Compute Classes

Packet names its servers for clarity, leveraging a formula that includes use case, generation, architecture and size.

As our server lineup grows, it has become harder to scale the very simplistic naming convention we started with (e.g. Type 1, Type 2, etc).

While we deeply value simplicity, we place a premium on clarity.  As such, to help users identify and leverage the right configurations more easily, we are introducing Classes in February 2018.

What are Classes?

Classes are groupings derived from common use cases.  If you're an AWS user, this will seem pretty familiar.  Packet organizes servers into the following classes:

  • Tiny (t) - Smaller instances, aimed at dev/test, controller nodes, etc.
  • Compute (c) - Compute focused, with a modest RAM footprint.
  • GPU (g) - GPU focused with  generous RAM 
  • Memory (m) - Memory heavy, with a generous RAM to core ratio.
  • Network (n) - Focused on network use cases, such as ingress / load balancing.
  • Storage (s) - Scale out boxes for storage scenarios.
  • Accelerator (x) - FPGA and other specialty accelerator focused servers.

Server Naming Convention

All instances have a name that follows a formula to help you understand its purpose and broad specifications:  class + generation + size + architecture (optional)

By piecing these features together, we arrive at a name, such as c1.small.x86 - indicating a small compute class server, in its 1st generation with an x86 processor.  Since most of our systems are x86 currently, we'll provide nicknames so you can use "c1.small" for shorthand.

What are Generations?

Generations are updated when we significantly upgrade or change the components.  

This is usually around a major processor refresh, but often also relates to the underlying hardware configuration / chassis.  As such, a "1st" generation system is not necessarily outdated, it is simply our first iteration of a particular configuration.

What About Previous Server Names and API Calls?

Packet previously labeled servers using Types, such as "Type 1" which was represented in the API as "baremetal_1".  

These will still work - we have simply made an alias in which baremetal_1 translates to its new name:  "c1.small.x86".  

However, new instances (such as our AMD configuration introduced in February, 2018) can only be accessed by the new class-based name or nickname.  In this case: "c2.medium.x86" or just "c2.medium" for short. 

List of All Servers / Names

The following represents our current lineup of servers, publicly available in February 2018.  Previous names are mentioned in parentheses.

Tiny (t)

  • t1.small 

Compute (c)

  • c1.small 
  • c2.medium 
  • c1.large.arm 
  • c1.xlarge 
  • c1.large.arm.xda

GPU (g) 

  • g2.large

Memory (m)

  • m1.xlarge 
  • m2.xlarge 

Network (n)

  • There are no 'n' class servers yet.

Storage (s)

  • s1.large 

Accelerator (x)

  • x1.small