ELASTIC FILE SYSTEM

• This is also considered a file system and is also optimized for low latency
• But unlike EBS it supports multiple access by EC2 instances at once
• It uses a file manager interface that uses standard file system semantics such as
: _lock files _rename files _update files _
• It also uses a hierarchy structure very similar to on-premises systems
• It allows you to store files that are accessible network resources
• O-premises network file systems will have users access files by browsing
• They connect to a server perhaps by a MAP drive that has been configured for
them
• When you connect to them you see a tree view of available folders and files
• Various LAN network systems provide this functionality like file or Storage Area
Networks (SANs) or Network Attached Storage (NAS)
ELASTIC FILE SYSTEM
• Amazon EFS provides the ability for users to browse cloud network resources
• EC2 instances can be configured to access EFS using configured mount points
• Mount points can be created in multiple AZs to attach to multiple EC2s instances
• Like your traditional LAN servers EC2 instances are connected to a network file
system EFS
• Performance factors make EFS desirable for a wide variety of workloads and use cases
and can meet the demands of 10’s, 100’s to 1000’s of EC2s
• From a user standpoint, the result is the same, the user accesses the network
resources like they have always done only now it is done using cloud resources
• EFS is a highly available fully managed durable service that scales to petabytes with
low latency and high throughput
• Being a managed service, there is no need for you to provision any file service to
manage the storage or provide any maintenance of those servers
ELASTIC FILE SYSTEM
• This makes it a simple option to provide file level storage in your environment that
uses standard operating system API’s, since any application designed to work with
standard operating system will work with EFS
• It supports both versions of NFS 4.0 and 4.1 and uses standard file semantics such as
strong consistency and file locking
• It is replaced across AZs in a single region making EFS a highly reliable storage service
• Like your traditional LAN servers EC2 instances are connected to a network file
system EFS
• As the file system can be accessed by multiple instances it makes it a good storage
option for applications that scale across multiple instances allowing for parallel access
of data
• The EFS file system is also regional so any application …….
• EFS was not available across all regions when this course was written
ELASTIC FILE SYSTEM
• The storage classes found in EFS include:
• Standard
• Infrequent access (IA)
• Both storage classes are available in all regions where EFS is supported providing the
same level of availability and durability from a standard storage class to the
infrequent access storage when it is enabled
• That process occurs when a file is not written or read to for a set period of time,
which is configurable, with options for this period range being: 14,30,60 or 90 days
• Depending on your selection EFS will move the data to the in storage class to save on
cost once that period has been met
• However, as soon as that same file is accessed again the timer will be reset and it is
moved back to the standard storage class
• Again, if it has not been accessed for a further for a further period it will then be
moved back to IA
• The only exception to the data not being moved to the IA storage class is for any files
that are below 128 KB in size and any metadata and any metadata of your files that
will all remain in the standard storage class
ELASTIC FILE SYSTEM
• If your EFS file was created after 13th February 2019, then it can be witched on and
off for its life-cycle management feature
• EFS can be used for a number of different use cases and work loads and as such each
use case may require change of performance throughput/IOPs and latency point of
view
• As a result EFS has introduced two different performance modes that can be defined
during the creation of your EFS file system
• These being:
• General purpose
• Max I/O
• General purpose: it offers low latency and all round file performance with a limit
mode of 7K file system operations per second on ??
• If you have a large scale architecture where EFS is likely to be used by thousands of
EC2s concurrently and is likely to exceed 7K operations per second you should
consider Max I/O
ELASTIC FILE SYSTEM
• This mode offers virtually unlimited throughput and I/OPs
• The downside is that your operation latency will take a hit compared to the
general purpose class
• If your application sits below 7K operations per second then the general
purpose class can be used with the added plus point of low latency
• However, if your testing confirms 7K operations per second may be reached or
exceeded then choose the Max I/O
• When using general purpose mode EFS provides a CloudWatch metric percent
I/O limit that will allow you to view your operations per second as a percentage
of the top 7000 limit
• This allows you to make the decision to migrate and move to the Max I/O file
system should your operations be reaching that limit
ELASTIC FILE SYSTEM
• I addition to the performance modes EFS also provides
• Bursting throughput: it is the default mode and the amount of throughput
scales as your file system grows
• Provisioned throughput:
• Data throughput patterns on file systems goes through periods of relatively low
activity with occasional spikes and burst usage
• EFS provisions throughput capacity to help manage this random activity of high
peaks
• Bursting throughput is the default mode and the amount of throughput scales
as your file system grows
• The more you store the more throughput is available to you
ELASTIC FILE SYSTEM
• The default throughput available is capable of bursting to 100 MebiBytes
• However, with the standard storage class can burst to 100 MebiBytes/TebiBytes
used within the file system
• For example, say that you have 5 TebiBytes of storage within the EFS system
your burst capacity can reach 500 MebiBytes/second
• The duration of the throughput bursting is reflected by the size of the file