EBS, EFS And Storage Gateway Flashcards
What is an instance store and why would you use one? Is there a cost?
Instance stores are physical storage attached to the physical host. As there is no network involved they are very fast and are used often for buffers, cache, scratch data - i.e temporary re-creatable data that needs very high IOPS (100’s of thousands). There is no cost to the instance store.
For IO1, GP2, ST1 - what is the maximum size you can provision?
16TiB
How would you use EFS as a cache for a web applications session state for multiple servers?
You would create an EFS volume and map it each individual server, they could then write session state data to the mapped volume.
What are the 4 types of EBS volume?
GP2, IO1, ST1, SC1
When you restore a snapshot of an EBS volume, when is the data available for use? Are there any latency impacts when using a restored snapshot?
Immediately. There are latency impacts if you don’t initialize (pull the blocks down from s3 ) the volume prior to using it with dd or fio.
What are the 2 PERFORMANCE modes for EFS? Why would you use them? Which has the higher latency and throughput?
General purpose and MaxIO. GP IOPS scales to 1000’s of concurrent connections, good for web servers, content sharing - general purpose use case. MaxIO Is higher latency but also higher throughput and is used for Big Data, media processing or anything that requires high IO.
Can you use EFS on a Windows EC2 instance?
No. Linux only, because Windows is arse.
Which storage type presents to an EC2 instance as a disk volume?
EBS
If I have a GP2 Drive sized at 100GB, how many IOPS do I have?
3*100=300IOPS
What are the two THROUGHPUT modes EFS supportS?
What are the 2 PERFORMANCE modes for EFS? Why would you use them?
Throughput
Bursting: Throughput will scale with file system size allowing for short bursts
Provisioned: Allows you to specify MB/S up to 1025MB/Sec.
Performance
General purpose and MaxIO. GP IOPS scales to 1000’s of concurrent connections, good for web servers, content sharing - general purpose use case. MaxIO Is higher latency but also higher throughput and is used for Big Data, media processing or anything that requires high IO.
I have a small group of machines in an autoscaling group which require access to an EFS volume. How do I secure access to that volume to allow only those machines in my ASG to access it?
You would use security groups with an inbound rule for your EFS volume allowing access from Ec2 instances in your ASG - those instances will have their own SG assigned on creation by the ASG so your inbound rule on EFS would reference your EC2 instances SG
You have a database that requires reliable IO of up to 230,000IOPS. Which storage solution would you use, why and what are the caveats
EFS cant be used as this is file storage, not block. EBS GP2 tops out at 3KIops or 16kIOPS with an additional fee. IO2 tops at at 32K (or 64K for Nitro). The only option is storage attached directly to the instance host - which leaves an instance store. As this data is ephemeral, we need to make sure that we have a replication and backup/restore mechanism as instance stores do not have an automated backup mechanism.
Is there a performance impact to snapshotting an EBS volume?
Yes. Snapshotting is a highly IO intensive activity and should ideally not be run during peak times.
On an EBS volume, do you pay by the amount of data stored, or the amount of storage you provision? Is this also the case for EFS?
You pay on the amount of storage you provision.
For EFS you pay based on the volume of data you store.
Why would I use an IO1 over a GP2 drive?
Max IOPS for GP2 is 16K. You would need IO2 if you had critical applications which required high, sustained consistent IOPS - I.e a large Database workloads.
