THE CHALLENGES

As DataComm Cloud is offering a new cloud service to the target market, the challenges are how to provide a competitive cloud service compared to the market giants such as Google Cloud, Microsoft Azure, and Amazon Cloud Service. Operating VMware cloud for more than five years and running Kubernetes cloud already, DataComm Cloud was also looking for extending their cloud service with OpenStack for enterprises in Indonesia. Their OpenStack operation requires the use of block storage and object storage. If they choose a storage design where the block and object storage come from different storage systems, this will add complexity for monitoring, managing, and future support capabilities.

Also, the cloud infrastructure and the application requirement will likely scale out along with the actual business need. Application and users will consume more capacity over time, and therefore the anticipated storage capacity might not be able to 100% meet the original plan. In addition, DataComm Cloud is not only operating one datacenter but up to 4 different datacenters, to provide high reliability, high availability, data security, and edge cloud service to enterprises all over Indonesia, With frequent happened power shortages suddenly in the country, it's important to consider data backup and disaster recovery ability among the current 4 datacenters, which may extend to 5 or 6 data centers very quickly.

Furthermore, the DataComm IT team pays lots of attention to Power Usage Efficiency (PUE) since they own more than one datacenter facility, and the new DCs are already on the plan. All the incoming datacenter build-up shall set PUE as mandatory criteria for equipment arrangement.

THE SOLUTIONS AND BENEFITS

Considering the OpenStack infrastructure and the potential blooming cloud business, the DataComm Cloud team takes more than one year to evaluate the right solution on the backend storage to OpenStack.

For storage needs, OpenStack cloud architects usually choose Ceph as the storage system for its scalability, flexibility, rich native support, and high availability. Also, Ceph storage is the most popular open-source software-defined storage in recent years.

With Ceph, users get all the benefits of open-source software. It supports multiple protocols such as object storage, block storage, and file system. With Ceph, OpenStack got what it requires to run with predictable performance and reliable integration.

WHY AMBEDDED ?

There are many options to have Ceph storage, DataComm Cloud could choose to deploy and manage the ceph cluster by their engineers or choose one Ceph storage vendor to provide the service.

Ambedded offers a Ceph storage appliance solution to the datacenters and enterprises. This solution is for the IT users who intend to deploy a Ceph cluster but without an experienced team to have initial deployment, optimize configuration, and maintain daily operation.

The Ceph Appliance is a turnkey solution integrated with ARM-based micro-servers, Ceph storage software (optimized to adapted micro-server architecture), and web-based GUI management ("UVS Manager"). This solution facilitates non-experienced Ceph users to deploy, configure and manage Ceph clusters without an entry barrier.

THE RESULT

The Datacomm Cloud has deployed three Ceph clusters across three datacenters using the Ambedded Mars 400 Ceph storage appliance. Each Ceph cluster has mixed SSD OSD & HDD OSD for different performance purposes use. Mainly for the RBD use by OpenStack Cinder and Glance. The three datacenters have a backup for each other with high availability, so the enterprise customer would not worry if any sudden disasters cause data loss. With the Ambedded Mars 400PRO Ceph appliance, the Datacomm Cloud team doesn't need to start a large cluster to run with their OpenStack IAAS, simply four units per site to start the operation with their availability concerns. The future capacity & performance requirement can be scale-out by adding more Ceph appliances simply. Also, thanks to the low-power design on this Arm-based microserver, each cluster consumes less than 500Watts per site, showing a significant power saving compared to run a Ceph cluster on traditional servers in the same Ceph capacity.