I would say following are 5 major pain points for the Amazon Web Services against Google Compute Engine, this article is entirely based on my opinion.....
1. Shared Block Storage
One of the Traditional practice followed in the IT Infrastructure for very long time, sharing a block storage across the group/cluster of servers, this would leverage the Operating System Clustering (MSCS/RHEL HA) and Database Clustering (Oracle RAC etc.,)
Suddenly when the world was moving toward cloud computing platform, sysadmins and infrastructure specialists faced a tough task on how to handle grids and clusters and had to change their tried and tested architecture to fit the cloud infrastructure. However nobody in the public cloud space yet to address this issue
At-least in GCE, a volume can be mounted to multiple instances as read only which does not require any additional subscription or services.
2. Network and IO
It is very clear that Google will be ahead in this space since they have a dedicated network connecting their Global Datacenters which means customers can even setup synchronous replication between regions, as tested the average bandwidth between regions recorder steadily over 300 Mbps whereas AWS who relies on public internet for the inter region connectivity recorded average 40 Mbps
Another interesting feature of GCE is their cloud platform network approach is very straightforward which makes the traditional network engineers/architects happier unlike AWS which is very confusing in the beginning and not so great NAT feature.
3. Block Storage
AWS is catching up to the trend by providing EBS (Elastic Block Storage) for their Virtual Instances otherwise customers would be unhappy with their not so great ephemeral (instance store) storage which cleanly sweeps the data when the Instance gets rebooted (like swap space in Solaris). So choose the EBS backed AMI (Amazon Machine Images) when deploying instances else you will be back to the start line :-)
GCE by default provides persistent disk which behaves like a normal disk in the server/desktop and can go upto maximum of 10 TB where in AWS maximum size of EBS is 1 TB
* Both the providers offers disks resources via Network
4. System Images/Boot Time
Machine Images are nothing but Server Images to maintain the standard or the sysadmins creates one based on the Organizations policy.
Interesting feature with GCE is that when the Image Resource is created, it becomes a global resource which means the images can be accessed from any region.
In AWS, users needs to initiate the AMI copy for cross region access else the AMI will be accessible only to that particular region.
In GCE, average system launch and access time is 30 seconds whereas in amazon the average launch time is 2 minutes, i ended up launching a Linux VM and got root prompt in 25 seconds which i never realized would happen when i started as Linux Admin.
5. Graceful Infrastructure Update
Hope many would know about the infamous incident of AWS outage where most of the VMs gone offline, AWS informed its users that it would reboot the Instances in-order to apply system/security updates which made some users go unhappy.
But GCE has the feature of Live Migration, it will migrate the customer VMs to the reserved location/host and do the maintenance activity, so the users need not worry about their VMs going down. Also one additional feature is Automatic Restart which users can set this feature incase of any system generated event such as hardware or host failure itself, GCE claims the reboot will happen under three minutes :-) This feature is one such enterprise-grade offering in IaaS and will surely benefit large enterprises moving to cloud.
These are observations i did on the AWS and GCE infrastructure after few weeks of usage, please do write to me incase of any suggestions/changes
No comments:
Post a Comment