Tuesday, July 19, 2011

Third world networkers guide to storage in the cloud

So i have been spending alot of time trying to get some light through the data center darkness in my head. I'm proud to say I can guide myself through that maze for about 3 meters before the evil dark lords show up. Every time I open the Data center stuff, I feel like I know nothing, there's vmware that I've used for ages but now has far too many fancy facelifts, storage is a pain I'm wishing would go away everytime I open a book on it -they are all different.

I have concerns, some small some huge. That the entire DC network model is 'flat' worries me. I've spent years learning to get as much as possible to layer 3 as possible. All of a sudden I have to deal with all this layer 2 mishmash competing for space in a single data center. I also get the sense that standards bodies especially for storage are way way behind or playing politics or waiting to see what techology and or vendor most customers adopt. needless to say Im quite silly in this area.

Lastly when I look at all the capacity - processing, memory, bandwidth etc going in there (think several cisco VBlocks - add a multiplier of more than 10) you suddenly realize the sales strategy to sell this has to be solid. I also decided tha t electicity costs more than bandwidth and storage combined.

Cloud computing success relies heavily on high-speed bandwidth. Whether streaming movies, backing up your data or running applications from the cloud, the ability to get data quickly from the cloud to the computer , phone or  (Insert fancy gadget name) is a key requirement for rapid adoption. 

Our access is primarily 3G, followed by 'others'. Bandwidth to the home has absolutely not kept up with another key ingredient for cloud storage/backup success: hard drive prices (depending on where you buy the drives), It is still way cheaper for most users to store backups at home. Obviously you can tell I'm thinking about mass adoption for cloud destined backups.
The Ramac weighed over a ton and was delivered via cargo airplanes. 
If bandwidth prices were to drop at the same rate as storage prices, I'd probaby have 800Mbps for less than KSH.10000 today a month. If someone were to then sell me a diskless
 workstation, who knows, I would probably consider it.

The hard drive IBM shipped in 1956:

* Stored 5 megabytes (MB)
* Cost $11,000 per megabyte
* Was 60 inches long x 68 inches high x 29 inches deep
* Weighed about 1 ton

In today’s dollars that would mean:
A $179 16 GB iPod Nano:
* Stores 3,200 times more data
* Would cost: $1,429,176,320
* Requires 8 semi-truck shipping containers to hold the data

A petabyte of storage would:
* Cost: $93,662,499,307,520
* Require a building the size of 10,814 football fields to hold the drives
* Require 472 of the world’s largest data centers to hold the drives


Source: http://blog.backblaze.com/2011/06/21/94-trillion-petabyte/

Sunday, July 17, 2011

third world networkers guide to a mobile workforce

The Monster 696 - 'my dream-mobil'
For a couple of months now I have depended on a neighbor /workmate for my daily commute to work - being without a car and all. One of the advantages/disadvantages is inadvertently work makes it into our 'in car' conversation.

I worked the last project with him, his expertise being in transmission. I learnt a lot from this guy. He ran circles around most of us when it came to DWDM and was highly instrumental in our CRS-1/7609-s DWDM 10G core transmission strategy and its success. Some of our sites have well over 50G(bps) in capacity.

So anyway, the guy goes like - hey after all this work we have done, how come we still can't work from home. Being one not to lie about stuff like that, I outright mentioned that I do at least 30% of my work at home.

While I was indisciplined before, working on the CCIE ensured that I have a comfortable office at home, and a very well set routine. Infact were it not for meetings and factors that I outline next, I can easilly do 90% from home, probably deliver more while at it.

Now my work requires bursts of concentration, just 2 hrs a day is enough to come up with draft technical documents, another 2 hours to read through team submissions if any and the rest of the time is ideally spent critiquing design points with the team or alone. (white boarding and discussing various points is the best part and it ensures we're on the same page). This by the way can and should be offsite. It should also be done often and probably be mandatory for a design team to meet at least once a week to brainstorm ideas.

Good planning includes setting commitments, responsibilities, measurable goals, objective metrics for tracking, and following up on all of this with an actual review. Tracking and reviewing measurable performance factors leads to accountability. With clear guidelines and expectations documented (probably signed off too), and the right technologies, anyone in a technical field especially the creative end of things - programming/design can pretty much work from anywhere.

So in conclusion that conversation had us come up with draft task lists, proper deliverables and ways of dealing with requests as they come to the team for instance someone has to take ownership of all meeting requests, email requests support escalations etc. There might also be need to have a 'weekly mandatory must be in office person'.

Most of this is still an ongoing process and we have also realized how much actually goes un-done even when guys come to 'work'.

In summary -
  1. Set very clear goals and tasks. (run everything like a project no matter how minute).
  2. Measure them with clear deadlines - throw in a a line like, if you miss any deadline, YOU HAVE to WORK from the OFFICE daily for a month.
  3. Set up a team mailing list or portal (preferred), you need a tool to track things.
  4. Ensure someone is responsible for adhoc requests. 
  5. All meeting requests should go to the team not individuals, The team should have a way of ensuring attendance and post meeting sharing. Some meetings should really go unattended, having an evaluation criteria shared with everyone would help ensure you're only invited to meetings relevant to your team.
  6. Let everyone know that 'the team' works together and share your plans.
  7. Set up a collaborative portal or method of remote sharing stuff. We for instance have or are in the process of launching some serious teleworking solutions designed by us, I say we eat our own dog food on this one.
Refine the plan as you move along....either way I hope in the end to have the culture that I must be at my desk slowly change.

In planning, the annoying bit is our processes always have an 'input', that which triggers a design change, it is on a rare day that we get accurate inputs. With that in mind, our/your planning is about the interdependencies,linkages and coordination of the different parts of the network/business such that having a plan makes dealing with the unexpected much easier, not harder.

Good planning reviews results - a stable easy to manage network and assumptions regularly, If you feel like things are too quiet, maybe ask to see customer's every once in a while. Their crappy networks should keep your juices flowing. (please note doing a good job here can get you fired since at some point people start wondering what the heck you do, tell them the networks stability is your KPI:-), tell them they owe you.

Trust me! reacting to the unexpected when there is no planning for each and every command run on the network, documentation and anything that goes hand in hand with a well designed network is harder than doing it right the first time. That and an idiot can reverse all your work in a second - literally (for instance lock all Route reflector configs change control should be much stricter there).

If you succeed to even get 50% time approved to work remotely, please share with me how you went about it.

Friday, July 15, 2011

Third world networkers view on the data center CCIE - if it happens

There is also a rumour of a ccie data center that if confirmed will definately get done.

Wednesday, July 6, 2011

Third world networkers wish for marketing departments

Life is hard, margins are shrinking

Oh well ... I'm seated here having done some kick ass projects. Some were/are fun, others not so much. If we were a factory producing raw material, we have produced it in excess. The capacity is there but hmm where are the damn users.

Let me re-tell this with an analogy of sorts. Most of our parents by the time they could afford to build or buy houses went out and bought huge/built 7 bedroom maisonettes hoping you and your siblings would live there forever.

Unfortunately for them you moved out/on , so did your siblings. Now they have all this idle rooms. A few entrepreneurial types converted the houses into guest houses and are making steady income from the house, others need someone like me- a true mugikuyu to talk to them about how to maybe get a return on that investment.

When you build a network or a data center and for some weird reason the dimensioning data came from somewhere other than actual customer requirements ie you make some assumptions, look at the budget and buy the biggest boxes, you risk having lots of idle capacity.

But what if you have a marketing dept. that looks at this capacity, adds value, packages it and sells it?  I think a really good marketing department should help planning and engineering teams figure out how to deliver product(s) that customers need and want. They should be directly involved in service definitions.

It starts with a deep understanding of what customers need and making sure the planning and engineering is getting continuous customer feedback and interaction data.  They also need an understanding of what we are building, how much it cost, how to maximize revenue off it maybe be by branding or product extensions (branding extension is for instance giving two customers 1Mbps links but throttling one link to 128kbps and selling it cheaper).

Thats the kind of marketing department (in whatever form) I'd like to work with in a technology company. In most such companies, engineers produce raw materials, marketers should take those and sell products. If you are in marketing, head over to engineering and look at the raw materials.

What I know from running a few businesses or seeing them run is in the end it is always about sales. The money. You can bullshit all you want but that is the bottom line.

So if for instance you outsource marketing end to end including product creation, packaging etc without having the expertise to judge or manage the results, you are pretty much screwed, learn the metrics to keep track of, your engineering team knows and can help. Be 'with it'. (My opinion has always been - outsource what you are an expert in unless you want to get conned').

Now what brought on this post????????????

Third world networkers guide to successful an almost succesful service model (cloud)

So there is a huge debate over at skunkworks mailing list home of an elite technical group. Someone raised a storm with some statements about cloud computing.

Now the term cloud has been abused several times. Over and over people keep bringing it up in contexts that I don't feel fit the word. Im not trying to clear that. Not today. In fact this post has nothing to do with cloud. More on that some other day.

I love the entire idea behind cloud. Public cloud from the likes of Amazon have been a huge game changer in the startup arena. I use the AWS for some services.

Private clouds will definately make a huge difference in our lives in Kenya if well priced and positioned otherwise I might very well use a public cloud elsewhere. I doubt a public cloud platform will happen soon here. I also believe cloud is not the main issue at the moment. Awareness is.

Security is of huge concern for me. I am especially scared of handing my data to a service provider that 'exports' my data to some 'cloud' hoping someone over there will keep it safe.I get scared if said person doesn't even ask me before handing the data over.

Another challenge is changing how managers and project managers think about processes.Cloud computing breaks most of the models in use today for a service portfolio. I can already see some confusion in meetings I attend. So for instance if you came from the ITIL Prince2 school of thought, the vertical blocks (cost analysis, business plan, service overview, technical plan, implement,operate) can't/won't scale.

Cloud computing gives you layered portfolios. Visualizing this can be a pain to comprehend for even experienced users. So for that mailing list the question really is how exactly is a novice regular user supposed to comprehend it and take advantage of it. How do you show value?

Start small, start by creating value services, created demand for a service, be open to audits, to discussions, let your technical people talk to the customers (note I said technical not your sales drones).

Personally I'd start from small communities like an estate of lets say 50 houses,Ensure each and every house has access to a 'community network either cable or wifi or whatever, show them how voip between your home and a sentry gate with a few IP camera's and community wifi can secure the 'estate'.Let kids have online multiplayer games, host a local movie database, or music, or photos, install a microphone, record your amateur guitar, stream it to your neighbor....

Have an estate site where users and home owners can access and check on their houses. host community bulletins online, maybe a 'for sale' board online, start engaging locally....and on and on and on....thats the kind of activism I want to be involved in...

Tuesday, July 5, 2011

Third World Networkers guide to 'culture' - not mine - netflix's

Third World Networkers guide to 'startups'

Everyone In Kenya tends to be involved in one form of business or another. Whether you know it or not, all those 'trials' are startups. Most never make it. Sad.

I have no illusions. A startup is not yet really a business. it's more like checking out the market, getting customers, working towards structures trying to figure out a repeatable business model that when finally found should be able to replicate and keep growing ideally without much of the founders day to day baby sitting.

Since I work for a tech. company and really didnt feel like muddling the 'disclosure' waters, our first startup was a farm:-), because I love being in one. We've done some really crazy experiments, figured out how the local vs export market works for passion fruits, tomatoes and green beans (michiri:-)).

I know what a charcoal cooler is, I know how irrigation works, we have customers. Things look good over there. But how do you start transitioning this to a 'business'. It's two years down the line, the structures and processes are in place and you'll be surprised at how much a farming business is alot like umm a tech business (those lessons are coming in handy at another forum right now).

So anyway, we 'the founders/shareholders' had to come up with roles for each of us, remember this is a farm so one of us had to pretty much 'live there', I did the financials and other tech stuff - like arranging for irrigation, automating it where possible, we had one other partner dealing with our 'partners' and labor. It's worked well. We tried to 'follow a typical business' cycle - heck we even filled a business model canvas - more on this later ( and I'll never do any business again without fully going through the process).
Business model canvas - work with it!
Why this, why now - well, there's alot of hooha on startups and entrepreneurship in Kenya today. It's really awesome. Alot of it is noise and people hoping to make things work. Since there's a really high chance I'll be going 'that' way this time for a tech company, It is interesting to try and co-relate the tech and farm.

This will also form a guide on what I finally settle and decide to do academically! - don't ask....:-)The basics are the same.

They are both businesses, they both call on leadership skills - so the leaders ability to articulate the vision,The right kind of ambition and an ability to achieve the said vision. I'm still putting together a proper story (I like stories and believe if I cant tell a nice clear story about my plans, agenda then my vision and strategy are blurred and there's probably going to be a misunderstanding).

I just love the idea of creating a culture, setting clear objectives, manage expectations - all things I have been learning but can't really exercise where I work .... and Im also scared and worried - the usual....

Oh well...i guess just a filler post today....and not quite a guide now is it:-)

Friday, July 1, 2011

Third World Networkers Guide to End-2-End broadband Service

So lets take a typical user on the Safaricom network. He has a mid range phone, lets say a Nokia E72, or an IDEOS from Huawei.

He clicks on the you his 'buy this' Icon, selects a video somewhere on the internet, pays via MPESA, the processor screams and cycles around calls up a couple of routines, fires up the video player or browser, creates a moving object on his screen, he's happy. Now lets take a look at exactly what goes on in the background from the moment he 'clicks'.

For a network operator to offer really great service, they need to have full end to end control of the network, the enduser device and the content. The End 2 End view of the mobile service in this case the 'video' is best premised from the perspective of the user. His/Her ability to watch their movie in a frictionless manner after paying for it is the goal.

From an implementation perspective, severalservice providers are involved:
  • - The Mobile network operator - lets say Safaricom.
  • - The guy providing video,
  • - The guys pocessing the payment
Does the consumer in most cases know of the multiple 'service providers' involved? nope! most don't, in most cases they shouldn't.

For most of us users, the mobile view we hold is that the operator is responsible for the access and content. Unfortunately the Operators 'span-of-control' is often limited to internal content or external content endorsed by the operator eg content served by Mobile network operators like bernsoft for Safaricom. Other than that there are multiple considerations involved in optimizing E2E service delivery.

While  a network operator will undertake alot of effort to ensure a memorable experience to the end user; maybe by using agreements sometimes commercial and standards, or sometimes co-operation with content providers, it is not always the case. Guys like Google and Facebook make alot of effort to co-operate with network operators. Others like CNN; not so much.

Components to consider if you want to have E2E QOS:
  1. The UE - aka handset - Unless its a certified unmodified device, the operator can't claim control here.
  2. RAN - Radio Access Network: On the one hand, the mobile service provider has full control over the nodes that make up the RAN. We now have QOS and standards defined to take care of this area.  On the other hand, there is another aspect of the RAN over which the mobile service operator has only very limited control: the literal, over‐the‐air portion.
  3. Backhaul : If you own it, you control it.
  4. Core Network: mechanisms can be implemented here that participate in network
    management and optimize the QoS since the operator almost always has control.
  5. Operator Owned content like the safaricom portal content: Content includes both the applications as well as any digital media that are part of the customer’s service subscription. For this there is full control by the mobile operator and it is almost always optimized to as close to the user as possible.
  6. Internet and other external content: Only at the point of Ingress/Egress. So if Facebook suddelny goes offline, don't expect the mobile/network operator to know or even care. Poorly written applications can introduce QoS issues not only for the user of the application,but for other users as well. I've seen and had to mitigate this several times on our network.
  7. Branded content: It sometimes happens that external entities offer branded content. For this the operator has limited control.
Also consider mobility,viruses and general end user demand variations. All this add up to the complexities involved in running and optimizing a network.