Hi Folks. After a long absence from blogging attributed to a move of apartment, job and some grandchildren, I thought it was time to take up the proverbial pen.
One of the big changes that took place was a change of jobs - in my case moving from EMC to Oracle - Blue to Red literally. In making that move there was some substantial levels of orientation and technology to absorb.
Let's face it, it is not every day that you are expected to learn about all those high-end enterprise applications. This takes time. You probably know what I am talking about - assets absorbed from Sun, StorageTek and a host of other acquisitions based in the Cloud coupled with databases, big data, data architecture strategies and the list goes on.
One of things that did come up was to understand what enterprise and industry wisdom mean to people. At my previous employer, enterprise applications were big complex constructions using by association big complex storage arrays, with big complex SANs and big complex backup environments. If possible, this should all be wrapped in a big complex virtual infrastructure based on VMware, preferably using Cisco UCS server blades and switches. Fine. No problem - as there was also value to be derived for the client.
However, the nature of complex applications at EMC was at that time looking at things like large Microsoft Exchange deployments, or database environments supporting ERP, Microsoft SharePoint and indeed mainly looking at the storage ramifications of thousands of users hitting disk in parallel.
However, the enterprise is vastly more complex than that. The ability to interact and manage the user experience of customers through sophisticated CRM offerings, the ability to manage oceans of structured data using massively parallel databases and indeed prepare for new techs like big data is a daunting task.
It is no wonder that the world of x86 server virtualization only penetrated to a certain level of an organization. The industry wisdom of "it must be on x86", "must use horizontal pools of infrastructure resources", "must use consolidated centralized storage" and "must be virtualized in VMware or some other hypervisor to be Cloud ready" is not a sufficiently complete answer to tackle these themes.
That did not stop every Tom, Dick and Harry from trying to sell a hypervisor of some sort with all the associated paraphernalia. This in turn generated a spree of IT organizations trying to recreate what they already had but just delivered "slightly differently".
Let's face it folks, Cloud is not "new" in the technology sense - no matter how many times the marketing departments of certain organizations try to convince you. More accessible -yes, easier to use - yes, new - no!
It is no wonder then that it has taken the IT industry some time to grapple with what Cloud and this "industry wisdom" all means for them. This has led to some important questions arising -
- "must it always be x86?",
- "should I forget all the last 30 years of IT knowledge?",
- "what is the role of IT in an organization?",
- "Dare a different approach be used not based on x86 exclusively?",
- "can a normal Unix be used in conjunction to a Linux/Windows on non-86?".
- "can I get sustained competitive advantage be being only commodity x86/Linux based?"
With the current industry trends and wisdom, I believe that it is incumbent on CFOs, CIOs and indeed CTOs to take a little step back. CTOs have the chance to examine technology for 'fit for purpose' and innovation certainly, whilst the CIO needs to ensure its practical application for the good of the organization.
The recent trend of CIOs reporting to CFOs is an interesting change. The CFO represents the ability of an organization to challenge what it is doing and why it is doing this. Innovation and creative thinking driven by financial stewardship!
CFO Trigger to Innovate and Question "Industry Wisdom"
Rather than viewing CFO oversight as a burden, it may well be that this "get back to basics" is a blessing in disguise. IT organizations get entrenched in their ways. There are many reasons to continue doing some things as they are - but plenty that need to change in light of CFO oversight.
However, there are some things that simply need to have a much higher level of efficiency. I have been meeting many customers using the Oracle database, and am surprised to see how few of them really use all the features that are available.
Indeed, the IT organization says "thou be too expensive matey", and then goes off on a spending binge to bring in a complete new stack of VMware with SAN storage, switches, servers and what have you - whilst reinventing ITIL so that it fits VMware operations. Ummm. That is not necessarily cheaper.
Don't get me wrong. This is not a mistake - but it is based on a pattern of break it all up so that it fits on a small x86 server and then virtualize it - how else do you sell a VMware/chargeable hypervisor license?
Well there are some workloads that fill up the entire machine - databases supporting large ERP installations. They can be broken up - but certain features like live VM migration of a running DB instance to another machine don't make a lot of sense when the VM has hundreds of GB/TB storage and 1/2 a terabyte of RAM.
In this case, the reverse pattern - antipatterns - is perhaps a better one - literally using machines in an aggregation to service the workload - spreading the DB instance across multiple machines - this already exists in Oracle RAC. This is not new - and works pretty well.
We see other patterns like this in the Google datacenters - where search workloads are "spread" across servers working in concert - wow sounds like a cluster!
The CFO can and indeed should work with CIO/CTOs to start questioning investment decisions. The one thing I saw with IT organizations using VMware/SANs/SANswitches/etc was that from a financial perspective something was not quite right:
- IT Investment that was leveraged on the back of a genuine critical business project was spread around the datacenter infrastructure
- Storage arrays were expanded to accommodate the project, new switches were acquired, new licenses purchased etc.
- Everything was mixed up - to the point that IT could not say where the money went - and the business could not verify that it had received what it actually wanted.
- Billing and chargeback were either non-existent or so primitively applied that it failed to gain traction
This was all justified through the IT organization rejigging its numbers. Not a good sign! The strength of the original investment was diluted through spreading it around such that the business unit really needing those resources was effectively "short changed" and trying to simply "make do".
Why is this important for the CIO?
At a time of crisis and oversight, it is incumbent on the CIO to question the approach put on his/her table for transformation or rejuvenation projects.
Verify whether there is another approach. Verify the patterns that are emerging. Remember, the reason we had wholesale x86 virtualization was that the servers running their particular workloads were not using hardware to its maximum capability or IT had simply decided for isolation reasons to add more servers etc.
- If servers are being used to their practical maximums - does the IT organization still need to invest in a chargeable hypervisor?
- If isolation can be provided differently than by encapsulating an entire operating system stack (a VM) - does the IT organization need to still make virtual machines and acquire more licenses for a hypervisor and more resources for server/SAN etc?
x86 was also the commodity route - costs less - doesn't matter what the workload it - just get ton loads of them - we'll worry about the datacenter space/power we need later. Well mainframes do a hell of a lot of work - albeit complex to manage. Big iron Unix boxes were simpler to operate but also doing a lot of complex workloads in parallel.
- If a larger server can intrinsically do the job of many smaller servers with high levels of reliability, isolation and performance - do we need to invest in many smaller servers? - do we need to get x86 only? - must it be Linux?
If we examine the world of tomorrow - the idea of holding structured information of some sort even if acquired from unstructured data sources as in the case of big data is still valid. Logical isolation is still needed and we do not need to throw out the last 3 decades of database innovation just because some marketing scheme is based on that.
The same and more applies for applications - value added apps are available to an increasingly mobile customer base demanding access 24x7 and from any device. The web and indeed technologies that encapsulate applications like Java are being used to address these needs. SOA architectures were developed to decompose applications to reusable components that could be recomposed to add more value and shorter time-2-value.
These are the real enterprise applications that drive value. That is where the investment should flow. Investment needs to be concentrated into those areas to make the dent in the business revenue that is needed. This implies a "vertical" architecture model of investment.
However, IT organizations need to manage large estates. They think "across" those estates - a "horizontal" pattern of activity. Good for IT - bad for concentrated vertical investment based on projects.
However, I believe that horizontal manageability can be achieved through management tools whilst allowing investment to be applied vertically for maximum results. The CIO can create the conditions where this is encouraged and create a level of transparency in datacenters and IT organizations that has hitherto been lacking.
This gradual transition to running IT as a business itself with the same level of fiduciary responsibility that the rest of the business is subjected to can and should be the driver for the recomposition of IT services and the new role that IT needs to play in ever tighter markets.
In the next series of blogs I will take some of these themes and see how that can be done from an architectural pattern and technology point of view for CIOs in the context of business driven projects.