GOMAINFRAME.COM
Your Subtitle text
Mainframe Cost Argument
Why I'm So Adamant That Mainframe Architecture Is Superior to Distributed Computing Architecture
As I've traveled around the world, it constantly astounds me to see IT executives who believe that the best way to scale up their computing environments (add more capacity) is to add more and more distributed servers. They buy these servers for $5,000 to $50,000 -- and when they deploy them, they underutilize these servers (these "cheap" servers are usually deployed at 10-20% of their potential capacity). And for some reason, these executives think they are saving money by buying cheap hardware; buying tons of extra software (including virtualization and provisioning software); buying extraneous network components, and performing loads of integration work themselves; paying innumerable people large salaries to manage distributed servers, storage and network devices -- instead of buying a turnkey, completely integrated, comparatively easier to manage mainframe environment!
The distributed computing model of adding small servers to scale capacity is broken -- primarily because it calls for constanly having to add people to manage additional capacity. Further, running very large data centers with thousands of servers wastes tons of electricity in both power and cooling functions (and wasting electricity is bad for the planet). Our industry has got to break these wasteful habits if we want our computing systems to serve our ever-growing demand for more computing power in the future.
IBM Competitors Who Criticize Mainframe Costs Don't Know What They're Talking About
Mainframe competitors and some IT executives criticize mainframe computing based upon acquisition costs are not comparing apples to apples. Yes -- from an acquisition point of view -- MAINFRAMES DO COST MORE FROM AN ACQUISITION POINT-OF-VIEW THAN LESS ROBUST, LESS SECURE, LESS FLEXIBLE, HARDER TO MANAGE, LESS INTEGRATED DISTRIBUTED COMPUTING ENVIRONMENTS. But from a total-cost-of-ownership perspective, mainframes outshine distributed computing environments when it comes to operational cost (mainframes cost significantly less to operate than typical distributed environments -- usually by a very wide margin).
What you are paying for when you buy a mainframe is a mature, highly-integrated computing environment. Over 50 years of development work has gone into integrating mainframe hardware, operating environments, and related tools and utilities. When you buy "industry standard" hardware -- you are paying for the hardware and some integration work (but you're paying for far, far less integration work than has been performed on a mainframe). In order to integrate your environment, you get to pay your staff or third party consultants to perform the integration work that you need done in order to glue-together your "industry standard-based distributed environment". And you get to pay various vendors for infrastructure components that YOU have to glue together (components that are already included in the mainframe environment).
Oops I Forgot... Additional Costs That Show-up After You've Deployed Your Distributed Computing Architecture
What tends to happen is that competitors focus their attention on mainframe hardware/OS acquisition costs -- and those costs are high when compared to bare-bones industry standard hardware costs. What mainframe competitors fail to include are all of the additional costs related to making their systems as robust as a mainframe. So, for instance, competitive quotes will omit costs such as:
Until recently, IBM has done a less than stellar job of dealing with the mainframe acquisition cost issue. Typically, when asked about mainframe acquisiton cost, an IBM sales rep will avoid the question and instead start talking about mainframe TCO (total cost of ownership -- the points mentioned in the preceding section). And, for a small percentage of mainframe prospects, this approach has worked (because these prospects are very aware of their true operational costs and see the mainframe as a vehicle to help significantly reduce their operational costs). But, for the most part, this TCO argument falls on deaf ears -- largely because it has become easier for IT executives to purchase low cost servers to add computing capacity. Note that these same managers are rarely responsible foradditional operating costs that impact the ultimate price that they are paying for additional computing power (such as power and cooling costs or real estate costs).
It should also be recognized that IT executives can easily obtain purchase orders for ancillary software (for such things as security software, infrastructure software, etc). The financial departments that service these IT executives have been set-up to make the purchase of small price tag purchases (such as low-cost, industry-standard servers) easy. This, in turn, makes it easy for IT departments to build distributed server environments -- but these same financial departments choke when they have to deal with larger, all-at-once purchase orders for large, integrated servers...
To IBM's credit, the company has recently (and finally) started to address its cost-of-acquisition issue through certain price reductions and by building lower cost "package pricing". Clabby Analytics believes that IBM's new "Solution Editions" for mainframes is a huge step in the right direction. For more information on IBM mainframe solution sets, go here (SOLUTION EDITIONS).
How Clabby Analytics Would Argue the Mainframe Cost Issue
First and foremost, I'd get the street prices for competing servers (such as an HP SuperDome or a collection of distributed servers) -- and I'd compare the acquisition costs to a mainframe environment of equivalent computing power. Using the solution edition pricing (mentioned above) -- I would show IT buyers how close mainframes and SuperDomes/distributed computing environments really are in acquisition costs. And I'd be sure that the head-to-head acquisition costs include the additional networking costs for the SuperDome (much of the mainframe networking cost is included in the base price of the mainframe -- running over an internal high speed bus architecture). Once you start rolling in the networking costs, it is amazing how close distributed computing environments and mainframe environments are in terms of acquisition costs (see Figure 1) for an example of the network cost penalty paide by distributed computing users.
Figure 1 -- The Networking Cost Penalty for Distributed Computing Environments
I would then start tearing into distributed computing architectural costs on the basis of:
As I've traveled around the world, it constantly astounds me to see IT executives who believe that the best way to scale up their computing environments (add more capacity) is to add more and more distributed servers. They buy these servers for $5,000 to $50,000 -- and when they deploy them, they underutilize these servers (these "cheap" servers are usually deployed at 10-20% of their potential capacity). And for some reason, these executives think they are saving money by buying cheap hardware; buying tons of extra software (including virtualization and provisioning software); buying extraneous network components, and performing loads of integration work themselves; paying innumerable people large salaries to manage distributed servers, storage and network devices -- instead of buying a turnkey, completely integrated, comparatively easier to manage mainframe environment!
The distributed computing model of adding small servers to scale capacity is broken -- primarily because it calls for constanly having to add people to manage additional capacity. Further, running very large data centers with thousands of servers wastes tons of electricity in both power and cooling functions (and wasting electricity is bad for the planet). Our industry has got to break these wasteful habits if we want our computing systems to serve our ever-growing demand for more computing power in the future.
IBM Competitors Who Criticize Mainframe Costs Don't Know What They're Talking About
Mainframe competitors and some IT executives criticize mainframe computing based upon acquisition costs are not comparing apples to apples. Yes -- from an acquisition point of view -- MAINFRAMES DO COST MORE FROM AN ACQUISITION POINT-OF-VIEW THAN LESS ROBUST, LESS SECURE, LESS FLEXIBLE, HARDER TO MANAGE, LESS INTEGRATED DISTRIBUTED COMPUTING ENVIRONMENTS. But from a total-cost-of-ownership perspective, mainframes outshine distributed computing environments when it comes to operational cost (mainframes cost significantly less to operate than typical distributed environments -- usually by a very wide margin).
What you are paying for when you buy a mainframe is a mature, highly-integrated computing environment. Over 50 years of development work has gone into integrating mainframe hardware, operating environments, and related tools and utilities. When you buy "industry standard" hardware -- you are paying for the hardware and some integration work (but you're paying for far, far less integration work than has been performed on a mainframe). In order to integrate your environment, you get to pay your staff or third party consultants to perform the integration work that you need done in order to glue-together your "industry standard-based distributed environment". And you get to pay various vendors for infrastructure components that YOU have to glue together (components that are already included in the mainframe environment).
Oops I Forgot... Additional Costs That Show-up After You've Deployed Your Distributed Computing Architecture
What tends to happen is that competitors focus their attention on mainframe hardware/OS acquisition costs -- and those costs are high when compared to bare-bones industry standard hardware costs. What mainframe competitors fail to include are all of the additional costs related to making their systems as robust as a mainframe. So, for instance, competitive quotes will omit costs such as:
- Virtualization costs;
- Provisioning costs;
- Security costs (add up the cost for licensing and managing security across hundreds of distributed servers);
- Networking costs;
- Physical plant costs (wiring/cabling);
- Real estate costs (a single mainframe can replace 250 industry standard servers -- at a fraction of the footprint);
- Power/cooling costs (including electricity costs and costs related to power supplies and cooling equipment);
- People related management costs (I've seen a major mainframe environment being run by 4 people!);
- Infrastructure software costs (mainframe managers tell me that mainframe infrastructure software licenses cost far less than the total cost for infrastructure licenses in distributed computing environments);
- Support costs (mainframes offer the most reliable hardware in the industry with the best meantime-between-failure [MTBF] of any commercial system. If stuff doesn't go down, then it doesn't need to be fixed -- hence support costs are lowered);
- Controlling/accounting costs -- mainframes offer the best control/account software (for internal billing purposes) in the industry;
- Availability/resiliency costs -- how much does it cost to build a disaster recovery plan and implement it across a network of distributed servers. I can assure you that managing business resiliency on a centrally controlled mainframe costs a lot less...
Until recently, IBM has done a less than stellar job of dealing with the mainframe acquisition cost issue. Typically, when asked about mainframe acquisiton cost, an IBM sales rep will avoid the question and instead start talking about mainframe TCO (total cost of ownership -- the points mentioned in the preceding section). And, for a small percentage of mainframe prospects, this approach has worked (because these prospects are very aware of their true operational costs and see the mainframe as a vehicle to help significantly reduce their operational costs). But, for the most part, this TCO argument falls on deaf ears -- largely because it has become easier for IT executives to purchase low cost servers to add computing capacity. Note that these same managers are rarely responsible foradditional operating costs that impact the ultimate price that they are paying for additional computing power (such as power and cooling costs or real estate costs).
It should also be recognized that IT executives can easily obtain purchase orders for ancillary software (for such things as security software, infrastructure software, etc). The financial departments that service these IT executives have been set-up to make the purchase of small price tag purchases (such as low-cost, industry-standard servers) easy. This, in turn, makes it easy for IT departments to build distributed server environments -- but these same financial departments choke when they have to deal with larger, all-at-once purchase orders for large, integrated servers...
To IBM's credit, the company has recently (and finally) started to address its cost-of-acquisition issue through certain price reductions and by building lower cost "package pricing". Clabby Analytics believes that IBM's new "Solution Editions" for mainframes is a huge step in the right direction. For more information on IBM mainframe solution sets, go here (SOLUTION EDITIONS).
How Clabby Analytics Would Argue the Mainframe Cost Issue
First and foremost, I'd get the street prices for competing servers (such as an HP SuperDome or a collection of distributed servers) -- and I'd compare the acquisition costs to a mainframe environment of equivalent computing power. Using the solution edition pricing (mentioned above) -- I would show IT buyers how close mainframes and SuperDomes/distributed computing environments really are in acquisition costs. And I'd be sure that the head-to-head acquisition costs include the additional networking costs for the SuperDome (much of the mainframe networking cost is included in the base price of the mainframe -- running over an internal high speed bus architecture). Once you start rolling in the networking costs, it is amazing how close distributed computing environments and mainframe environments are in terms of acquisition costs (see Figure 1) for an example of the network cost penalty paide by distributed computing users.
Figure 1 -- The Networking Cost Penalty for Distributed Computing Environments
I would then start tearing into distributed computing architectural costs on the basis of:
- Virtualization/provisioning costs -- mainframes are the industry's "GOLD STANDARD" in consolidation, virtualization, and provisioning. My recommendation: buy a mainframe -- set up its virtualization/provisioning/workload management environment -- and jump to the next level in server management (a set of products and services called "service management") as opposed to jumping backward into x86 virtualization and provisioning.
- (Go here for a discussion of mainframe service level management ); and,
- (Go here for a YouTube discussion of why I think mainframes should be used as centralized management hubs:
http://www.youtube.com/watch?v=vbHNU-QmksU)
- (Go here for a video of my recommendation to use a mainframe as a central virtualization/provisioning management environment for mainframe and distributed computing users: http://www.youtube.com/watch?v=vbHNU-QmksU).
- Security costs -- particularly the cost for licensing and managing security across hundreds of distributed servers.
- Networking and phyisical plant costs (see Figure 1 -- above. Enough said...);
- Real estate costs -- a single mainframe can replace a server farm of 250 industry standard servers -- at a fraction of the real estate footprint.
- Power/cooling costs -- including electricity costs and costs related to power supplies and cooling equipment.
- People related management costs (I've seen a major mainframe environment being run by 4 people!).
- Go here for mainframe customer comments on mainframe management
- Infrastructure software costs (mainframe managers tell me that mainframe infrastructure software licenses cost far less than the total cost for infrastructure licenses in distributed computing environments).
- Support costs (mainframes offer the most reliable hardware in the industry with the best meantime-between-failure [MTBF] of any commercial system. If stuff doesn't go down, then it doesn't need to be fixed -- hence support costs are lowered).
- Controlling/accounting costs -- mainframes offer the best control/account software (for internal billing purposes) in the industry.
- (Go here for a discussion of IBM's Tivoli Usage and Accounting Manager [IBM site]: http://www-01.ibm.com/software/tivoli/products/usage-accounting/);
-
Availability/resiliency costs -- how much does it cost to build a disaster recovery plan and implement it across a network of distributed servers. I can assure you that managing business resiliency on a centrally controlled mainframe costs a lot less...
