Docstoc

When running Linux on IBM System z makes sense... makes sense

Document Sample
When running Linux on IBM System z makes sense... makes sense Powered By Docstoc
					When running Linux on IBM System z
         X
 makes sense...
       cents


Velocity Software Inc. / Hill Gang – July 14th

                                Len Diegel
                                Len@velocitysoftware.com
                                L @ l it ft
Agenda

• Today’s “legacy system”! Times have changed.
       h       i         ff i          i    ll i
• How changes in IT are affecting your installation.
   – IT Trends & Directions
   – TCO Considerations
• Measure, Track, and Report - Why bother?
• Results – What should you expect?
• Tools of the trade: The good, the bad, & the sad
Background
                                               p
• This session is based on actual customer experience &
  history as it relates to migrations from x86 (commodity)
  server “farms” to Linux on z.
  The i          d     bl     th          t d l     th
• Th issues and problems they encountered along the way.
• Their efforts to resolve those issues and problems.
             adapted overcame            .
• How they adapted, overcame, and . . survived ! !
                                         survived
Definition: “Legacy System”

 “A legacy system is an old computer system
or application program that continues to be
used because the user (typically an
organization) does not want to replace or
redesign it.”
        en.wikipedia.org/wiki/
   g y                           g
“Legacy can be a burden or heritage
         with dignity”…

          Karl-Heinz Strassemeyer (IBM Corp)
 IT in 2010 has a new set of issues.

Unfortunately, the “legacy systems” of the 90s are
   creating more issues to deal with than the
   mainframe of the 70s.
    A nightmare to manage & administer.
    Limited to horizontal capacity increases. (see below)
    Increasingly expensive to maintain.

  Today’s legacy systems are the byproduct of the
  “add another rack” IT generation of the 90 s . . .
   add          rack                      90’s
Challenges to the IT Professional . .

 Environmentals –
    Space, Heat, Power, “Green” Efforts
 Administrative
    Maintenance, Upgrades, Charge Back, Capacity
     Planning, Performance DR       Security
     Planning Performance, DR, Data Security, and more
 Internal IT Pressures
    Migrations, Consolidations, Compliance Issues,
       g       ,               ,    p             ,
     Maximizing Resources, Flexible Capacity
 Industry Strategies & Directions
    Cloud / Storm / SUN-down / Sky-BLUE/ Cloudburst
 What everyone wants from IT . .

Secure, high speed access to
           time
  data any time, from any
device, and from any where.
      f
 Even from a dive boat in
Turks Caicos working with
Neale on breakfast for this
         event . .
What IT Execs are thinking:
Response from 100 IT executives at companies using Linux operating
system on IBM mainframes.
                                                  IBM s
93% of respondents projected that their use of IBM's IFL (Integrated
Facility for Linux) would increase or remain steady.
The two main reasons:
1.   Advantage of computing capacity available on their mainframe's
     and IFLs.
2
2.   Li
     Linux on th mainframe would b more cost-effective th other
                the i f       ld be        t
                                        cost- ff ti than th
     platforms.
     platforms.



          * http://finance.yahoo.com/news/Survey-Predicts-Continued-prnews-15547427.html?.v=1
     What is the greatest problem with your
     primary data center ?

“Power and Cooling will be a top 3 issue with all CIOs in the next 6
              “P
to 12 months.” Michael Bell – Gartner Group
     Spending                                                             Installed Base
      (US$B)                                                                 (M Units)

    $300                                                                         50
                            Power and cooling costs x8
                            Server mgmt and admin costs x4                        45
    $250
                            New server spending                                   40
                                                                                  35
    $200
                                                                                 30
    $150                                                                         25
                                                                                 20
    $100
                                                                                 15
                                                                                 10
     $50
                                                                                 5
       $0                                                                        0


  Source: IDC, Virtualization 2.0: The Next Phase in Customer Adoption,
           Doc #204904, Dec 2006
IBM’s Global CEO Study
                                                geos
• Over 1500 CEOs were interviewed across all geos.
• Most intensive study conducted in this space.
                                        p    y
• More than 80% see an increase in complexity as their
  number one IT issue. Fewer than 50% say they are ready.
• Challenges differ from region to region:
          Government R l i
   – US / G            Regulations
   – Japan / Power shift to emerging markets
   – China / “Thinking Global” as opposed to closed society.
• Conclusion: “No matter what the industry, those that can
  manage and react to complexity will have the advantage
                  can’t
  over those that can t.”
IBM’s Mainframe Cloud Agenda

• Focus on Mainframe Strengths
   – Extend System z’s massive virtualization capabilities
• Leverage System z efficiencies
• Offer System z (mainframe) customers value beyond that
         i th     l tf         l ti      dd i
  seen via other platforms / solutions and doing so
  transparently to the broader community. (90 Day Free
  Trials on IFLs)


                   Keeping                     clouds”
                  “Keeping your head above the clouds STG Tech Conference 09
                  IBM’s Cloud Website: http://www.ibm.com/ibm/cloud/
 System z for Cloud


 As                            cost-          2009
“As the IT world prepares for cost-cutting in 2009, server
virtualization seems like one of the few technologies bound for
continued growth and success. . . . The IBM System z may be the
b enterprise-class virtualization platform available today . . .”
             i
best enterprise- l     i    li i      l f        il bl d        ”



 Enterprise Strategy Group – January 2009 Report “IBM System z: The Enterprise Server
                                Virtualization Platform.”
 IBM’s Project “BIG Green”
• Consolidation of several thousand Linux servers to 30 IBM
  z10s. (may end up being fewer than 20..)
• 80% Reduction in Energygy
• 85% Reduction in Floor Space
   – 1,600 for z10s vs 11,600 square feet for Distributed
• Categories as a % of Gross Savings:

                                             Facitilites
                                             Connect
                                             Labor
                                             Software
                                             HW Maint
                                             HW Purch
    Internal IT Study for TCO

    •   Determine consolidation ratio       Server Migration : x86 vs. zLinux
        for given workloads (Linux,
        Windows, etc.)                       0
                                            20
    •   Estimate annual cost to operate
        virtualized servers for 3-5         15
        years.
    •   C
        Compare this to stand-alone
                   hi         d l           10                         x86/other
        provisioning or public Cloud                                   Virtual/IFL
                                             5
        offerings like Amazon
    •   N< PPR*PUR*CPF supports              0
        that the more servers you can            1st 2nd 3rd 4th
        consolidate to virtual, the lower        Qtr Qtr Qtr Qtr
        your cost.                           Consolidation Math

N= # of virtual servers / Proc Perf Ratio / Proc Util Ratio / Cores per Frame
Response Time Comparison
                           100
                            90
                            80
                     nds




                            70
   sponse Tim in Secon




                            60
                            50                                zVM
            me




                            40                                VMware
                            30
                            20
                            10
 Res




                             0
                                 10 20 30 40 50 60 70 80 90

                                 Number f Virtual M hi
                                 N b of Vi t l Machines
Throughput Comparison
             160
             140
             120
             100
              80                                zVM
              60                                VMware
   Tot TPS




              40
              20
     tal




               0
                   10 20 30 40 50 60 70 80 90

                   Number f Virtual M hi
                   N b of Vi t l Machines
But keep in mind, not everything was meant to run on z . .
Cost Per Image for Linux Servers

                               400
 Cost Per Worklo Image in $k




                               350
                       i




                               300
                                                    S/A Servers
                               250                  Public Virt.
               oad




                               200                  Private Vi t
                                                    P i t Virt.
                               150                  zLinux Add
                                                    zLinux New
                               100
      P




                                50
                                 0

                                     IT Solutions
Detailed Cost Breakdown - 5 yr TCO

Platform            100 x3250           100 Amazon          VMware & Five       7 IFLs on z10
                                        Large EC2           x3950 – 8ways

Hardware            $5 000 000
                    $5,000,000          $2 800 000
                                        $2,800,000          $1,080,000
                                                            $1 080 000          $2 030 000
                                                                                $2,030,000


Software            $21,490,000         $20,840,000         $2,380,000          $1,520,000


Facilities and      $11,020,000         $4,020,000          $2,760,000          $1,210,000
Administration                          (admin only)

Total Cost for      $37,510,000         $27,740,000         $6,220,000          $4,780,000
100 Images

Cost per image      $375,100            $277,400            $62,200             $47,600


IBM SWG White Paper / April 2009 & IBM Internal Study of 4000 xSeries Distributed Servers
System z Rising

• More than 1800 customers      2000
  are using System z for        1800
  Li
  Linux.                        1600
                                1400
• In 09, IBM added 54 new
                                1200
   mainframe
  “mainframe” customers         1000
  for the purpose of Linux       800
• Customer experience and        600
  results support the current
       l           h             400
                                 200
  strategy to migrate Linux
                                   0
   pp               y
  applications to System z.            2002   2004   2006   2008
                                 The number of Linux IFLs sold/year
                                  doubled between 2006 and 2008.
Unique Value of the IFLs

• Run at full MIPS even on
  sub-capacity models           1.72     391%
                                 4.4    1157%      IFL Value Increase
• If/when you upgrade, the
                                                                       1100% over G5
  IFLs move up in power at             12
                                       10
                                                                       55% over z9BC


  no additional charge.                 8
                                        6
• IFL’s have increased in               4                                              IFL Capacity
                                        2                                              %IFL Increase
  power by roughly 55%                  0



                                            G6




                                            z9 0
                                            z8
                                            z9 0
                                            z8 0
                                            z9 0

                                            z1
                                             9
          h      i
  over each previous
                                              6
                                               0
                                               0
                                               0
                                               0
                                               9
                                               9
                                               9

                                               0
                                                Machine Type
  generation or 1100% since                       IFL Value Increase


  G5.
           Note: IBM delivered more IFL MIPS in 1Q10 than in 2009.
Horizontal vs. Vertical Scale (x86 vs. z)
What Horizontal vs. Vertical Scale means to
you (x86 vs. z)

• How about adding a second engine to a car at the flip of a
  switch . . . without stopping.
• Increasing horse power from 180 HP to 450 HP at the flip
  of a switch . . . without stopping.
  Or,
• Or being able to do both . . . on the fly . . . without
  stopping.
• System z allows you to increase your capacity either
  horizontally or vertically and it’s the only platform to
  achieve this flexibility.
But don’t get yourself in a “hole” . .



                          Some applications run
                          better on z than others.
                              ’        d d
                            It’s a good idea to
                              understand the
                                ff
                              difference when
                          migrating to Linux on z.
                   What applications make a best fit?
WebSphere MQ Series
 W bS h       S i
DB2 Connect
CICS Transaction Gateway , IMS Connect for Java
SAP
                                                    High Availability,
WebSphere and JAVA applications development           h /        d dh
                                                    High I/O Bandwidth,
WebSphere Application Server (WAS), Portal         Security and DR Requirements
Domino                                             Simplifying the Support Model
Network I f
 N t           t   t     FTP, NFS      t
       k Infrastructure, FTP NFS, DNS etc..,        Shortening end-to-end Path Reqs
                                                               end-to-
Oracle Database                                    Reduction in Network Traffic
Applications requiring top end disaster recovery   Collocation of Applications
model
ComServer and Communications Controller for        Consolidation of diverse loads
Linux
Virtualization and Security Services
InfoSphere/Cognos
Communicate Pro (VoIP)                             “Selecting Projects for System z”
                                                         Bill Reeder, IBM Corp.
     What applications make a good fit?

UDB (DB2)                    Evaluate Server Choices
Informix, (IDS)              Correct Application Availability
Apache web serving           Supporting Applications
                              S      ti A li ti
                              Political Issues within the Org.
SAMBA                        Potential Porting Issues
TIM/TAM (LDAP Services)
            (             )
TSM
Existing Linux Workloads
   What applications are difficult to move?

ISV and IBM applications that have not yet ported their
application to run on Linux on System z
Applications that by design run at VERY High sustained
utilization which I will define here as >95%.
Stand-Alone single applications as the only Linux for System z
applications
Applications that are to internally sensitive to try and move
This URL is a link to the IBM software running and supported to
run under Linux on System z.
http://www-1.ibm.com/servers/eserver/zseries/os/linux/software.html
http //    1 ibm com/ser ers/eser er/ series/os/lin /soft are html
What workloads are moving to IFLs?




                          IBM Corp. 2010
              Measure Track
Why bother to Measure, Track,
and Report ?
      p
Len’s Travel Tips and other ways to reduce TCO . .
Len’s Travel Tips (continued)
  The question becomes . . . . . .
     When you hit the next “bump” on the road to better
                       performance,



do you keep buying new tires, or do you get the front end
aligned ? ? ?
Responsible IT professionals should always consider

• Performance Management to ensure service levels
  are met.
• Capacity Planning to ensure future needs are met.
    p
• Operational Alerts that detect issues such as
  looping, exceeding disk capacity, etc., and doing
  so for hundreds/thousands of servers concurrently.
• Charge back and accounting information to allow
  your business to charge for resources consumed.
Real Life Situations
• Financial in NY Area - Crit Sit / 2 months / SRM parms
  wrong – SLES8 server wouldn’t initialize
• State of Montana - Crit Sit / 1 month / Linux Server
  Looping, 1 IFL offline. POC Failed.
• University Crit Sit - Response Time issues.
• Financial in DFW Area - System spikes every 5 minutes .
  . Wrong choice of monitors . .
                              y              g
• Industrial – Decided to buy 20G of Storage rather than use
  best practices and set the correct parms..
• Industrial #2 – “We keep adding IFLs.” (now leaving the
              )
  platform . .)
Large Insurance Company
• Pain Points (history) –
   – Too many servers
     Space, Power,
   – Space Power Cooling
   – Inability to Allocate Resources Where Needed
   – New Apps - Time/Expense of Basic Trials and Testing
• Today – (strategy deployed in less than 4 months)
   – +700 Servers on 2 Boxes / Dynamic Capacity Options
   – 100,000 Active Users
   – Zero Production Outages Since 2005
     $3M/Year in Savings
   – $3M/Y i S i
               Some installations are smaller:

•   One CEC (system)
•   Two IFLs (one production, one test)
•   20 Linux Servers replaced, more planned
•   Proof of Concept – 90 Days (free trial – including zVPS)
•   Savings – space, power, admin, etc.
Typical Quotes
 •   "My TSM debugging could not have been done without Velocity.
     IBM s
     IBM’s DB2 Level 2 support was wondering where I got some of the
     info ;-)” …. and that level of detail…

 •    Using                                                       IFL.
     “Using any other monitor would cost me at least 5% of my IFL That
     means one (1) IFL for every 20 that I install just to monitor my
     system.”

 •   “They gave me their monitor for free.”
      –   What about the drag on your system?
      –   What b t d di ti              th     Tb t f t           f th  ft   ?
          Wh t about dedicating more than a Tbyte of storage for the software?
      –   What did it cost you in time to install? (Did it install?)
      –   What does it give you for data, health checks, support, etc…?
      –   Believe me, IT WASN’T “Free”.
          B li                       “F ”
Or, one or more of the following:

•   “Time to add another IFL.”
•   “It’s just the way Linux runs on System z.”
•   “You might consider using System p instead.”
•   “You might consider leaving that appl. on the x86.”
•   “This was your idea. You might consider a new job.”
•   “I told you this would never work on a mainframe.”
•   “DRT…. Dead Right There ”
    “DRT                   There.”
However, accomplishing the basics does
create certain challenges . . As y move
                      g          you
forward, what should you expect?
It’s hardly likely that you would trust your
 y        p f                        g
system’s performance to an astrologist . . .
Prerequisites

• A single suite of products that can deliver whatever is
  necessary to manage performance on z/VM and zLinux.
   – Keeps cost to a minimum.
   – Simplifies service, upgrades, education, etc.
     Single     d i t f
   – Si l vendor interface
• A vendor with a history for delivering best-of-breed
  p               pp
  products and support.
• A cross industry install base with references in all major
  geographies.
   Velocity Software Inc.
                               IBMers
• Founded in 1988 by former IBMers.
• Velocity products are used by hundreds of installations
  worldwide; many are IBM showcase customers.
• IBM uses Velocity’s products internally and makes frequent
  references to our products in their Redbooks.
• We have a working relationship with IBM System z labs in
  Endicott(z/VM), Poughkeepsie(zOS), and Boeblingen(zVSE).
• Velocity has participated in every VM Early Support Program
  since 1988 and has been a PartnerWorld member since 89. 89
• Velocity presents research at many conferences including:
   – SHARE, GSE, CMG, WAVV, and Local Linux User Groups
Velocity Software Delivers :
      l
• Total Data Accuracy
   – When necessary, identifies, prorates and corrects inaccurate data.
• Complete Data Collection
   – Variety of Platforms (NT, Linux, SUN, HP, etc)
   – Standard Interfaces (SNMP and NETSNMP)
     Concurrently across 100s or 1000s of servers.
   – C          l         100    1000 f
• A 100% Data Capture Ratio
                        y         y              p
   – Best in the industry . . . Why measure incomplete data . . ?
• Minimum impact on processors by monitoring agents.
   – Velocity’s target is .1% or less of ONE processor
Not All Performance and Systems Management Tools are Created Equal . . . .

z/VM – Linux Requirement Velocity Software Product   All Other Vendors

Support All Linux                    YES                        NO
Charge Back Accounting               YES                        NO
Capacity Planning Info.              YES                        NO
Low Cost Operation                   YES                        NO
Low O h d A
L Overhead Agents                    YES                        NO
Performance Serv. (zTUNE)            YES                        NO
Performance Education                YES                        NO
Included in IBM Redbooks             YES                        NO
Standard Open Linux Agent            YES                        NO
Network Perf. Monitoring             YES                        NO
Accounting – VM Level                YES                        NO
Accounting – Linux Process           YES                        NO
Accounting – Linux Applic.           YES                        NO
MXG and MICS Interfaces              YES                        NO
“Trying to determine performance issues without
Velocity’s suite, is like asking a doctor to give you an
       y        ,              g             g y
examination while wearing your clothes.”

And, the results are pretty much the same.

                                   D.O.A.
                                   DOA




                                  Quote from an IBM D.E. at WAVV
                                                2010
zVPS Overview

• zVPS is Velocity’s Performance Suite. It is a
  collection of several components including:
                           p               g
   – zMON
   – zVWS
   – zVIEW
   – zPRO (optional)
   – zTUNE (optional)
                     zVPS Overview (cont)
• Provides real time metrics on all facets of server
  performance including z/VM, Linux, and other
  distributed         (SUN Windows, etc.)
  di ib d servers (SUN, Wi d                )
• Full function capacity planning and interfaces to
  planning tools such as MXG and MICS
• Complete and accurate charge back accounting
                          pp
  information for Linux applications and z/VM.
• Operational alerts for up to 1000s of virtual
  servers and immediate detection of performance
  issues.
                   zVPS Example:
                            IBM System z


  zVPS can monitor                     z/VM
   and measure the
 activity of Linux on                         Distributed Servers (x86)
 System z in addition    Linux (IFL)    &
                                                       SUN
to distributed servers
  like SUN Solaris,                    zVPS           Linux
Linux, HP, Windows,
      and more.,,
                                                     Windows

Net: You can determine the level of use for            other
       any server in your network.
Data can be used for Trend Analyses

                         l                 h
              IFL CPU Utilization – 12 Months
  100
   90
   80
   70
   60
                                                  Prod 1
   50
                                                  Prod 2
   40
                                                  Test 1
   30
   20
   10
    0
        Jan      Mar    May    Jul    Sep   Nov
Hmm, CPU spikes every 10 minutes
Which user class is causing the problem?




           In this case it’s one of the Linux Servers.
Drill further -> guest Suselnx2
Drill more…
Auto Arrange
      fy      ff     gp        (          )
Identify the offending process (“smallstr”)




             Unique Capabilities of Velocity Software
zPRO
Cloud Enablement Tool for Systems Administrators
  Cloning f Li       S         S l i VSE
• Cl i for Linux Servers, Solaris, VSE, CMS
  users and more.
• Provides administrators the ability to define
  servers based on “golden images”.
• Manages z/VM security (RACF)
• Provides an easy-to-use, intuitive front-end to
  Dirmaint via a Tab approach and syntax checking
  for defining directory entries.
• Documentation / Command search capabilities
zTUNE

• Configuration analysis to ensure best possible
  configuration for performance.
       g            p
• Dozens of Automated “health checks” for both
  Linux and z/VM.
• Analyses of your Capacity and Performance data.
                          p
• Recommendations to improve p  performance and to
  optimize resources whenever it’s requested.

Velocity works as an extension of the local team.
 Summary
• We are committed to providing the best z/VM performance
  products available.
• Our zVPS Product Suite is a complete set of Performance and
  Systems Management tools designed exclusively for z/VM
  and Linux on IBM System z
• We provide excellent product support services, and education.
• Our install base is both global and diverse.
     l i S f          i h          i d l d i hi            by
• Velocity Software is the recognized leader in this space b
  most of IBM’s z/VM – Linux “showcase” accounts.

      Proven Performance from Velocity Software
The next step is yours . . .

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:3
posted:2/7/2012
language:English
pages:62