Possible MPU Server Purchases, 2013 - 2016

Assumptions:
  • Machines in service should not be more than six years old.
  • We can use virtual machines for non-critical "slave" roles.
  • Where we can replace server hardware with VMs we should do so, to release the server hardware for use elsewhere.
  • The LCFG master should be relatively young as it is so critically important.
  • The VM servers should be relatively young.
  • Servers are now purchased with 5 year warranty

MPU physical servers in summer 2013

Service Function Machine Age Machine Name
Student SSH 6 kubelik
Staff SSH 5 hogwood
Packages mirrors AFS 5 telford
NX 5 bakerloo
KVM host (users) 4 metropolitan
LCFG master 4 schiff
Offsite KVM host 1 4 piccadilly
Offsite KVM host 2 4 northern
RPM master 4 brendel
Disaster LCFG 3 sauce
SAN storage 3 IBM storage array
KVM host IF 1 1 jubilee
KVM host IF 2 1 hammersmith
Package cache 1 1 hare
Package cache 2 1 wildcat
KVM host AT 0 waterloo
KVM host IF 3 0 oyster

2013-14 Developments:

(Each fresh indentation indicates a dependency on the previous event.)

  • End Packages mirrors AFS as a separate service, folding it into the Services Unit's AFS service.
    • telford is either binned or handed to the Services Unit.

  • Introduce NX 2
    • NX becomes NX 1

  • Buy two large fast servers alef and beth.. (2 of Dell R720 - 6000 each)
    • Offsite KVM host 1 moves to alef.
      • northern is freed up.
        • NX 1 moves to northern.
          • bakerloo is freed up.
            • bakerloo is binned (?)
    • Offsite KVM host 2 moves to beth.
      • piccadilly is freed up.
        • NX 2 starts on piccadilly.

  • Buy two servers gimel and daleth. (2 of Dell R210-II - 1000 each)
    • LCFG Master and RPM Master move to gimel and daleth.
      • schiff and brendel are freed up.
        • Student SSH and Staff SSH move to schiff and brendel.
          • kubelik and hogwood are freed.
            • kubelik and hogwood are binned.

  • Disaster LCFG moves from sauce to juice.
    • sauce becomes the spare DL180.

MPU physical servers in summer 2014

Service Function Machine Age Machine Name
Student SSH 5 schiff
Staff SSH 5 brendel
NX 1 5 northern
NX 2 5 piccadilly
KVM host (users) 5 metropolitan
SAN storage 4 IBM storage array
Disaster LCFG 3 juice
KVM host IF 1 2 jubilee
KVM host IF 2 2 hammersmith
Package cache 1 2 hare
Package cache 2 2 wildcat
KVM host AT 1 waterloo
KVM host IF 3 1 oyster
Offsite KVM host 1 0 alef
Offsite KVM host 2 0 beth
LCFG master 0 gimel
RPM master 0 daleth

2014-15 Developments (and questions):

  • The IBM storage array warranty expires at the end of this period. We want to purchase earlier than 2015-2016, probably around Easter 2015, to ensure service overlap. (Allow 18,000)
    • Buy a new storage array vav
      • Move SAN Storage to vav
        • IBM storage array is released (and unsupported)
    • It may be possible to extend the service life of existing IBM storage array if it continues to prove reliable and a cost-effective maintenance contract is available.

  • Buy a new large server he. (Dell R720 - 6000)
    • KVM host (users) moves to he
      • metropolitan is freed up

  • Will the KVM hosts need more capacity?
    • Let's see what the other units' virtualisation plans look like.
    • Let's see the results of Iain's survey.

MPU physical servers in summer 2015

Service Function Machine Age Machine Name
Student SSH 6 schiff
Staff SSH 6 brendel
NX 1 6 northern
NX 2 6 piccadilly
Disaster LCFG 4 juice
KVM host IF 1 3 jubilee
KVM host IF 2 3 hammersmith
Package cache 1 3 hare
Package cache 2 3 wildcat
KVM host AT 2 waterloo
KVM host IF 3 2 oyster
Offsite KVM host 1 1 alef
Offsite KVM host 2 1 beth
LCFG master 1 gimel
RPM master 1 daleth
KVM host (users) 0 he
SAN storage 0 vav

2015-16 Developments (and questions):

  • Buy two new multi-CPU big-memory machines zayin and het (Dell R520/128GB - 4000 each)
    • EITHER
      • NX 1 moves to zayin
        • northern is freed up
          • northern is binned
      • NX 2 moves to het
        • piccadilly is freed up
          • piccadilly is binned
    • OR
      • KVM host IF 1 moves to zayin
        • NX 1 moves to jubilee
          • northern is freed up
            • northern is binned
      • KVM host IF 2 moves to het
        • NX 2 moves to hammersmith
          • piccadilly is freed up
            • piccadilly is binned

  • Virtualise Student SSH and Staff SSH ?

  • Time to get a new host for Disaster LCFG?
Topic revision: r11 - 09 Dec 2013 - 10:07:44 - AlastairScobie
 
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback
This Wiki uses Cookies