Feral Technology

Necessary ingredients for an IT solution architect

So Solution Architect is a job description that covers a multitude of sins, I'm taking it to mean the lead designer for some kind of IT project.

Usually for me that means an infrastructure project of some kind (either end-user or working in a professional services organisation). I think it's applicable to the software architecture people as well though.

Technical Depth

We all came from somewhere, you'll have a home territory, be it networks, storage, OS or wherever, that you're most comfortable with. Stay awesome at it, it's the lens you use to understand the rest, so keep it shiny.

Technical Breadth

Knowing how everyone else's stuff works. I'm not going to list it, but you should be able to have a conversation with almost any SME and be able to ask sensible questions and understand the answers. Knowing what the rest of the world does is important.

You can't know everything though - I tried to make a simple diagram of the technology stack we use, and narrowed it down to the fifty six key competencies and technologies you had to know about to grok the whole system properly. 

Knowing what you don't know and who to ask is often enough.

Experience, scepticism, and a great big bullshit meter

You have to have seen a few good and a few bad projects and shops and salespeople to have a good baseline. Ten years in the industry seems to be about the minimum to develop this, but YMMV.

Knowledge of your vertical

You need to know what it's for.

This is why it's hard to hire SAs as an end-user, knowledge of your vertical market is a big differentiator. Having worked in lots of verticals and knowing what changes between them helps. Ex-consultants, contractors and PS people have an advantage.

Knowledge of your company or client

You need to know the capabilities, budgets and politics, basically. Who's paying for it, who's building it, and who's running it.

Knowledge of the wider business world

You need to be able to talk to business people, customers, salespeople and other non-technical types and ideally add value to those conversations.

A great network of SMEs

Subject matter experts, you need them, be nice to them, help them when you can.

A great network of vendors/channel contacts

The people you get quotes, discounts and gossip from. All the components you design with have costs, lead times and foibles, they're not published, you should know all of them and these are the people who help you to stay on top of it.

Drawing skills

I've known a couple of very verbal architects who wouldn't draw, for some reason they had poor success getting their ideas implemented. Good diagrams are the basic metaphors and visuals design conversations are based on.

Writing skills

With paragraphs, structure and everything. Core skills.

Speaking and debating skills

You're going to have to present and defend your stuff, helps to be comfortable talking in front of a crowd, better if you enjoy it.

Comments are open, apologies if it seems a bit rambly, it's quite late here.

G'night.

 

Never mind the tiering - whole databases on SSD

Had an interesting discussion today, and the upshot is that I'm looking again at the possibility of sticking whole databases onto enterprise flash drives. I like this for a few reasons:

It's simple - easy to understand and explain, particularly to conservative operations folks. It's like running on FC disks, but much faster.

It's predictable - FAST, FASTCache, controller based flash caches, server side flash caches are fine, but in cache miss cases you're going to hit spinning disks and a penalty. This risk can be removed by making everything fast.

It doesn't need a bunch of software licenses - on EMC systems (probably others, not picking on EMC just know it's true there) at least, I'm going to be paying for software licenses for tiering or caching, and these costs aren't subject to the sort of price erosion that commodity storage is (especially SSD). @ianhf made the point that vendors do like to shift costs into software when hardware gets cheaper, I can avoid this and link my costs back to SSD pricing.

Now, this is quite expensive, but for important systems or systems with expensive software licensing dominating it's viable. The 1 TB SSD storage for the database will cost less than the clustered database servers talking to it, and far less than the database software licensing.

Is this the way forward? Anyone else thinking of doing this?

Vendor Blogger Survey

After watching twitter for a bit, I noted that it seemed to me that some vendor folks are very anti-competitor rather than pro-own-company.

This is tiresome. Anything any vendor employee says about a competitor is automatically viewed with suspicion and isn't believed. There have been some quite aggressive tactics recently, smearing competitors doesn't make it likely we'll want to do business with you.

An example of the right attitude came from one of our supplier account managers. He was in an rfp against another supplier when news came out that the competitors kit had had a high profile public failure. We asked him to comment, perhaps expecting some crowing or spin. The response along the lines of "Thank god it wasn't one of ours, it happens to all of us".

Good answer.

Anyway, intrigued I looked at the last twenty posts from vendor employed storage bloggers and noted them as self (about vendor) or other (dissing the competition).

Results

EMC:

Chad Sakac (@sakacc): 20 self, 0 other
Chuck Hollis (@chuckhollis): 18 self, 2 other
Barry Burke (@storageanarchy): 15 self, 5 other

NetApp:

Val (@valb00): 18 self, 2 other
Alex McDonald (@alextangent): 10 self, 10 other

HP:

Marc Farley (@3parfarley): 16 self, 4 other
Calvin Zito (@hpstorageguy): 17 self, 3 other

IBM:

Barry Whyte (@barrywhyte ?): 16 self, 4 other

Oracle were 100% self, didn't catch the authors name though.

I'm on my phone at the moment, so no great conclusion, but there's my data...

Oracle licensing changes - a tale of two RISC businesses

So, on 6th Dec 2010 Oracle changed the licensing multipliers for two CPU architectures. These apply to Oracle enterprise software, including Database, WebLogic Server and an increasing amount of the portfolio.

SPARC - the multiplier for Fujitsu SPARC has dropped for the latest chip, by 50%. This means that licensing an Oracle M5000 with the new 2.6 GHz quad core SPARC64 VII+ is 50% cheaper than the existing 2.53 GHz SPARC64 VII. Customers with older M5000's (initially came with a 2.15 GHz dual core processor) have an attractive upgrade and consolidation path.

This provides an easy business case for one last round of big SPARC iron. If you're running Enterprise Oracle DB and RAC the software licensing is generally a lot more expensive than the kit, even on proprietary RISC gear. These are pretty long lived systems, so the upgrade drives a another round of spending on high margin RISC gear and locks in another five years of maintenance revenues and defers the migration to Linux/Xeon for quite some time. Nice one, Oracle.

Itanium - the multiplier for Itanium CPUs doubles for new processors. The only people hit badly by this are HP, HP-UX customers running Oracle are faced with a doubling of license costs when they upgrade - this is going to prompt some serious thinking. At this level of cost Exadata's become competitive and a move to Linux/Xeon is bordering on a no-brainer - the high margin HP-UX/Integrity business is going to take a kicking, it's a legacy business, but a profitable one. Evil one, Oracle.

Oracle tends to view businesses (and customers) as a series of cashflows over time, this move enhances the cashflows from the SPARC business, and it's at the expense of Oracle software licenses, which sends the signal that Oracle are serious about their SPARC business. The Itanium move is likely to play out over the next 2-3 years and rob HP of a lot of the next round of HP-UX/Oracle hardware upgrades.

And IBM? Oracle already charge a lot for P Series kit, and IBM also have a database of their own. HP without a competing software stack aren't really in a position to do much about this, other than try to push their customers onto lower margin HP kit.

So, evil, but clever. That's Oracle!

 

Things I'll be writing about...

A few topics I had jotted down to write about. I won't write about them now, it's very late and I'm composing on an iPhone...

"Bad Equilibrium" in the workplace

The cloud's "Infrastructure Administrator" - the skillset and organisational challenges around tooling up to do private cloud.

Beliefs that people hold, while still working together.

Future of client computing, having been a VDI user for two years now.

Comment around news and press releases.

Things I've learned from my MBA, mainly things that weren't the "learning outcomes" the course authors were pitching for.

Introduction

Hello. First post, so a little introduction. I'm Martin - I'm an IT guy, and latterly a business guy. I like the combination, and respect that lots of other people really don't.

I was a freelance Unix contractor straight out of Uni, did three startups (crashed all three, but learned a lot) and fourteen years later was tempted into permanent employment by a payments company that interesting enough and doing important enough stuff that even with the attention span of a weasel it keeps me occupied.

On the technical side, my evolution has been Unix since 1992, specialising into Solaris, specialising into high end clusters and the like. After rates went through the floor in 2003 I jumped sideways into storage, jumping around design and implementation. 

I quite like working with vendors, partly because they're interesting and partly because they're funny. I've worked for vendors, disties and reseller - the whole grubby channel - and I sympathise with all of them.

At the moment I'm in the last year of a distance learning MBA, which occasionally provokes me into thinking about what I'm doing. 

So. I'll be sharing some of those thoughts, other things that occur to me, views on what's going on in enterprise IT and probably some twitter related gossip. I discovered Twitter about six months ago - it's good, with caveats.

If this works out, I'll upgrade the TypePad account, but it's freebie formatting until then.

All this is my own thoughts, and don't represent any positions of my employer, and possibly not even myself once I've thought about it.