Dec 04
The last time I was genuinely excited about ‘free’ was when Google sent me 100 bucks of free advertising on Adwords. The time before that was when EMC hooked me up with a nice little USB stick at the EBC.. Thanks again Joe, I still use it constantly. Free gadgets are one thing (more!), but free services is another. 
Through the course of the global financial crisis, there’s been a vapid shift in the IT infrastructure consulting market to provide ‘free consulting’. The loss leader approach to winning business has been around forever, but the more the ‘free consulting practice’ proliferates in IT infrastructure the more you have to wonder what is really being provided. Various books and management articles expound the brilliance of free, and possibly the absolute requirement for free to compete in our new global supersonic economy.
Makes sense if the bench is warm, pipeline is light, yet you don’t want to see 10 years of talent vanish due to P&L pressures for short-term gains and long-term painful rebuilding. For hardware vendors, you have a magic pool of ‘marketing development funds’, which are applied feudal style to strong upside product win opportunities (competitive accounts, account salvage ops, etc.). But for pure consultancy, you might have bench cycles to apply, or you might take an opportunity loss to deliver ‘free’ services.
A couple challenges with free consulting services:
- Free means no budgeting approval, therefore no real effort on the part of the buyer
- Free results in little to no emotional commitment from the buyer or the business
- Free diminishes stakeholder involvement (often related directly to issues that need the most attention)
- This sets an ugly stage, since the underlings then have zero interest in playing-ball
- Free devalues the work being done and casts a cheap light on the deliverables, no matter how well crafted
Any way you cut it, Free Equals Nothing in the services industry. You must have skin in the game at some point, some time, some where, with somebody in the food-chain. Otherwise, free is nothing more than a LOST leader, a waste of time and talent.
Tagged with: infrastructure • services
Nov 30
Like many good things in the US, cloud computing originated on the West coast inside the innovative web giants Google and Amazon. Both offer forms of storage and application services. Google aims today at the consumer and end-user with the ever-expanding office-type applications (Calendar, documents, mail, etc.). For many organizations (such as the city of Los Angeles) Google replaces the traditional Microsoft desktop application suite. Amazon is geared more for advanced consumers (like web developers) and the enterprise with a range of increasingly data center friendly storage and compute services.
An ever-increasing list of cloud-storage focused startups are on the scene, and now the major vendors are making entries to market, slowly but surely. Don’t forget the Telecom’s like AT&T, who like EMC, are closely modeled after the Amazon cloud model. 
The public cloud model threatens the premise of the traditional enterprise hardware and software industry, so we are beginning to see the major infrastructure vendors positioning for the private cloud market, as seen with the EMC/Cisco/VMware joint venture cloud services company Acadia and a slew of similarly minded partnerships.
Indeed, the cloud is a confusing marketing fueled circus at the moment, but I do think the impact on traditional infrastructure roles will be significant over time. I personally don’t think the private cloud is a cloud (I like to call it virtualized infrastructure with good engineering), and like some people would suggest, “not everything is a cloud“. What I’m talking about is the impact of public cloud infrastructure services on general IT infrastructure.
So here’s a preliminary speculation on the cloud impact to infrastructure roles:
- Tech Cycle One (2010-2012)
- Storage cloud services evolve for archive, file, and disaster recovery
- Compute cloud services mature to data center ‘friendly’ offerings
- Flurry of device and software entries to market to solve ‘first mile problem’ associated with cloud archive and storage (this is something I plan to write about soon)
- Fewer system administrator type roles required for small-medium businesses and startups (via cloud compute)
- Early adopters of large enterprise storage and compute services – primarily x86 and secondary storage, but limited impact to headcount
- Tech Cycle Two (2013-2015)
- Storage cloud services mature, consolidations and major acquisitions begin
- Compute cloud services become standard for small business and adoption begins in mid-market
- Significant reduction of standard back-office support (files, email) for small-medium businesses and startups
- System administrator type roles required for small-medium businesses and startups become less common place, in some cases not required at all
- Initial impact to large enterprise storage and compute services – primarily x86 and secondary storage, possible redux in skill requirements
As a practitioner, this is a good time to track the trends and stay ahead of the shift. This one may shake up the status-quo management model we’ve enjoyed for the last 15 years in distributed computing.
Tagged with: cloud computing • infrastructure • public cloud infrastructure
Nov 19
A few weeks ago I got a chance to check out the 
Global Switch data center in Sydney, Australia. Very impressive place.
First off, the location is about a 10 minute walk from the central business district of Sydney and a few hundred meters from Darling Harbour (tourist mecca). Not exactly a low-rent location or the most intuitive place to find high-end high-volume data center infrastructure.
Turns out the building historically served the wool industry as an inspection warehouse and was up for sale and was bought and retrofitted by the Global Switch team into a world-class data center. Again, not your typical estate choice for high-end DC infrastructure.
So many things impressed me with this building and stand out from most data centers you see:
- Multiple floors – this place is a serious four-layer cake of raised-floor data center space, the place is BUZZING with white noise and ‘electric hum’ as they put it. You’re average data center is a single floor in a building, or a building built to purpose for data centers, but rarely an entire building retro-fitted into multiple floors of data center. Interesting architecture and design and not a trivial feat of engineering to pull it off.
- Lease Holders Only- only available to DC ‘lease owners’ or systems integrator/technology businesses and not to direct customers. Reminds me of feudal Europe land ownership where the Kings grant the land, the Nobility set land owners setup the roles, and then lease to the serfs on conditions and terms specific to their space. This is a really good way to keep the riff-raff out of the facilities and run a ‘tight ship’ so to speak from a security point of view.
- Metro location – literally in the heart of a major metro area city, not in the low-cost outskirts like most data centers old and new. Plus, would a data center metro support production or DR or both? Both- which is really interesting given the proximity to the city.
Tagged with: carrier neutral • data centers • metro locations • sydney
Oct 23
It’s not often the EMC markitecture spiel breaks from the well worn rut. But Frank Slootman talking straight about the newly formed business unit at EMC is refreshing after years of EMC changing the deduplication party-line, spinning vague technical slants on product lines, siloed sales teams pushing all kit in all places, and generally confusing customers about their capabilities with deduplication.
In 2007, the stories of ‘our deduplication is in development’ where trumped in 2008 with the Quantum partnership announcement, and once again with the Data Domain acquisition in 2009. Finally, someone has a leadership position and is potentially leading a breakdown of siloed product lines (I like to hope). Frank may not have a lot to loose, but regardless I like his style and hope it works out for the sake of a lot of good technologies and people who build them.
Some key aspects I happen to agree with from the recent interview:
- ‘Let Disk Library be a VTL’ – Hallelujah- Franken-VTL will be no more. Bolting on an inline deduplication device behind a high-performance VTL with 4 times the throughput just didn’t make sense. This also will clarify to existing CDL/EDL customers there are 2 simple choices: High-speed non-deduplicated VTL – use the EDL/FalconSTOR. Deduplicated disk, use Data Domain.
- ‘Data Domain plays in the core, Avamar at the edge’- Absolutely. Both are fine products but have a different role and capability in the enterprise. Avamar can be beautiful for remote sites and relatively small backup payloads and supports a wide range of topologies for replication. Data Domain integrates with existing software infrastructure and scales in a completely different way (more suitable to the enterprise core).
- A hard-line regarding Commvault – Agree. In fact, for a company that has grown into the enterprise market with a solid product known to actually work as advertised, I’m skeptical of the aggressive play to take down the target-based deduplication market with software-only deduplication architecture. Sure, it works, but just because my car is drivable doesn’t put me in pole position for Indy. Where are the performance specifications and benchmarks, where are the design guidelines, and most importantly how do you break the cardinal rule of mixing heavy I/O and compute workload on media servers and magically whip the technologies that have struggled to scale deduplication compute and metadata scaling into the enterprise after 6-10 years of R&D and field experience? Also, why bilk customers who want to write to a non-Commvault disk device, when Symantec already tried that and royally enraged their customer base?
- An integrated line of business for backup solutions – Makes sense. When you see Avamar being positioned in the field for enterprise core backup, 3D3000’s pushed in one account, EDL’s in another, Networker’s solid development roadmap vs. perpetual field support challenges, and Data Protection Advisor being pimped like a utility, but not strategically positioned, you really have to wonder with all the guns blazing is there anyone really steering the boat here?
Deduplication Straight Talk from inside the Sausage Factory…IBM- time to take some notes on this one.
Tagged with: commvault • deduplication • emc • strategy
Oct 19
There’s an interesting trend with storage infrastructure outsourcing, where outsource vendors on-board capital assets and then gradually run them into the ground. You see this over and over, especially in long-term out-sourced environments (IBM, EDS/HP, Perot/Dell,EMC, etc.) – the onboarded hardware/software estate (which is rarely in great shape) evolves into a living graveyard over 3-5 years. The root cause boils down to contract economics:
- Outsourcer strikes a deal, commonly ‘your mess for less’ and onboards assets and management responsibilities
- Contract is structured around data volume / growth metrics (nothing to do with efficiency)
- Legacy designs are grand-fathered in and ongoing growth is cobbled onto existing asset base
- Over time, the asset base ages, hardware reaches end-of-life and software runs out of support
- New technologies mean architecture and planning work, not typically budgeted for as ‘project work’ in outsource contracts
- New technologies require upgrades and migrations, again not typically budgeted for due to competitive price pressures
- And time marches on and the outsourcer keeps up with operational needs, and architecture/planning is best-effort based
The problem with offloading infrastructure assets to an outsourcer is that there’s absolutely no incentive for the outsourcer to run lean or maintain a modern estate. The result is at first, negligable, but over time incredibly risky. Here are some observations of the down-side to letting your infrastructure run into the ground:
- Operational risk – mean time between failure metrics are only published for the duty-cycle / life of devices, not the after-life!
- No disaster recovery – when your asset base is out of support, non-standardized, and new ‘0ne-offs’ are added to the mix, disaster recovery is a long-shot
- Skills drain – so if the people supporting the living museum of hardware/software never learn anything new, are you really getting service or are you paying the bill for ‘not so innovative’ people and services?
- No way out – there is absolutely a breaking point where upgrading no longer becomes an option, and your only choice will be a costly and potentially complex ‘green-field’ deployment
Bottom line – it’s best to own your infrastructure and pay other companies to run it. If you can, hire good architects and planners, and worst case, hire them on a project basis. This way, you can at least control the design and potentially avoid a mess, while applying contractual pressure to the outsourcer to innovate and use modern technology.
Tagged with: capital costs • contracts • infrastructure • outsourcing • technology obsolescence
