20, X50, X70,X90.Right year)*
Ony Ear revisions Previous revisions becomes end But whereas here coming back to X the for x or the c series what are now using is what is the NVMe base SSD with the controller and flash translation and the controller Right That we're gonna talk
So on the same capacity is supported but as additional 18.3 b The Drives are also introduced If there is no change with regards to the capacity This expansion, shelf here, only one expansion shelf can be connected, But here if you see two expansion shelf.(DFS means direct flash shelf. An expansion shelf where you will have the NVMe basis as the direct flash modules. That expansion shelf is called as direct flash shelf. Okay)Right
,Right. That varies from model to model.

Excel Detailed

,The top one is controller zero, bottom one is controller one..Excel series, we don't have the nVM as a separate module It is there long with their drive- That increases amount of data written speed and memory of the nand states with controllers with flash modules (more performance)
, The size of the battery is less, so that's that's the reason why we were able to keep it within that drive itself..Our controllers, it supported both SaaS based SSD and embedded SSD, but you don't have to worry about this and embedded SSD but you don't have to worry about this and the parts until twenty twenty, , but now we no longer sell the SaaS based expansion shelf, as SaaS based, SSD, because its old , so we dont do to new customers.. ,We have used the DLC direct layer sells to manufacture our disks .RightRightRightExpansion shelf, when you connect to any model, the expansion shelf is same for all the models. Expand the expansion shelf is always of the height three u You can have up to twenty, twenty two, 20 four, 20 six, and 28 DFMs.
On the left side in, expansion shelf on each series, how many DFMs the DFM capacities you can see over here.. How many expansion shelf can i be connected to model for expansion shelf. The minimum drives put in the base enclosure Is what is 10 and 20 which is of course which will also be called for. C series and e series what about the logic has made up by all time, So let that cell which each and the quad and tell it so so like in short*How you manufacture the larger capacity drives *What are you*For the x and x l series that drives, we manufacture is based on a DLC type (triple layer cell)..
*

The DFM Flash Model with Direct RAM Details

We can do that The distributed NVR has speeds of over the modules, or the right of
*Individual when it is taken out
You used to. You were getting there, then we get a new vision of getting set out is all about a 30 minute, and it is an overview, as I'm sorry (A quick summary of the previous discussions on DFM flash and model with direct RAM)

Back End Cable.Details

So on the base enclosure, the minimum you will put in the x series, c series, and e series, the minimum drives you put in the in the base enclosure. The minimum drives you put is 10, maximum 20 Expansion shelf, fourteen, twenty eight , C series and e series the base and said okay Okay. For what is not working can see the three dimensional show that so you can go can you can just you can explore it, explore it is just this to click the module controller show this can go as I used I want to show you because where exactly why we becauseHere, this controller look like for the front same and if the customer should *If you already has it that's what we're changing
This where our

Direct Memory Model Details

We have tried memory You said that
, When So not that you should move the so connect to the port*You You are and four ports for expansion for application use

  • So if you also you want to connect there and support it If you is a card to come here because but not can you So you get can if because But the canWhat The series the .The the .Okay We have so there we for of can ,OkayNot you,The can,What. the the the the the the the all of is. ,And For the 2814You If You To,Yeah*The that is
    *The You What a*Here,If *So you have for one so you on the. what *For . is is the *The is and has to can has.Right has has has and said said has said and have Right to is for has to,that*The,The is is is That. And,What for
    The or And If Yeah. * The or hasIsYeah.The to to that, right If thatThat all If so withRight
    Alright So. Okay. Okay. So to So with That Yes. Yes,Yes
    So The This That Yeah YeahSo here are the ports out. I mean, everything Right? * Yeah, Yeah for , Right to.The *To *Why did you have to say *Yes of Yeah, Okay to You So.You That what said *Right And with what it you has Then that can,You willYou itTheYou*The thatYou theYou knowYeah