With an excess of $1 trillion in assets under management in more than 100 markets, this global investment management firm generates hundreds of petabytes of data per year. The firm’s IT team was charged with looking for ways to increase service velocity and agility, deliver more consistent user experiences across devices and gain cost efficiencies. This led the firm’s storage architecture team to consider Ceph, an open source, massively scalable distributed object and block storage system running in the Linux kernel on a Red Hat OpenStack platform.
In particular, the firm was drawn toward Ceph because of its ability to auto-scale to the exabyte level and beyond, and the fact that the architecture is self-healing and has no single point of failure. Also of interest was the fact that Ceph is a portable storage platform that can be used in public, private and hybrid cloud environments.
But to adequately evaluate how the solution would perform, the firm needed to test Ceph within an environment that closely emulated their own.
Initial conversations between the investment firm and an OEM partner identified the need to perform a full proof of concept (POC). However, neither the OEM nor the investment firm had the ability to construct an emulated environment in which to test OpenStack and Ceph.
Because the lab services capabilities and engineering resources of the Advanced Technology Center (ATC) are well known in the industry, the OEM turned to WWT to perform the full POC.
The POC preparation took approximately three weeks, and the test itself was conducted just over two. The OpenStack pod within the ATC had two OpenStack Regions with Ceph Object Gateways, and 12x Cisco UCS Nodes (rack and blade) consisting of:
- 2x OpenStack Control
- 4x OpenStack Compute
- 2x CEPH Object GW and Replication
- 4x CEPH Storage
The ATC ran a Kablue policy routing via a virtual machine for 35-70 millisecond latency generation and Citrix VDI instances.
The POC tested access to Ceph’s object storage system between sites at variable latency and loss scenarios as well as other test parameters. Among tests performed, the customer’s network latency and traffic characteristics were mirrored and varied during the POC to provide a thorough assessment of the capabilities of OpenStack and Ceph.
The POC achieved all predefined success criteria, including the successful emulation of the investment firm's production environment (including data packet speed and other performance measures to prove the ability to use OpenStack in their multi-region environment). By using the ATC, we were able to validate that Ceph — the massively scalable, open source, distributed storage system — remained viable in the customer’s complex IT environment.
The proof of concept resulted in the following outcomes:
- Confidence: The tests performed in the POC were deemed a complete success. They provided valuable lessons learned during the testing process and assured the firm that, should they move forward with a Ceph solution, that decision would be based on sound data.
- Quality assurance: The test plan ensured all steps were executed and all observations captured for accuracy.
- Risk reduction: The POC tested access to Ceph’s object storage system between sites at variable latency and loss scenarios to identify potential bottlenecks and any areas of risk.