7
Copyright The TANEJA Group, Inc. 2013. All Rights Reserved. 1 of 7 87 Elm Street, Suite 900 Hopkinton, MA 01748 T: 508.435.2556 F: 508.435.2557 www.tanejagroup.com PRODUCT PROFILE ACCELERATING IOINTENSIVE VIRTUALIZED APPLICATIONS Shifting Into High Gear with Astute Networks ViSX G4 Flashbased Performance Storage Appliances MAY 2013 Virtualization is a great boon to IT shops for it’s flexible and costefficient infrastructure that consolidates previously physically sprawling applications and servers. But current hypervisor technologies are still evolving support for highend performance requirements, especially for shared but externaltotheserver resources like storage. Many IT shops trying to convert deeper into their portfolio based on initial virtualization success have run up against frustrating limits as to which of their more demanding applications can be effectively virtualized and still be delivered with good service. Virtualizing IO intensive applications remains a significant challenge, especially in IT organizations that are building their virtualization environments with traditional storage arrays in order to enable virtual machine migration and provide sufficient data protection. Most mission critical applications are IO intensive, relying on databases or providing enduser communication channels (e.g. email). VDI efforts in particular can become IO performance constrained even at small scales. If virtualized IO constraints with traditional storage are limiting the number of vms per server, or preventing virtual hosting for IO intensive apps like databases, email, and VDI, then flash technologies seem to promise a tailormade highperformance solution. But flash can be an expensive investment to acquire and hard to deploy and manage to assure a solid ROI. With flash, performance can be bought, but costconscious IT organizations must choose wisely or risk implementing expensive and ultimately unsatisfactory or limiting solutions. An ideal performance solution would drop in to an existing virtual environment, currently constrained or otherwise, and accelerate performance without rearchitecting the storage layer, without reimplementing data protection, and without creating new burdens for the virtual server administrator. It should supercharge the environment to enable greater vm density, deliver better than physical performance for designated apps, and support VDI working at scales that make sense for wholesale adoption. In this product profile we will examine in detail the new Astute Networks ViSX G4 Performance Storage Appliance to see how it fits this ideal profile – deploying without disruption to truly accelerate virtualization performance, increase VM density, and further the adoption of virtualization across IO and missioncritical applications. POOR IO PERFORMANCE STALLS VIRTUALIZATION INITIATIVES The main virtue of virtualization is that it decouples applications from physical infrastructure, enabling both to be managed more independently. This freedom allows IT shops to build architectures that deliver great service in a highly costefficient manner, and support provisioning new service orders quickly and flexibly. For compute power, virtual hypervisors can pool together scaleout clusters of commodity servers and carve them up logically into virtual machines. Then to support virtual machine elasticity and to enable dynamic migration across that compute pool, centralized storage arrays are interconnected to the server cluster. This architecture works well

PRODUCTPROFILE( ACCELERATING*IO INTENSIVE ...docs.media.bitpipe.com/io_10x/io_109901/item_702090...Copyright *The*TANEJA*Group,*Inc.*2013.All*Rights*Reserved.* 1*of*7* 87Elm*Street,*Suite900

  • Upload
    others

  • View
    3

  • Download
    0

Embed Size (px)

Citation preview

Page 1: PRODUCTPROFILE( ACCELERATING*IO INTENSIVE ...docs.media.bitpipe.com/io_10x/io_109901/item_702090...Copyright *The*TANEJA*Group,*Inc.*2013.All*Rights*Reserved.* 1*of*7* 87Elm*Street,*Suite900

 

Copyright  The  TANEJA  Group,  Inc.  2013.  All  Rights  Reserved.   1  of  7  87  Elm  Street,  Suite  900    Hopkinton,  MA    01748    T:  508.435.2556    F:  508.435.2557   www.tanejagroup.com    

PRODUCT  PROFILE  

ACCELERATING  IO-­‐INTENSIVE  VIRTUALIZED  APPLICATIONS  Shifting  Into  High  Gear  with  Astute  Networks  

ViSX  G4  Flash-­‐based  Performance  Storage  Appliances  MAY  2013  

Virtualization   is   a   great   boon   to   IT   shops   for   it’s   flexible   and   cost-­‐efficient  infrastructure   that   consolidates   previously   physically   sprawling   applications   and  servers.  But  current  hypervisor  technologies  are  still  evolving  support  for  high-­‐end  performance   requirements,   especially   for   shared   but   external-­‐to-­‐the-­‐server  resources   like  storage.  Many  IT  shops  trying  to  convert  deeper   into  their  portfolio  

based  on  initial  virtualization  success  have  run  up  against  frustrating  limits  as  to  which  of  their  more  demanding   applications   can   be   effectively   virtualized   and   still   be   delivered   with   good   service.  Virtualizing   IO   intensive   applications   remains   a   significant   challenge,   especially   in   IT  organizations  that  are  building  their  virtualization  environments  with  traditional  storage  arrays  in  order  to  enable  virtual  machine  migration  and  provide  sufficient  data  protection.    

Most   mission   critical   applications   are   IO   intensive,   relying   on   databases   or   providing   end-­‐user  communication   channels   (e.g.   email).   VDI   efforts   in   particular   can   become   IO   performance  constrained  even  at  small  scales.  If  virtualized  IO  constraints  with  traditional  storage  are  limiting  the  number  of  vms  per  server,  or  preventing  virtual  hosting  for  IO  intensive  apps  like  databases,  email,  and  VDI,  then  flash  technologies  seem  to  promise  a  tailor-­‐made  high-­‐performance  solution.  But  flash  can  be  an  expensive  investment  to  acquire  and  hard  to  deploy  and  manage  to  assure  a  solid  ROI.  With  flash,   performance   can   be   bought,   but   cost-­‐conscious   IT   organizations  must   choose  wisely   or   risk  implementing  expensive  and  ultimately  unsatisfactory  or  limiting  solutions.  

An   ideal   performance   solution   would   drop   in   to   an   existing   virtual   environment,   currently  constrained   or   otherwise,   and   accelerate   performance   without   re-­‐architecting   the   storage   layer,  without   re-­‐implementing   data   protection,   and  without   creating   new  burdens   for   the   virtual   server  administrator.   It   should   supercharge   the   environment   to   enable   greater   vm  density,   deliver   better  than  physical  performance  for  designated  apps,  and  support  VDI  working  at  scales  that  make  sense  for  wholesale   adoption.   In   this   product   profile  we  will   examine   in   detail   the  new  Astute  Networks  ViSX   G4   Performance   Storage   Appliance   to   see   how   it   fits   this   ideal   profile   –   deploying   without  disruption   to   truly   accelerate   virtualization   performance,   increase   VM   density,   and   further   the  adoption  of  virtualization  across  IO  and  mission-­‐critical  applications.  

 

POOR  IO  PERFORMANCE  STALLS  VIRTUALIZATION  INITIATIVES  The   main   virtue   of   virtualization   is   that   it   decouples   applications   from   physical   infrastructure,  enabling   both   to   be   managed   more   independently.   This   freedom   allows   IT   shops   to   build  architectures   that   deliver   great   service   in   a   highly   cost-­‐efficient  manner,   and   support   provisioning  new   service  orders  quickly   and   flexibly.   For   compute  power,   virtual   hypervisors   can  pool   together  scale-­‐out   clusters  of   commodity   servers  and  carve   them  up   logically   into  virtual  machines.  Then   to  support   virtual   machine   elasticity   and   to   enable   dynamic   migration   across   that   compute   pool,  centralized   storage   arrays   are   interconnected   to   the   server   cluster.   This   architecture   works   well  

Page 2: PRODUCTPROFILE( ACCELERATING*IO INTENSIVE ...docs.media.bitpipe.com/io_10x/io_109901/item_702090...Copyright *The*TANEJA*Group,*Inc.*2013.All*Rights*Reserved.* 1*of*7* 87Elm*Street,*Suite900

 

Copyright  The  TANEJA  Group,  Inc.  2013.  All  Rights  Reserved.   2  of  7  87  Elm  Street,  Suite  900    Hopkinton,  MA    01748    T:  508.435.2556    F:  508.435.2557   www.tanejagroup.com    

Product  Profile  

enough   for   consolidating   business   applications   that   only  make   intermittent   demands   on   IO   or   can  tolerate  IO  latency.  

While  early  phases  of  virtualization  tended  to   focus  on   less  critical  and   less  IO  demanding  business  applications,  they  still  required  scalable  capacity  and  data  protection  at  the  same  level  as  when  they  were  physically  hosted  so  it  was  natural  to  simply  leverage  existing  array  designs  and  deployments.  But   now   many   organizations   are   attempting   to   virtualize   increasingly   IO   intensive   applications  including   key   databases,   email   servers,   and   VDI   deployments.   These   applications,   once   virtualized  and   concentrated,   severely   strain   virtualization   architectures   with   contention   at   the   traditional  shared  storage  layer.  

The  consolidation  of  applications  bAy  way  of  virtualization  aggregates  IO  across  virtual  machines  and  delivers  an  IO  load  to  the  storage  layer  that  is  almost  fully  random,  defeating  many  traditional  storage  performance   features.  Traditional  HDD  spindles   are  best   for   serial   IO,  but   aggregating   lots  of   them  cleverly   can   provide   high   performance.   However,   upgrading   SAN’s   to   high-­‐performance   arrays  delivering   enough   spindles   from   traditional   capacity   storage   vendors   unacceptably   increases   total  cost   and   complexity   for   virtualization   projects.   Left   without   a   viable   solution,   this   virtualized   IO  performance   bottleneck   effectively   limits   overall   vm   density   and   stalls   out   further   virtualization  initiatives.  

Mixed  Results  For  Flash  “Optimizations”  Flash  technologies  are  everywhere  these  days,  and  at  first  blush  it  should  be  easy  to  “just  add  some  flash”   to   an   environment   for   a   quick   performance   boost.   But   a   big   problem   with   many   flash  “solutions”  is  simply  getting  cost-­‐effective  performance  from  the  flash  investment.  Assuming  that  cost  is  an  objective,  it’s  critical  to  put  the  right  flash  implementation  at  the  right  place  in  the  IO  path  to  not  only  achieve  the  desired  performance  but  cost-­‐efficiently  make  best  use  of  the  flash  resources.  Flash  solutions   are   not   equal   in   cost,   effort   or   performance   payback   (see   sidebar   on   Cost   Evaluation   of  Performance  solutions).  

Where  can  flash  technology  be  applied  to  address  the  virtual  IO  performance  bottleneck?  Let’s  look  at  a   few   of   the   common   options   in   terms   of   acceleration   of   virtual   IO,   and   cost-­‐efficiency   as   an   IT  investment:  

• Flash  in  the  server  –  Dedicated  server  flash  provides  a  local  IO  performance  boost,  but  at  a  cost  that  scales  with  the  number  of  servers.  Server  flash  generally  can’t  be  shared  as  a  pooled  resource  and  it  can  be  difficult  to  balance  workloads  to  make  best  use  of  the  flash  resources.  

Cost  Evaluation  of  Performance  Solutions    

When   comparing  performance  solutions,   it   is   important   to   keep  an   eye  on   the  relevant  metrics.    Traditional   storage  arrays  often   tout  price-­‐per-­‐storage-­‐capacity   figures   (e.g.  $/TB)   that  miss   the  point  when  investing   in  performance.  When  evaluating  IO  performance  accelerating  storage,  the  better  primary  comparison  metric  would  be  based  on  price-­‐per-­‐performance  (e.g.  $/IOP)  or  price-­‐per-­‐“workload”  (e.g.  $/VDI  desktop).  

When  evaluating  solutions  with  these  kinds  of  price-­‐per-­‐performance  metrics  it  is  also  important  to  consider  not  just  the  acquisition  cost  (CAPEX),  but  the  costs  of  implementation  and  operations  (OPEX)  as  well   as   costs  avoided,   the   potential   return   on  opportunity,   and   future  expansion.   For  example,   while   the   list   acquisition   price-­‐per-­‐performance   of   the   ViSX   G4   is   designed   to   be  favorable  compared  to  other  flash  solutions,  when  the  total  costs  are  tallied  up  the  ViSX  G4  can  far  outshine  the  alternatives  in  existing  virtualization  infrastructures.  

Page 3: PRODUCTPROFILE( ACCELERATING*IO INTENSIVE ...docs.media.bitpipe.com/io_10x/io_109901/item_702090...Copyright *The*TANEJA*Group,*Inc.*2013.All*Rights*Reserved.* 1*of*7* 87Elm*Street,*Suite900

 

Copyright  The  TANEJA  Group,  Inc.  2013.  All  Rights  Reserved.   3  of  7  87  Elm  Street,  Suite  900    Hopkinton,  MA    01748    T:  508.435.2556    F:  508.435.2557   www.tanejagroup.com    

Product  Profile  

The  use  of  server  SSD  can  impede  storage  vMotion  and  server  fault  tolerance  features.  There  are  emerging  global  shared  cache  solutions  but  these  may  require  high  speed  interconnects  to  support  complex  cache  replication  schemes  and  are  still  very  limited  in  their  ability  to  scale  capacity  and  performance.      

• Flash   in   the   storage   array   –   Flash   SSD’s   added   to   traditional   storage   arrays   provide   a  storage  boost  over  HDD’s,  especially   if  auto-­‐tiering   is  operational.  On  the  other  hand  tiering  implies   that   not   all   IO   will   be   treated   to   equal   performance.   And   storage   controllers   not  designed   for   flash   limit   the   performance   potential   of   the   flash   investment.    All   flash   storage   can   provide   a   big   performance   improvement,   but   can   be   very   expensive  unless  it  optimizes  flash  usage  (e.g.  native  de-­‐dupe,  et.al.).  And  storage  networks  may  have  to  be   upgraded   (server   HBA,   switches,   cables,   et.al.)   to   provide   enough   bandwidth   at   lower  latencies,  making  the  total  cost  of  this  option  even  more  expensive.    

• Flash   in   the   network   –   Some   solutions   host   flash   as   a   cache   in   the   storage   network,  effectively   pooling   and   sharing   the   cache   with   the   added   benefit   of   a   non-­‐disruptive  implementation.   Unfortunately,   globally   shared   cache   can’t   be   reserved   or   allocated   to  specific   performance   demanding   applications   which   results   in   unpredictable   performance,  and   like   with   all-­‐flash   arrays,   the   network   can   be   an   expensive   bottleneck   limiting   flash  improvements.  

In   addition   to   delivered   performance   and   cost-­‐efficiency,   other   TCO-­‐type   criteria   should   be  considered   like   total   footprint,   power   and   cooling   implications,   and   the   impact   on   current  infrastructure,  operations  and  processes.   In  particular,   integration  with  and  support   for  hypervisor  storage-­‐related  capabilities   is   critical.  Hypervisor  vendors,   such  as  VMware,  are   increasingly   taking  over  the  storage  management  functions  that  are  typically  found  in  modern  storage  arrays.    Solutions  that   easily   integrate   with   hypervisors   and   simplify   the   daily   life   of   the   virtual   server   admin   are  increasingly  becoming  a  requirement.  

In   the   above   list,   networked   all-­‐flash   arrays  might   come   closest   to   an   ideal   virtual   IO  performance  solution  because  it  essentially  enables  sharing  a  resource  pool  of  flash,  and  with  available  hypervisor  integrations   enabling   allocation   of   those   flash   resources   to   specific  workloads.   But   Ethernet   based  storage   networking   still   constrains   high   performance   IO.  What   is   really   needed   to   boost   the  most  common   existing   virtual   infrastructures   is   a   type   of   networked   flash   storage   that’s   not   only  architected  to  internally  optimize  flash  resources  but  also  optimizes  TCP/iSCSI  traffic  to  fully  deliver  on  the  promise  of  flash  performance  over  standard  Ethernet  networks.  

An  Ideal  Solution  –  Networked  Flash  The   concept   of   enabling   high-­‐performance   IO   for   virtual   environments   by   intelligently   leveraging  networked   flash   over   common   Ethernet   iSCSI   storage   networks   is   exactly   what   Astute   Networks  designed   into   its  ViSX  product   line.  Astute’s  ViSX  solutions  are  architected  to  maximize  the  use  and  performance  capability  of   flash  as  shared  storage  while  optimizing  both  TCP  and   iSCSI   traffic.  They  call  this  “Networked  Flash™”:  

• Provides   shared   high-­‐performance   storage   with   full   data   protection   that   enables  virtualizing  demanding  IO  intensive  applications  

• Leverages  flash  investment  with  an  optimal  design  for  both  performance  and  efficiency    • Implements  without   disruption   to   server   operations,   existing   capacity   storage   or   virtual  

admin  management  processes    

With   an   Astute   Networks   VisX   solution   powered   up   in   the   storage   network,   virtual   server  administrators  treat  it  as  new  “tier  0”  of  storage  managed  the  same  as  storage  from  any  other  existing  

Page 4: PRODUCTPROFILE( ACCELERATING*IO INTENSIVE ...docs.media.bitpipe.com/io_10x/io_109901/item_702090...Copyright *The*TANEJA*Group,*Inc.*2013.All*Rights*Reserved.* 1*of*7* 87Elm*Street,*Suite900

 

Copyright  The  TANEJA  Group,  Inc.  2013.  All  Rights  Reserved.   4  of  7  87  Elm  Street,  Suite  900    Hopkinton,  MA    01748    T:  508.435.2556    F:  508.435.2557   www.tanejagroup.com    

Product  Profile  

array.   Virtual   administrators   simply   migrate   IO   intensive   virtual   machines   from   slower   capacity  storage  onto  the  new  performance  flash  storage  as  needed.  This  has  the  immediate  effect  not  only  of  ramping  up  performance   for   the   re-­‐hosted  applications,   but   also   reduces   contention  on   the   slower  storage  to  the  benefit  of  all  virtually  hosted  applications.  

ASTUTE  NETWORKS  VISX  G4  The  next  generation  Astute  Networks  ViSX  G4   is  designed   from  the  ground  up   to  handle   the  100%  sustained  random  IO  workloads  generated  from  demanding  high  performance  virtual  environments.  Each  ViSX  G4  is  capable  of  delivering  140,000  IOPS  from  its  2U  array  housing  up  to  45.6TB  of  usable  flash.  Significantly,  the  G4  has  built-­‐in  inline  de-­‐dupe  making  that  45.6TB  appear  effectively  as  250TB  of  total  flash,  all  at  full  flash  speeds.  

What  really  sets  the  ViSX  G4  apart  from  other  network  flash  storage  is  its  unique  DataPump  Engine™,  a  custom  ASIC  that  optimizes  network  storage  traffic  over  TCP  and  iSCSI,  and  internally  implements  fast   flash   RAID.   Without   the   DataPump   Engine,   other   networked   flash   solutions   are   naturally  constrained  at  the  network  layer,  making  their  flash  both  less  effective  and  less  cost-­‐efficient.  

In  order  to  fully  address  the  virtual  IO  performance  problem,  the  ViSX  G4  has  been  highly  optimized  for   hypervisor   environments,   integrating   closely  with   VMware   vSphere   (VMware   Ready   Certified),  Microsoft   Hyper-­‐V,   RedHat,   and   Citrix   XenServer   solutions.   However,   any   client   with   iSCSI   can  leverage   the   G4’s   high   performance   storage.   For   VMware   environments,   the   ViSX   G4   is   VMware  Ready   Certified   and   all   vSphere   storage   services   can   take   advantage   of   the   G4’s   high   performance  storage   resources   –   including   iSCSI   multi-­‐pathing,   virtual   disk   snapshots   and   virtual   disk   thin  provisioning,   storage   IO   control   for   performance  management,   and  profile   driven   storage.     The  G4  also  supports  iSCSI  boot  and  is  fully  manageable  through  vCenter.  

Flash  as  Flash  Can  Be  Astute   Networks   aims   to   solve   the   virtual   IO   performance   problem   at   the   lowest   price-­‐per-­‐performance  (e.g.  $/IOP  or  $/VDI  desktop),  as  compared  to  traditional  capacity  storage  aiming  at  the  lowest  price-­‐per-­‐capacity  (e.g.  $/GB).  Part  of  this  cost-­‐efficiency  is  in  optimizing  performance,  part  in  maximizing  effective  capacity.  

As   noted   above,   the   ViSX   G4   has   built-­‐in   de-­‐duplication   which   on   its   dual   8-­‐core   processors   with  256GB  RAM  enables  high  speed  inline  de-­‐dupe  processing.  This  means  that  45.6TB  of  flash  can  look  like  250TB  total  flash  to  clients.  This  native  virtual  capacity  advantage  is  in  addition  to  any  capacities  preserved   through   thin   provisioning   or   other   storage   virtualizing   technology   applied   at   the  hypervisor.   If   higher   capacities   are   desired,   multiple   ViSX   G4’s   can   be   chained   together.   VMware  vSphere   can   easily   “concatenate”   G4   resources   at   the   hypervisor   level   for   increased   IOPs   and  capacities.  ViSX is offered with both high endurance eMLC flash or lower cost, lower endurance MLC flash for customers seeking the lowest cost per GB.  

Each  ViSX  G4   serves  up   to  64  volumes   (128   targets)   to   the  virtual   environment  over  Ethernet   and  iSCSI.   The  G4   has   2x10GbE   and   4x1GbE  ports  which   can   all   be   active   or   used   in   any   combination.  With  the  leverage  of  the  DataPump  Engine,  the  ViSX  G4  delivers  the  performance  equivalent  of  over  800  15k  SAS  disks,  up  to  140,000  sustained  random  IOPS.  With  multiple  ViSX  G4’s  clustered  together,  users  can  realize  well  in  excess  of  1  million  random  IOPS.  

The  Optimizing  DataPump  Engine  Astute’s  DataPump  Engine  is  a  custom  ASIC  design  with  10  multi-­‐threaded  RISC  cores  and  supporting  DMA  engines.  The  DataPump  Engine  is  reported  to  deliver  5-­‐10x  the  performance  of  networking  on  a  

Page 5: PRODUCTPROFILE( ACCELERATING*IO INTENSIVE ...docs.media.bitpipe.com/io_10x/io_109901/item_702090...Copyright *The*TANEJA*Group,*Inc.*2013.All*Rights*Reserved.* 1*of*7* 87Elm*Street,*Suite900

 

Copyright  The  TANEJA  Group,  Inc.  2013.  All  Rights  Reserved.   5  of  7  87  Elm  Street,  Suite  900    Hopkinton,  MA    01748    T:  508.435.2556    F:  508.435.2557   www.tanejagroup.com    

Product  Profile  

traditional  x86  chip.   It  actually  processes  a   few  key   tasks   that   together  provide  wire  speed  storage  access   for   iSCSI   clients,   exposing   the   full   performance   capability   of   flash   despite   it   being   delivered  over  a  network  (for  the  virtualization  enablement  and  data  sharing  benefits).  

1. Accelerates  network  traffic  by  optimizing  the  TCP/IP  stack  2. Virtualizes  the  data  store  messaging  by  offloading  iSCSI,  CRC,  and  SCSI  protocol  processing  3. Optimizes  flash  performance  by  implementing  fast  flash  RAID  

It’s  clear  that  iSCSI  networked  flash  without  this  kind  of  unique  and  deliberate  network  acceleration  will  suffer   latencies  and  bandwidth  constraints   that  would   limit   its   total  effective  performance,  and  increase  its  relative  price-­‐per-­‐performance.  Only  by  optimizing  the  network  part  of  the  IO  chain  can  the  virtual  IO  performance  problem  be  solved  with  shared  flash  storage.  

Highly  Available,  Resilient,  and  Green  The  ViSX  G4  makes  effective  use  of  eMLC  flash,  and  to  ensure  data  integrity  over  time  incorporates  an  Automated  Flash  Life-­‐cycle  Management  process.  The  G4  is  actually  internally  overprovisioned  with  28%  reserved  flash  (not  counted  in  the  45.6TB  usable)  as  part  of   its  data  protection  package.  As an example, each of its 24 hot-swappable 400GB eMLC Flash modules actually has 512GB.  This  extra  capacity  is   used   to   pre-­‐reserve   and   pre-­‐erase   cells   to   speed   up   garbage   collection   and   overall   write  performance.  It  also  significantly  extends  the  life  of  the  Flash  modules.  

For  array  data  protection,  the  internal  controller  supports  flash  RAID  0,  1,  10,  5,  and  6.  In  addition  the  flash   modules   in   the   G4   are   hot   swappable,   as   are   the   redundant   power   supplies,   cooling   fans,  network   and   power   connections.   Astute also offers ViSX Replicator, which supports both synchronous and asynchronous replication for even higher levels of data availability.

And   of   course,   flash   storage   has   an  impressively   green   footprint   compared  to  traditional  HDD  storage,  even  more  so  if   you   compare   the   number   of   spindles  that   would   be   required   to   get   flash  performance   at   the   de-­‐duped   flash  capacity.   Each   ViSX   G4   unit   is   spec’d   at  only  500  watts.  

High   Performance,   Simple  Management    ViSX   arrays   are   manageable   through  Astute’s   FlashWRX   web   browser  interface,   CLI,   or   SOAP/XML   API,   but  since  the  ViSX  G4  is  internally  integrated  with   all   flash   optimizations   and   error  correcting   baked   in   by   default,   there   is  little   need   for   deliberate   management  attention   outside   of   gross   hardware  failures.   Operating   the   ViSX   itself   is  mostly   a   matter   of   powering   up   and  ensuring   network   connectivity   while  data   volume   and   storage   related  

 

Figure  1  –  FlashWRX  Web  Management  interface  showing  ViSX  RAID  configuration  

Page 6: PRODUCTPROFILE( ACCELERATING*IO INTENSIVE ...docs.media.bitpipe.com/io_10x/io_109901/item_702090...Copyright *The*TANEJA*Group,*Inc.*2013.All*Rights*Reserved.* 1*of*7* 87Elm*Street,*Suite900

 

Copyright  The  TANEJA  Group,  Inc.  2013.  All  Rights  Reserved.   6  of  7  87  Elm  Street,  Suite  900    Hopkinton,  MA    01748    T:  508.435.2556    F:  508.435.2557   www.tanejagroup.com    

Product  Profile  

operations   on   the   G4   for   virtual   environments   would   be   performed   from   hypervisor   storage  management  interfaces.  

For   vSphere   administrators,   the   ViSX   G4   comes  with   a   new   VMware   vCenter   plug-­‐in   to   centralize  management   for   storage   operations   and   the   ViSX   appliance   all   within   vCenter.   Like   other   iSCSI  storage  arrays,  3rd  party  storage  applications  work  naturally  on  the  ViSX  storage  volumes,  often  with  significant   performance   improvement.   For   example,   network   backups   can   be   taken   at   high   speed  (and  recoveries  if  ever  needed  would  run  much  faster  too).  

SMASHING  VIRTUAL  IO  BARRIERS  WITH  THE  VISX  G4  Overall,   the   ViSX   G4   promises   to   deliver   great   value   leveraging   its   unique   conjunction   of  virtualization,  flash  and  networked  storage:  

• Delivering  the  full  unconstrained  performance  benefits  of  flash  technology  • Delivering  the  data  protection  benefits  of  shared  storage  • Delivering  the  efficient  management  benefits  of  seamless  hypervisor  integration  

The  ViSX  G4  Flash  Performance  Storage  appliance  presents  a  clear  solution  to  the  virtual  IO  problem.  Because   it  deploys  easily   into  any  current  virtualization  environment,   it  could  support  a  number  of  significant  virtualization  initiatives  including:  

• Virtualizing  IO  intensive  databases  • Accelerating  key  performance  sensitive  applications  like  MS  Exchange  or  Sharepoint  • Increasing  total  vm  density  and  improving  server  infrastructure  utilization  • Reducing  vMotion  churn  • Enabling  successful  VDI  deployments  at  larger  scales  • Supporting  a  shorter  RPO/faster  RTO  service  level  

Is  It  Really  That  Simple?  Implementing  the  ViSX  G4  is  directly  simple.  Rack  the  2U  ViSX  G4  unit(s),  plug  them  in  and  configure  them  to  the  iSCSI  Ethernet  network.  Declare  new  storage  volumes  in  the  hypervisor  and  just  storage  vMotion  hot  vm  datastores  over  into  them  from  their  slower  performance  constrained  locations.  The  ViSX  G4  offers  an  exceptionally  quick  upgrade  to  high  performance  IO  with  no  disruption  of  existing  infrastructure.  It  really  is  true  performance  acceleration  in  a  matter  of  minutes.  

Is  It  Actually  Cost  Effective?  When  comparing  performance  solutions,  especially  with  flash,  be  sure  to  factor  in  just  how  optimally  the  solution  delivers  the  full  promise  of  flash  performance.  For  example,  a  flash  SSD  trapped  behind  a  traditional   HDD   storage   controller   accessed   over   a   slow   network   doesn’t   provide   the   same   local  performance   per   TB   as   server   SSD   flash.   At   the   same   time,   server   flash   isn’t   a   readily   shareable  resource  in  a  virtualization  environment,  limiting  its  total  performance  boosting  contribution.  

The  ViSX  G4  not  only  positions  flash  at  the  best  point  for  virtualization  sharing  and  enablement  –  and  with   the   lowest   disruption   -­‐   it   also   incorporates   several   features   that   taken   together  multiply   the  returns   from   your   investment   in   flash.   As   noted,   in-­‐line   de-­‐duplication   makes   the   flash   go   much  farther  while  the  DataPump  Engine  enables  high  performance  over  common  networking  protocols.  

Bottom   line   is   that   the   ViSX   represents   a   significant   opportunity   to   cost-­‐effectively   accelerate   IO  performance   in   virtual   environments,   and   that   enablement   brings   financial   rewards   beyond   the  investment   in   the  ViSX.  We  believe   that   for   a   $50k   investment   in  ViSX,   one   could  expect   to   readily  

Page 7: PRODUCTPROFILE( ACCELERATING*IO INTENSIVE ...docs.media.bitpipe.com/io_10x/io_109901/item_702090...Copyright *The*TANEJA*Group,*Inc.*2013.All*Rights*Reserved.* 1*of*7* 87Elm*Street,*Suite900

 

Copyright  The  TANEJA  Group,  Inc.  2013.  All  Rights  Reserved.   7  of  7  87  Elm  Street,  Suite  900    Hopkinton,  MA    01748    T:  508.435.2556    F:  508.435.2557   www.tanejagroup.com    

Product  Profile  

increase   vm   density   to   the   point   of   avoiding   $100k   of   server   acquisition.   A   thorough   total   cost  accounting  for  all  the  performance  benefits  should  yield  multiple  ROI  factors.  

TANEJA  GROUP  OPINION  The  virtual  IO  performance  bottleneck  presents  both  a  challenge  and  an  opportunity.  With  the  right  approach  that  leverages  existing  investments  in  virtual  infrastructure  and  staff  expertise,  overcoming  performance  constraints  can  not  only  move  virtualization  initiatives  forward,  but  can  accelerate  them  to  provide  an  ROI  well  past  expectations.  Virtualization  provides  many  benefits,  and  suddenly  being  able  to  virtualize  much  deeper   into  a   formerly  virtualization-­‐resistant  application  portfolio  can  also  energize   related   initiatives   such   as   building   private   cloud   service   offerings.   Newly   high-­‐powered  environments  might  even  foster  ideas  for  innovative  new  applications.  

Key  to  the  ready  adoption  of  the  Astute  Networks  ViSX  G4  approach  is  that  it  will  drop  in  to  whatever  virtual   environment   infrastructure   that   exists   today.   Delivering   non-­‐disruptive   high-­‐performance  flash   solutions   avoids   causing   forklift   upgrades,   ripping  and   replacing,   or  otherwise   re-­‐architecting  productive   environments.   Astute   has   deliberately   aimed   at   networked   flash   as   the   best   way   to  leverage   flash   technology   because   it   puts   in   the   right   place   to   be   most   effectively   shared   while  preserving   current   data   protection   and   storage   management   practices.   With   a   focus   on   tight  hypervisor   integration,   the   ViSX   G4   solves   the   virtual   IO   performance   constraint   with   almost   no  impact  on  the  virtual  administrator.  

Because  of   features   like   inline  de-­‐dupe  and   the   custom  DataPump  Engine,   other  performance   flash  solutions   will   be   hard   pressed   to   match   the   IOPS/$   of   the   ViSX   G4.   For   currently   constrained  environments  the  combination  of  high  performance  with  cost-­‐efficiency  and  easy  adoption  should  be  immediately  attractive.  When  benefits   like   increasing   total   vm  density,  performance   improvements  for   all   vms   (by   removing   contention   for   shared   traditional   storage),   and  new  service  opportunities  are  factored  in,  the  ViSX  G4  should  have  an  obvious  ROI  justification.  Ultimately,  the  simple  fact  that  the   ViSX   G4   can   help   virtual   admins   deliver   a   “better   than   physical”   experience   to   important  applications  is  probably  reason  enough.  

.NOTICE: The information and product recommendations made by Taneja Group are based upon public information and sources and may also include personal opinions both of Taneja Group and others, all of which we believe to be accurate and reliable. However, as market conditions change and not within our control, the information and recommendations are made without warranty of any kind. All product names used and mentioned herein are the trademarks of their respective owners. Taneja Group, Inc. assumes no responsibility or liability for any damages whatsoever (including incidental, consequential or otherwise), caused by your use of, or reliance upon, the information and recommendations presented herein, nor for any inadvertent errors that may appear in this document.