AWS: On Demand Capacity Reservation (ODCR) and preventing Insufficient Capacity Error (ICE)

 What is ODCR, and why does it matter?

On-Demand Capacity Reservation is one of the ways how we can reserve capacity on AWS. This feature helped to reserved capacity especially the large EC2 instance or special instances like (P series) that is very little in the Availability Zones (AZ). The service will not charge you if you have an EC2 running 24 by 7. It will ensure when you need to start/stop your EC2, the same instance family/types for that particular AZ are reserved for you. It is not commitment-based, like a Reserved Instance or Saving Plan, where you need a minimum of 1 year. ODCR is only activated once you enable it in the AWS Console. You can cancel it anytime if you do not needed.

What is an Insufficient Capacity Error (ICE)?

ICE is an exception where there are not enough EC2 instance families available in a particular Region or AZ. This is not common in AWS, but it has happened before, especially for large EC2 like 32x or special EC2 like P series. 

How does ODCR help to prevent ICE?

ODCR signals to AWS that you needed the instance in that particular AZ, so you reserved it using ODCR. Unfortunately, ODCR is not foolproof; this means ODCR only happens if there is an available Instance for that AZ; you can ODCR it. If there is no instance, ODCR won't work. For example, if you ODCR the EC2 today because you needed it but so happens no more instances are available, your ODCR will fail (https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/capacity-reservations-using.html). So always ODCR your instance early.

The strategy for better use of ODCR:

1. ODCR critical application: You do not need ODCR for all your systems; identify mission-critical applications for your business and ODCR them.

2. ODCR-only baseline EC2: This means you only need to ensure that the minimum EC2 supporting your application is protected. If you have Auto-scaling enabled, the new EC2 provision by Auto-Scaling does not need ODCR.

3. ODCR critical EC2: You can identify critical workloads, such as Databases, self-managed Active Directory, or software licenses, that are tied to your CPU core. 

What happens if you get ICE?

ICE can happen, and what you can do is follow the following strategy:

1. Provision your EC2 in another Availability Zone. AWS has a minimum of 3 AZs per region, but you can try other availability zones for your desired instance.

2. You can provision your EC2 to use other Instance families or types first, wait 20-30 mins and retry back your desired Instance Type\Family. AWS will try to create the instances finished by other customers, and it will take a few hours to completely provision the new instance type.

ODCR is not a bad feature; it is useful if you know how to use it better. Charges only happen when you ODCR the instances, and your EC2 is NOT running. Running EC2 won't get additional charges for ODCR but rather normal running costs. If you shut down your EC2, the charges for ODCR will be the same as those for your running EC2.


Comments

Post a Comment

Popular posts from this blog

SharePoint 2013: Distributed Cache Services

Image
What is Distributed Cache Services(DCS)? - It is a Platform Service, it runs at the Farm Level.   -- You can decide to run it on all servers in the farm or you can use it to run on a 1 or a couple of servers.   -- Suggest to have dedicate servers for DCS. - It provides in-memory cache for the entire farm. - It is based on Window Server App Fabric Distributed Caching. - It uses to cache Authentication Token(to avoid multiple authentication processes) and MySite Social Feeds. - DCS enable by default and allocate memory based on total memory on your server.   -- You should at least dedicate 50% of RAM or more just to the DCS service for overhead. What is it NOT : - It do not uses Window Azure Cache services and thus it cache on local server (no cloud). - It uses Window Server App Fabric version that comes together with the installer: You are not recommended download any version from Microsoft and upgrade it independently.  - You are not recommended go to the "C
Read more

SharePoint 2013 for Developer: Baby Journey (1)

Introduction to SharePoint 2013 for developer (1): Introduction a. SandBox is not the way to go for SharePoint 2013 or SharePoint Online 2013  SharePoint 2013 App Model is the replacement for SandBox but SandBox still exists in SPS2013 for backward compatibility purpose. SharePoint 2013 Farm\SandBox Solution :  -CAB File with a .wsp extension -Container of files distributed as a unit (no change from previous version) -Contains manifest with instruction for installer (no change from previous version) -Use powershell to deployed  Good News : - Nothing changes on Farm\SandBox solution: so 0 relearning - Bad News:  - Because nothing change, 0 enhancement to that as well.   File system in SharePoint 2013  - same path as SharePoint 2010 or MOSS(2007) -c:\program files\ common files\ microsoft shared\web server extension\ NOTE: it contains 2 folders (14 and 15) - 14 : Is for 2010 and 15 is for 2013 WHY? * SharePoint 2013 have 2 UI mode ** It is for Migration from
Read more

File Structure (Part 1)

Image
There are few ways for people who arrange their folder structures and which i do not wish to invent it. My approach is to arranging my folder structures based on Features not by types What is Type? Angular have alot of types. for example, we have Controller, Services, Directives which are the commonly used. We do not wish to arrange our file structure like this: This file approach is suitable only for SMALL application, for enterprise application, we have tons of Modules and alot of programmers are working on different modules. So before we jump to multiple modules, we should first arrange our file structures by modules. Let say build an intranet application which consists with Employees modules, eTimelog modules and eLeave modules. I would like also to create main folder to stores all my framework (layout, sidebar, navigationbar, footer-bar etc on my site. here is my simple file structure: - common: putting all my modal, services\repositories\utilities class to be
Read more