Oxford University

OUCS Annual Report

4. Central Servers


The role of OUCS in providing central computing services has been gradually wound down over the years, so that today the orientation of OUCS facilities and services is a far cry from the 1980s "mainframe" days of the ICL 2980, or even of the Vax systems provided in the late 1980s and early 1990s.

The University's distributed computing strategy, which has been in place since the early 1990s, has resulted in the winding down (in 1997) of the central Vax service. Nevertheless, the two central Unix systems, Ermine and Sable, continue to grow in use and importance, though their usage is much less now focussed on logins, interactive sessions and general-purpose Unix computing. Rather, they provide the platform from which a wide range of client-server functions have been launched (eg Web serving, Email serving, FTP serving).

Gradually, these various Unix-based services have been devolved onto dedicated (and much cheaper) stand-alone systems, typically commodity PCs running Linux. Several of these have been described in sections 2 and 3 above.

Ermine & Sable

Notwithstanding these overall strategies, the general usage of Ermine and Sable continues to rise, both for interactive logins and for server-processes [figure 34 & figure 35]. So the goal of winding down these services remains elusive. OUCS continues to monitor the types of activity for which these systems are utilised, and to develop plans for devolving the functions to smaller dedicated systems.

Figure 34 Figure 35

The number of users on Sable dropped from 18,000 to 15,000 in September 1998, after the 4,000 new undergraduates were pre-registered with Herald accounts for email. Thus 25% of the new undergraduates elected to create Sable accounts for themselves. Sable was still heavily overloaded during peak times but the stability problems have somewhat lessened, especially after an operating system patch-kit applied in January 1999 [figure 36].


Sable Availability (Figure 36)

4 Weeks Ending

#

CPU

Availability

Time

# breaks

MTBI

MTTR

Sessions

Hrs

DEC

Overall

Prime shift

Lost

 

 

 

 

 

 

 

 

 

 

 

 

 

09-Aug-98

279104

980

99.9

99.9

99.7

0.8

2

336

0.4

06-Sep-98

255647

905

100.0

100.0

100.0

0.0

0

-

-

04-Oct-98

294697

350

100.0

100.0

100.0

0.0

0

-

-

01-Nov-98

559433

491

99.9

99.9

100.0

0.2

1

672

0.2

29-Nov-98

635628

441

99.9

99.9

100.0

0.2

1

672

0.2

27-Dec-98

397630

374

99.8

99.8

99.3

1.5

5

134

0.3

24-Jan-99

391282

267

99.9

99.7

99.7

1.7

2

335

0.8

21-Feb-99

614022

593

99.8

99.8

99.5

0.7

2

336

0.3

21-Mar-99

596339

559

99.9

99.6

99.4

2.3

3

223

0.8

18-Apr-99

361732

272

99.2

99.2

99.9

5.4

2

333

2.7

16-May-99

618989

423

99.9

99.9

99.5

0.8

3

224

0.3

13-Jun-99

628284

488

99.9

99.9

99.7

0.4

2

336

0.2

11-Jul-99

428119

353

99.9

99.9

100.0

0.3

1

672

0.2

 

 

 

 

 

 

 

 

 

 

Total

6060906

6496

99.8

99.8

99.7

14.3

24

363

0.6



Ermine Availability (Figure 37)

Ermine's user-base has remained fairly steady at about 8,000 users, and has had a fairly uneventful year [figure 37].

4 Weeks Ending

#

CPU

Availability

Time

# breaks

MTBI

MTTR

Sessions

Hrs

DEC

Overall

Prime shift

Lost

 

 

 

 

 

 

 

 

 

 

 

 

 

09-Aug-98

79122

1883

99.9

99.9

100.0

0.2

1

672

0.1

06-Sep-98

68502

1778

100.0

99.4

97.9

4.3

1

668

4.3

04-Oct-98

80687

1461

100.0

100.0

100.0

0.0

0

-

-

01-Nov-98

93169

705

100.0

100.0

100.0

0.0

0

-

-

29-Nov-98

96489

635

100.0

100.0

100.0

0.0

0

-

-

27-Dec-98

79047

970

100.0

100.0

100.0

0.0

0

-

-

24-Jan-99

70345

707

100.0

100.0

100.0

0.0

0

-

-

21-Feb-99

97585

937

99.9

99.9

99.6

0.7

1

671

0.7

21-Mar-99

96572

831

100.0

99.7

99.3

2.2

1

670

2.2

18-Apr-99

72915

542

100.0

100.0

100.0

0.0

0

-

-

16-May-99

96271

631

100.0

100.0

100.0

0.0

0

-

-

13-Jun-99

92400

455

100.0

100.0

100.0

0.0

0

-

-

11-Jul-99

88128

762

99.4

99.4

98.2

3.9

1

668

3.9

 

 

 

 

 

 

 

 

 

 

Total

1111232

12297

99.9

99.9

99.6

11.3

5

1745

2.3


The reliability and availability of these systems (as well as the HFS) has improved, but it still not wholly satisfactory [figure 38].

Figure 38

Special Printing Devices

OUCS operates a number of special printing devices, which are typically more expensive than any one department or unit can afford, and only by aggregating demand across the University are they cost-effective. They include:

Usage on all these devices has been steady, and Operations staff make a point of providing professional quality output.

Computer Room Operations

The computer room is staffed from 08:00 until 22:00 each weekday. Operations staff are responsible for all operational duties associated with the large number of computers and network equipment housed there. This includes the central Unix computers, the HFS and its robot tape library, all the network servers, the servers for the OUCS departmental network (including teaching and public area computers), servers associated with the NSMS service, as well as the large supercomputers housed on behalf of the Materials Department and Oscar, the main OLIS library and other library systems, servers for central administration and the Bioinformatics system.

All this equipment is protected by a large UPS system, and multiple airconditioning systems (for resilience).

In addition to standard operational services, the evening operations staff also provide a degree of help to users in the Work Area. Evening staff consist of one permanent employee plus a collection of casual staff, mostly drawn from the ranks of postgraduates.

[ Table of Contents | Previous section | Next section ]

[ Oxford University | OUCS Home Page | Search OUCS Pages | Feedback ]

Last updated: 29-May-00.
© Oxford University Computing Services.