isilon Scale out NAS  
Scale Out NAS
 
menu Index

White Papers


menu Clustered Storage Revolution

menu Breaking the Bottleneck

menu Economic Benefits of Clustered Storage

menu Power of Infiband and Clustered Storage


What Now?

menu Contact Us / Free Trial

menu Request A Call Back

menu Visit Layer47.com

menu Go to Forum

Discuss Now






   
 

Frequently Asked Questions - US


jetNEXUS Menu How is Scale out NAS different from NAS or SAN?
jetNEXUS Menu What benefits are there of using Scale out NAS?
jetNEXUS Menu What are the benefits of namespace / filesystem consolidation?
jetNEXUS Menu
What sort of performance can I expect from Scale out NAS?
jetNEXUS Menu How do you scale up with Sclae out NAS?
jetNEXUS Menu How does RAID work on Scale out NAS?
jetNEXUS Menu What types of data and protocols are supported?
jetNEXUS Menu Who is using Scale out NAS?
jetNEXUS Menu Is Scale out NAS expensive?
jetNEXUS Menu Who are the market leaders in Scale out NAS?



Isilon Scale out NAS

 



    What is a Hardware Load Balancer? How is Scale out NAS different from NAS or SAN?

    Whilst NAS and SAN are designed to fit different environmental requirements and use different protocols for host access to storage resources, today’s technologies in this space share a common architecture.
    A NAS or SAN storage system will have one or two controllers or heads. These provide hosts with access to the data stored in the disks underneath and also provide.

    1. Performance (CACHE, CPU, Host connectivity ports)
    2. RAID / Data Parity (to protect from data loss)
    3. Management of the storage environment (LUNs, Volumes, Snapshots, Migration / replication etc)

    Scale out NAS has a totally different architecture and is typically designed for rapid data growth in file or unstructured data environments without the performance or management limitations associated with traditional NAS / SAN systems. There is no concept of storage controllers or NAS filer heads. The system is composed of Hardware “nodes” which contain both disk and CPU / CACHE resources. Each node is a peer in the cluster, there is no “master” node.
    These Nodes are clustered together over a common backplane (Isilon uses 40Gbps Infiniband) to form a single storage resource and data is “striped” across the cluster instead of individual RAID groups.







    Why use a Hardware Load Balancer? What benefits are there of using Scale out NAS?

    Scale out NAS delivers many benefits, particularly for those companies and organisations who store a lot of digital or rich media content. The lack of a controller or controllers in the architecture means the system no longer has natural performance limitations in both I/O capability and data throughput to and from the cluster. As each node includes on-board CPU / CACHE and Host connectivity ports, the system scales linearly.  Every time you add more capacity, you are also adding performance (I/O and Throughput). 

    Some clearly defined benefits for Scale out NAS are:

    • Linear Scalability of both performance and capacity with no bottlenecks
    • Massive performance due to pooled resources
    • Namespace / Filesystem aggregation (single volume storage, like a giant Z: drive)
    • More resiliency through clustered architecture, inherently highly available
    • Ease of management (97% easier to manage than traditional SAN technology – Forrester)
    • Flexibility, meet performance or capacity needs by adding hardware on the fly with no downtime

      This is particularly important to users in the life sciences and media space.
      It is very effective as a storage platform behind Illumina or Nanopore Gene Sequencer technology for example.
      Its also used a lot by people working in post production with AVID or Final Cut Pro editing software.
      Both sets of users need massive, scalable, performance storage which presents data in a single volume.

       




    How does Hardware Load Balancing Work? What are the benefits of namespace / filesystem consolidation?

    Users and workgroups who need to share or collaborate on particular data and files can only do so efficiently and seamlessly if the data exists in the same LUN / Volume on the underlying storage array. If the data is spread across multiple volumes or multiple storage arrays, then host access is fragmented and the operational workflow is affected.

    A good example of this is Post Production for the Television and Film industry.

    There are many stages to post production work, some of which include:

  • Data ingest
  • Transcoding (from one file format to another, typically to alter the bit rate)
  • Colour correction
  • Editing
  • CGI and Graphical Rendering
  • Subtitling

  • Throughout this process the engineers and their workstations involved need access to the same files and data.


    If this data (which can be very large) is spread across multiple volumes or multiple storage arrays then it is very difficult to access the same files from multiple workstations on the network. This issue comes as a result of the natural limitations of Operating systems and storage in provisioning large enough LUNs / Volumes to accommodate the data (2TBs is often a common threshold). A single file in this industry can be many TBs in size, so this problem is commonplace.

    Scale out NAS provides a single volume for data which can exceed 2PBs (2000TBs) in size and can be accessed concurrently by any host on the connected network. This allows for significant productivity enhancements and time / efficiency savings in operational workflow. It also greatly simplifies the management of the storage environment and reduces the complexity of backup tasks.





    What are Load Balancing Methods? What sort of performance can I expect from Scale out NAS?

    The available performance from a Scale out NAS array depends largely on the configuration as with many other types of storage, but there are some generalisations to be made in comparison to SAN or NAS architectures.

    Even the smallest Scale out NAS storage systems will typically have more cache and more CPU resource than a SAN or NAS of corresponding cost. This is a result of “stacking” cluster nodes together, each of which has multicore CPUs and in the case of Isilon 4GBs of CACHE per node. Moreover, these are brought together to create a coherent pool of performance resource that can be utilised by the cluster across the Infiniband backplane, very much like supercomputing clusters.

    So in an entry level 3 node cluster you have:

  • 6 x multi core CPUs (traditional NAS / SAN would have just 4, even in more expensive arrays)
  • 3 x 4GB cache modules, 12GB total (traditional NAS / SAN would normally ship with 4-8GBs)
  • 40Gbps capable backplane
  • 3 x Network connections

    The performance of any storage array whilst having many factors, is typically a combination of three things, available CPU resource, CACHE and the number of drives (or spindles) that make up the RAID group where the data resides for a given read / write request.
    The performance available from this clustered / pooled resource is massive, particularly with regard to sequential reads (media streaming is a good example.)

    The data is not being served from a single RAID group (typically 7 disks with one spare, RAID5) as with traditional systems where these disks can become saturated with read requests quickly. Instead, data is striped across all nodes in a cluster no matter how many nodes are installed, so any single file will be served from many more drives (there are 12 in each node, 36 in an entry level cluster) and in many cases if the file is “hot” can be served directly from the cache. In either scenario, the data is delivered across the 40gbps backplane to the node where the request originated and sent on to the network at single stream speeds of up to 400MB/sec (on 10 gig links.)

    This kind of file striping also helps to mitigate issues with “hotspotting” on RAID groups which is a problem in more traditional systems. This is one reason why many web2.0 companies use Scale out NAS storage, as it gives them the ability to deal with a massive increase in traffic to a single file (such as a video showing a major global event) which can become very popular suddenly. (MySpace, Facebook.)

  •  




     What is server Health Checking? How do you scale up with Scale out NAS?

    By its nature, Scale out NAS is many times more scalable than traditional systems.

    The system is made up of “building blocks” called nodes. Each node has disk and performance resources and is adopted to the cluster once racked and powered on. The process of scaling up capacity and / or performance in an Isilon cluster takes just 60 seconds and can be done with no downtime and with no experience of managing storage devices at the push of a button.

    To summarise:

  • Performance and / or capacity can be added on the fly in 60secs
  • There is no downtime involved and data is 100% available whilst the node joins the cluster
  • No changes need to be made (no volume management tasks, no RAID configuration)
  • The work can be done by someone without storage knowledge

  • The Isilon scale out NAS system also allows you to scale performance OR capacity independently through the use of Accelerator nodes (a 1U blade server with CPU, CACHE and HOST / BACKPLANE ports) and EX nodes (a 2U tray of drives with a SAS connection to existing nodes.)

    Unlike traditional systems where performance and capacity planning up front is critical and must take futures in to account when sizing your environment, with Scale out NAS you can start with what you need today knowing that you are able to scale up when required to meet new challenges in your storage environment, Pay as you Grow.

    Isilon IQ can scale to 96 nodes and more than 2,500TBs of capacity in a single cluster.

     



    What is Persistence? How does RAID work with Scale out NAS?

    With the leading solution in this marketplace from Isilon Systems, RAID is not used for data protection. The files are broken up in to blocks and written to all nodes in a cluster no matter how many nodes exist (each node has 12 drives) and parity blocks are written to protect data from loss and to enable files / blocks to be reconstructed in the event of an outage or corruption.

    The available parity settings are:

  • N+1 which is equivalent to RAID5
  • N+2 which is equivalent to RAID6 (or RAID DP, Double Parity)
  • N+3 and N+4 for more critical data.

    N+1 will allow for a single drive failure or a single node failure and give 100% data protection
    N+2 will allow for two drives or two node failures and give 100% data protection
    N+3 and +4 follow the same structure

  • In addition to providing more powerful protection algorithms than ANY traditional storage system, this data protection method is also more flexible.

    It is possible to set a GLOBAL protection level of N+1 and then set different individual protection levels against folders within the filesystem (the CEOs holiday pictures for example....) which can be set at +2, +3, or +4 as required. This gives much greater control over data protection than with traditional RAID technologies whilst being much easier to implement and manage.



     

    What is SSL Acceleration and Offload? What types of data and protocols are supported?

    SAN and NAS systems cater for different types of data and usage.

    SAN storage is very good at supporting highly transactional environments such as Databases and systems requiring random read / write access to blocks of data, particularly where the number of individual I/O requests is very high.

    NAS is traditionally used to support groups of users who require file level access to sequential data such as back office documents, Word, Excel, PowerPoint etc.

    Scale out NAS systems usually support NAS protocols, CIFS (windows file sharing), NFS (Linux / Unix file sharing) and HTTP / Webdav .

    Isilon IQ comes licensed for all of these protocols and is designed to support large un-structured file data like documents, images, video and web content etc.

    It currently does not support block protocols such as FC and ISCSI but can be used for database / email and other more block orientated applications when used as NFS storage behind VMWare or similar server virtualisation technologies.


     
    Why Offload SSL? Who is using Scale out NAS?

    Some of the world’s biggest names are using Scale out NAS,

    MySpace – Isilon IQ is used as the storage platform behind this popular web 2.0 site
    Facebook – Isilon IQ is used as the storage platform for this social networking site
    BBC iPlayer – Isilon IQ is used for content delivery
    NASA – Isilon IQ is used to store satellite images from across the globe
    Second Life – Linden Labs use Isilon to power the SL virtual world
    European Bio-Informatics - Used to store Human Genome Project Data
    Medical Research Council - Used for life sciences research, gene sequencing

    Clustered Storage is particularly well suited to:

  • Media / Broadcast industry
  • Oil & Gas exploration
  • Medical Imaging and Gene Sequencing
  • ISP / Telco and Hosted Storage environments
  • Web 2.0
  • And anywhere else where there is a lot of file based data which is growing quickly, such as document management systems and archives.

  •  



    What is a Next Generation Load Balancer? Is Scale out NAS expensive?

    No. In fact it can be very cost effective, particularly over the total life of the system.

    Typical reductions in storage management and complexity can approach 97% (independently verified by Forrester, see document links in the NAV to the left)

    Users can achieve up to 27% increase in storage utilisation over traditional systems (Forrester again)
    The outright capital purchase cost for Isilon IQ is on a par with systems from Network Appliance and EMC, HP etc. It is an enterprise class storage system with all the features and functions you would expect and some that may surprise you.

    Need a quote? Click here
     



    What is Web Acceleration? Who are the market leaders in Scale out NAS?

    The market leaders in this space and recognised by Gartner as being ahead of the rest is Isilon Systems. Isilon have been around since 2000 and support some of the largest global names.


    Their storage solution Isilon IQ has been designed from the ground up to meet the challenges resulting from the rapid growth of digital media and unstructured data.


    From NBC’s coverage of the Olympic games to Satelite images from NASA and the entire Human Genome Project, Isilon is in use throughout a huge variety of organisations and the list is growing rapidly..

    For more information click here
    For Pricing for a clustered storage please click here




 
 


   
       
     

For more information please visit:

www.layer47.com