File and Object Storage

 View Only

A Financial Benefit of Operating IBM Ceph

By Greg Deffenbaugh posted 27 days ago

  

A component of selling data storage solutions is helping customers evaluate Return on Investment (ROI) and Total Cost of Ownership (TCO) of the proposed solution. In most cases these are academic exercises and don’t reflect reality because assumptions are at the core of inputs. For example, if the cost of downtime is $10K/minute and the solution eliminates one 4-hour outage a year, the return for that component is $2.4M which dwarfs the other components of the calculation. Further, you don’t really know how many outages would be avoided, so you have to “assume” a number. These calculations also tend to favor the solution that meets the minimum requirements the most cost effectively.  Therefore, the calculations assume that the requirements are defined accurately and completely, and even then, tend to undervalue new, innovative features.

Selling Software Defined Storage (SDS) adds inaccuracy to the effort. Evaluating the savings of a solution that runs on “commodity” hardware (which will be less expensive than purpose-built storage controllers or proprietary, single-source hardware) is hard to quantify do to lack of accurate hardware costs. Because of these fallacies, I try to look at financial benefits more realistically, but with less rigor. 

Ceph Provides Multi-Protocol Storage Without Compromise.

We often talk about providing file, block and object storage services from a single Ceph cluster which is common when providing storage services for OpenShift or when the total storage footprint is fairly small. However, for larger deployments, Ceph clusters tend to be more purpose built, i.e. build a cluster whose primary function is S3 storage for a data lakehouse, S3A HDFS replacement, block storage for VMware, high-performance files storage for analytics, etc. 

Today, most organizations have dedicated NAS (file), SAN (block) and S3 object storage solutions with teams aligned to support each solution. Some organizations have a two-vendor strategy, so they have 2- NAS, 2-SAN and 2-S3 object solutions resulting in 6 sets of SMEs to support the infrastructure. This is not very efficient, for example both NAS teams understand the common problem space but tend to have limited knowledge of the “other” NAS solution. Applying Solution-A management practices to Solution-B is usually a very bad idea.

Have one SME leave for greener pastures, and another go on vacation when something goes wrong and you may struggle to be able to apply the right knowledge to the problem.

IBM Ceph gives you the ability to make much better use of your valuable IT personnel.

Ceph is architected with a set of common services that manage the hardware, data placement and protection, cluster services, etc.  with separate “protocol stacks” for file, block and object. Anyone who manages one of the protocol stacks will have a very good understanding of the core functionality (LIBRADOS & RADOS).

Adopting Ceph as a storage standard in your organization requires building a single Ceph architecture and administrative team with people having sub-specialties. For instance, NAS specialists will focus on CephFS because of their understanding of integration into authorization and authentication systems, file sharing use cases, data archiving / protection strategies, etc. The ‘old’ S3 teams will align to RGW management for the wide variety of object use cases. You may even have sub-sub-specialties. Deploying RGW for archiving is different from supporting analytics with S3A and data pipelines.

Now take on the scenario where you are down a couple of headcount in one area and something goes wrong. Everyone on the expanded Ceph team knows the basics – can: bring a failed node online, pull logs, restart process, install updates etc. The RBD block storage SME can look at an S3 cluster and determine if the issue is in the core functionality or something in the RGW or network layer. 

This brings up another common customer Ceph discussion. IBM Ceph is based on open-source Ceph.  “Why should I pay for IBM supported distribution?” Suppose your entire team has 300/400 level knowledge of Ceph, some have 500 level knowledge in a particular protocol or use case. IBM likely helped the 500-level person acquire their knowledge and further provided confidence that Ceph was being deployed properly for the use case. That support is also available to help the 300/400 level person when the 500-level person is on Vacation.

I recently posted what I thought was a simple question to an internal IBM Ceph slack channel. Got responses from a development engineer, a support engineer and a couple of product managers providing ideas. In the end, we came up with a strong answer that was “supportable”. We don’t allow customers to access our internal slack channels, but we do provide access to the SMEs who answered my question. (Google search didn’t offer an answer and just pointed me to the documentation.) 

When asked about upstream vs IBM distribution of Ceph, I think of the Ray Parker Jr. lyric – “if there is something strange in the neighborhood datacenter, who ya gonna call?” There are a lot of books, blogs and chat histories you can use to learn to setup and run upstream Ceph. But at 3:00AM, books, blogs and chats are a tough way to deal with an issue, much better to be able to reach out to another team of SMEs.

Since the IBM Storage Division took over the responsibility to support the upstream development of Ceph in January 2023, Ceph has evolved into a solution designed to be managed by traditional storage teams. Ceph technology is being evaluated as the next-generation storage architecture, providing cloud-like storage scalability and availability, multi-protocol support, and very good economics.

I can’t tell you how many SMEs you will need or how much operational savings you will get deploying Ceph, but you will get better value from your limited IT personnel while helping your organization adopt cloud-like storage infrastructure on-prem. I also believe that deploying Ceph will help with your personnel retention. Ceph is “cool” technology that can solve a very wide range of business and technical challenges. Learning to effectively deploy and manage IBM Ceph provides a lot more job satisfaction than running a GUI on a black box system.

Image sourced from pixaby.com


#Spotlight
0 comments
11 views

Permalink