Bill and Storage Optimization

How to go about optimizing storage costs? What factors should a proficient organization take to ensure an effective utilization of investments in storage technology? 

While the topic is too broad and too complex to provide a recipe book, this article explores some aspects to consider and goes into some detail in some options.

***

Bill was at a meeting about optimising storage costs for an application.

Andrew (Andy) Bennet said: “Our director wants us to save 10% of our application costs and today, I would like to know my options for storage.”

Bill said: “Storage cost optimisation is always a tricky topic. What are your views on cost savings, Kelly?”

Kelly Khan, in charge of the Storage Tower, said: “There are different ways of optimizing storage spend. It seems in this case we want to optimize a specific application. That’s probably doable. More problematic would be saving storage costs as a whole.”

“Yes, our storage spend was 15 million last year.” Said Bill, after typing something on his tablet.

“We need to consider different aspects.” Said Kelly. “When we’re dealing with on-prem equipment, we must consider the depreciation of the assets. Shutting down a storage unit doesn’t mean that its costs disappear. Yes, it’ll stop consuming electricity and cooling, but the depreciation (D&A) will stay for the duration of the amortization period.”

“How long is that?” Asked Amanda Mills, from Architecture.

“Typically five years.” Said Kelly. “Also, by and large, we’re always in the middle of that life, so we have to reckon with another couple of years for most devices. Costs will not disappear just like that.”

“Are we talking about Operational Expenses (OpEx) or Capital (CapEx)?”

“Depreciation and amortization (D&A) costs are OpEx, Amanda. They reflect the capital investment in equipment, what we call financial assets. Of course, those investments are CapEx by definition.”

“Sure, I understand that, Bill. But, what about my application? Is there anything we can do?”

“There are a few options, Andy.” Said Bill. "I have extracted some information from our TBM system and these are the details for your application." Bill shared his tablet’s display on the conference room screen.

Table 1 - Sample server allocation to Tier 1 Storage

“The servers associated with your application all consume storage from a Tier 1 device. Each of the servers receives a share of the storage device proportional to its allocated storage.”

“What are those tiers, Bill?” Asked Andy.

“Kelly and her team helped us classify the various storage technologies, considering price and performance. When we get an invoice, we map the purchased item to one of the four tiers to simplify matters.”

“Why can’t we have the amount we actually paid?” Asked Amanda.

“Two reasons. The purchasing system doesn’t know what we’re going to use the asset for. Thus, we don’t have an invoice for any of the hardware items we have in the data center.”

“What’s the other reason, Bill? You said there were two…” Said Kelly.

“The other is again the concept of D&A. The financial system calculates the depreciation globally, and again, we don’t get a breakdown by device.”

“What is the OpEx for each of the tiers?” Asked Kelly.

Bill showed another table on the screen.

Table 2 - Cost per Storage Device by Tier

“Your application is using the most expensive storage devices, Andy.” Said Bill. “One thing we could do is move them to a lower tier.” Bill changed the display on the screen.

Table 3 - Sample Server Allocation to Tier 2 Storage

“Everything else being equal, you’d save your application 25% of its storage costs, or $10,000.”

“And how much would that save me in total?”

Bill tapped again on his tablet. “The storage footprint for your application is just under 7.4% of the total OpEx. By saving 25%, you’d go to a total saving of about 5.5%.”

“How do you get to that number?”

“If we go down to 75% of the storage costs, and these represent 7.4% of the total, multiplying those two numbers we get that 5.5%.”

“Thanks, Bill. We're not there yet. What else could we do?”

“Apart from playing with the tiers, the next lever you have is the allocated storage. Do your servers need that amount of storage? Is there any way you could bring those numbers down?”

“Are you running databases on these servers, Andy?” Asked Amanda.

“Sure, like any other application of this family. They all use the same database technology.”

“You could look at resilience and redundancy. How many instances of the database are you running? How many standby servers do you have? If your application follows a specific template, you might have streamlining opportunities. Maybe you can get rid of some test instances.”

“What did you mean by redundancy, Amanda?”

“Data redundancy. How many copies of the same data does your application have? That might give you further options.”

“I see. I guess I can also look at the retention period for some data tables.”

“Exactly, Andy. You might archive some data to Tier 4 devices. That would yield additional savings.”

“Thank you, ladies. This has been extremely informative.”

“Before we close the meeting, Andy, a word of caution. Some figures we’ve been talking about are just that: numbers. Before you make changes, please talk to your capacity planners and your DevOps team to validate any of the suggestions we’ve been discussing. You may want to get somebody from Amanda’s team as well.”

“Yes, that makes sense, Bill. I will do that.”

***

Copyright (c) Guillermo Cuadrado, 2022

​​​​