The Nectar Research Cloud is used for a very broad spectrum of workloads, so we offer a range of flavor classes. A flavor defines the size of compute resources (number of virtual CPUs, memory and storage capacity) that can be assigned automatically to virtual machine instances in a cloud configuration.

Nectar flavors make it easier to support different user requirements and enable efficient use of the Nectar Cloud infrastructure. 

Each flavor has an associated "cost" in Nectar Service Units (SU) per unit of time. For convenience, these are shown in the tables below as SU per hour and SU per year. The flavor costs correlate to the compute resources available to the flavor; e.g. the number and priority of the virtual CPUs and the amount of RAM. Other cost adjustments are applied for various reasons. For example, deprecated flavors have a relatively higher SU cost to encourage users to stop using them. 

The SU costs for each flavor are subject to change.

The following standard flavor classes are offered on the Nectar Research Cloud:

  • Tiny (t3)
  • Balanced (m3)
  • RAM Optimised (r3)
  • CPU Optimised (c3)
  • Preemptible (p3) (Coming soon)
  • GPU Visualisation (g1) (reserved instances only)
  • GPU Compute (g2) (reserved instances only)
  • Huge RAM (h4) (reserved instances only)
  • Huge RAM (h3) 

General Availability flavors

The t3, m3, r3 and c3 flavors are available for use by all projects.

Tiny (t3)

The Tiny (t3) flavors are a great choice for testing and prototyping or even just simple workloads with 1GB RAM per virtual CPU core available in 1, 2 and 4-core, with smaller 10GB root disks.

Name

VCPUs

Memory (GB)

Memory (MB)

Disk (GB)

SU per hour

SU per year

t3.xsmall

1

1

1024

10

0.014

125

t3.small

2

2

2048

10

0.029

250

t3.medium

4

4

4096

10

0.057

500

Balanced (m3)

The Balanced (m3) flavors are suited to most workloads. Balanced (m3) flavors offer 2GB of RAM per VCPU.  This ratio suits typical applications, and it allows efficient use of Nectar infrastructure. (If your applications need more memory, you can use the RAM optimised flavors; see below). All Standard (m3) have a 30GB root disk.  The largest one provides 32 virtual CPUs and 64GB RAM.

Name

VCPUs

Memory (GB)

Memory (MB)

Disk (GB)

SU per hour

SU per year

m3.xsmall

1

2

2048

30

0.029

250

m3.small

2

4

4096

30

0.057

500

m3.medium

4

8

8192

30

0.114

1,000

m3.large

8

16

16384

30

0.228

2,000

m3.xlarge

16

32

32768

30

0.457

4,000

m3.xxlarge

32

64

65536

30

0.913

8,000

RAM Optimised (r3)

For workloads that require more memory, the RAM Optimised (r3) flavors are recommended. They provide a ratio of 4GB of RAM for each VCPU.  The root disk size is 30GB.  The largest r3 flavor provides 32 virtual CPUs and 128GB RAM.

Name

VCPUs

Memory (GB)

Memory (MB)

Disk (GB)

SU per hour

SU per year

r3.xsmall

1

4

4096

30

0.052

457

r3.small

2

8

8192

30

0.104

913

r3.medium

4

16

16384

30

0.209

1,827

r3.large

8

32

32768

30

0.417

3,654

r3.xlarge

16

64

65536

30

0.834

7,308

r3.xxlarge

32

128

131072

30

1.688

14,616

CPU Optimised (c3)

For workloads that use more CPU cycles there are the CPU Optimised (c3) flavors. These flavors are best suited to compute-intensive workloads.  The CPU Optimised (c3) flavors go up to a maximum size of 32 virtual CPUs and 64GB RAM.  The VCPUs are given a higher priority for CPU cycles on the physical server relative to other flavors.

Name

VCPUs

Memory (GB)

Memory (MB)

Disk (GB)

SU per hour

SU per year

c3.xsmall

1

2

2048

30

0.043

380

c3.small

2

4

4096

30

0.087

760

c3.medium

4

8

8192

30

0.173

1,519

c3.large

8

16

16384

30

0.347

3,039

c3.xlarge

16

32

32768

30

0.694

6,077

c3.xxlarge

32

64

65536

30

1.388

12,155


Preemptible  (p3)

Preemptible instances or elastic compute enable access to additional compute resources for a short timed period at a low Service Unit (SU) cost. Preemptible instances will be a quarter (¼) of the SU cost of comparative m3 flavors.

All preemptible instances will be killed after a 24 hour period, instances can be killed before 24 hours depending on current resource restraints.


Name

VCPUs

Memory (GB)

Memory (MB)

Disk (GB)

SU per hour

SU per year

p3.xsmall

1

2

2048

30

0.007

63

p3.small

2

4

4096

30

0.014

125

p3.medium

4

8

8192

30

0.029

250

p3.large

8

16

16384

30

0.057

500

p3.xlarge

12

24

24576

30

0.086

750

p3.xxlarge

16

32

32768

30

0.114

1000

p3.3xlarge

32

64

65536

30

0.228

2000




Reserved Flavors

For more information please see GPU flavours & reservation system.


Compute GPU (g2)

g2 instances come with Nvidia A100 GPUs


Name

vGPU RAM


VCPUs

Memory (GB)

Memory (MB)

Root Disk (GB)

Ephemeral Disk (GB)

SU per hour

SU per year

g2.xsmall

812

24

24576

30

200

0.719

6,300

g2.small

1015

30

30720

30

250

0.899

7,875

g2.medium

1624

48

49152

30

400

1.438

12,601

g2.large

2030

60

61440

30

500

1.798

15,751

g2.xlarge

4060

120

122880

30

1000

3.596

31,501


Visualisation GPU (g1)

g1 instances come with Nvidia A40 GPUs


Name

vGPU RAM


VCPUs


Memory (GB)

Memory (MB)

Root Disk (GB)

Ephemeral Disk (GB)


SU per hour

SU per year

g1.small

8
16

32

32768

30

300

0.365

3,200

g1.medium

1230

60

61400

30

500

0.685

6,000

g1.large

2460

120

122880

30

1000

1.370

12,000


Huge RAM (h4)

h4 instances are available in larger RAM sizes than h3 and also come with fast ephemeral disk


Name

VCPUs


Memory (GB)

Memory (MB)

Root Disk (GB)

Ephemeral Disk (GB)


SU per hour

SU per year

h4.xsmall

24

180

184320

30

250

4.677

40,969

h4.small

32

240

245760

30

500

6.236

54,625

h4.medium

48

360

368640

30

1000

9.354

81,938

h4.large

64

480

491520

30

2000

12.472

109,251

h4.xlarge

128

960

983040

30

2000

24.943

218,502


Restricted Availability

The following flavors are restricted in use.  Access to these flavors needs to be requested through:

  • The allocation system for h3 flavors
  • Via a support request for the private flavors
  • The allocation system, and then the GPU reservation system for GPU system flavours

Huge RAM (h3)  

The Huge RAM flavors are available for applications and workloads that require a very large amount of RAM. The Huge RAM (h) flavors offer 7.5GB of RAM per VCPU core and start at 180GB RAM. The RAM consumed by the Huge RAM flavors is a very limited resource.  Users will need to provide a strong technical justification on the allocation request form to be granted access to these flavors.

Name

VCPUs

Memory (GB)

Memory (MB)

Disk (GB)

RAM per VCPU

SU per hour

SU per year

h3.large

24

160

184320

30

7.5

2.598

22,761

h3.xlarge

32

240

245760

30

7.5

3.469

30,347

h3.xxlarge

48

360

368640

30

7.5

5.196

45,521


Private flavors

Nectar nodes may offer additional flavors to selected users. These private flavors and their details (including SU costs) are Nectar node specific. The flavors are frequently associated with hardware whose purchase was funded by or for a specific project. Access to private flavors may be requested via a Nectar Support ticket.