Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 16 Current »

The sizing is referencing Accelario server

CPU, RAM and Network sizing by data size

Given:

Recommended:

 Shape

Data size

Network

CPU

RAM GB

Instance type for AWS:

small

up to 500 GB

1 Gbps

4

8

Not supported

medium

500 GB - 1TB

1 - 5 Gbps

8

16

Not supported

large

1 - 5 TB

5 -10 Gbps

16

32

c5.4xlarge
c5a.4xlarge
c6a.4xlarge
c6i.4xlarge

X large

5 - 10 TB

10 - 20 Gbps

32

64

c5.9xlarge
c5a.8xlarge
c6a.8xlarge
c6i.8xlarge

XX large

10 TB and higher

10 - 40 Gbps

64

128

c5.18xlarge
c5a.16xlarge
c6a.16xlarge
c6i.16xlarge

Boundaries of NFS throughput are defined by 2 parameters:

  1. Disk throughput (throttling) - real speed is usually 10% less than defined due to NFS protocol behavior

  2. Network throughput - real speed is usually 30% less than defined due to TCP protocol implementation

The NFS throughput is defined by lower value either of parameters

Examples

Examples of expected NFS throughput (NFS read/write speed) by different settings:

Given:

Expected:

Network

Disk throughput

IOPS

NFS version

NFS throughput (NFS read/write speed)

1 Gbps

125 MB/s

2000

 3 or higher

100 - 120 MB/s

5 Gbps

600 MB/s

4000

 3 or higher

500 - 600 MB/s

10 Gbps

1000 MB/s

4000

4,1

800 - 1000 MB/s

Equation:

MIN("DISK THROUGHPUT", "NETWORK BANDWIDTH") - 30 PERCENT
e.g
Netwokr = 10 Gbps = 1.25 GB/s
Disk throughput = 1 GB/s
Then expected NFS throughput will be
min(1.25, 1) = 1 GB/s - 30 percent ~ 800 MB/s, 
we can expect sometimes the pick can be up to 1000 MB/s

The system requirements for the NFS server can vary based on several factors such as the number of concurrent sessions and the volume of data that has to be moved. Here are some of the factors to be considered while configuring the NFS server and client to maximize performance.

  • Network traffic must be well-balanced across client networks, and the networks themselves must never be overloaded.

  • Ensure that there is sufficient disk bandwidth and that it is not driven at more than 60 percent of its random I/O capacity to minimize the chances of bottlenecks.

  • Ensure that the percent system time of the CPU does not exceed 50% for optimal NFS performance.

  • Use the formula, Virtual memory = RAM (main memory) + swap space to calculate the memory requirements depending on the type of workload deployed on the server.

  • No labels