Amazon’s Virtual Private Cloud

Amazon has released their Virtual Private Clouds allowing people to extend their network and services to Amazon’s cloud services. Werner Vogels (Amazon’s CTO) describes VPC in his blog. The Virtual Private Cloud is particularly important because it opens the doors wide open for Enterprises to use cloud services. It gives them the ability to embrace cloud services without requiring the sophistication of abstracting their applications or re-writing them. Let me explain.

It seems the majority of people adopting Amazon’s services up until now have been web companies who saw Amazon’s Elastic, internet-facing infrastructure as a way to achieve scale without redundant internet connections, BGP, HA’d firewalls, load-balancers etc etc. It was internet-facing and this is exactly what they needed. However the Enterprise was left at the door. I believe even the most avant garde IT manager ruled Amazon out for a number of reasons.


  • Enterprises have applications that run on private networks. They are not internet facing.
  • Security is an issue and having a thin layer of firewall services is operating on the edge.
  • Anyone can attack your decision to host Enterprise services in the cloud using the classic FUD.
  • Integration of legacy systems or interaction with other Enterprise systems.


Amazon’s VPC addresses these concerns. No Longer are systems sitting on the internet they are sitting on a private network that can only be routed to and from the Enterprise network. Network access can be governed via corporate firewalls, visibility of networks can be governed via Enterprise routing policies. It reflects a paradigm that most IT managers already use for communication between primary and secondary data centres when WAN links fail – Internet VPNs.

You can now carve off some IP address space from your internal network, host it at Amazon and redistribute the route into your Enterprise network so Users and other IT systems can access, replicate, integrate etc. Impressive.

Amazon should have called VPC the ‘floodgate’ as it is now a real no-brainer for an Enterprise to start mass adoption of services from Amazon. I also predict that the early movers will be people replacing disaster recovery environments with DR environments hosted at Amazon. This would allow them to utilise their DR environment as a new production environment (to achieve greater scale) or reduce their operating costs.

Think of something as generic as file servers base on an Open Solaris ZFS file system. Snapshots can be created and replicated from the Enterprise to a system in located on a private network at Amazon. They no longer have to be encrypted and sent to a server located on the ‘public’ internet. They can be copied from a file server deep within the Enterprise to a network that no one has to know is located in the Cloud 😉

What is Cloud?

You hear this quite a bit certainly from Infrastructure people – What is Cloud? How do you define Cloud?

It’s a really easy answer and I believe these series of headings are a good basis for it.


Probably the first rule of cloud is elasticity in the sense of dealing with scalability to meet demand and also when demand falls away. Instead of procuring resources to meet the peaks but then operate at a lower mean ‘cloud’ enables people to grow and contract. This is obviously found in Amazon’s EC2, Azure and Google App Engine but also in Amazon’s Hadoop service.

If you aren’t elastic then you aren’t Cloud.

It’s Month-to-Month

A little weird as a second point but the thing I love about Cloud is it is month-to-month. There is no commitment beyond me paying for what I use. If you look at the Cloud as a deregulated resource market where commodity services are traded in short time frames and commitments are on a usage basis then you get my idea.

Cloud means less commitment and I believe this drives efficient use of resources. Squirrels store nuts for winter – people don’t.

Standard Interfaces using an API

Further to the point about deregulation how this was actually enacted (and done perfectly by Amazon) was via an API and a tool-set. They abstracted their resources, presented them via an API and then provided some tools to use the service. The ‘portal’ came later – initially it was a tool-set used via the command line.

There is beauty in this. The abstraction of the interface simplified the access to the resources. You can fire things up and shut things off. You could experiment, simulate, practice and then turn off an entire environment. Then you’re done.

Eventually Consistent

‘Eventually Consistent’ is one of my favourite themes and I was put onto it when CouchDB first released. Whether it is Big Data (datasets) or large amounts of file storage the Cloud is replicated and eventually consistent. I suppose what I like about this is that replication is performed as part of the Cloud and as a user I don’t really need to know about it. I also like the scale that eventually consistent architectures give you. You have a big application? How about you spread your data out across the globe and service the user out of their closest geo? Why stop there how about you replicate entire data sets (over time) and put the data and the files close to the user. CouchDB takes this a step further and allows topologies where you can maintain equipment within you’re organisation and still replicate to the Cloud.

Truly Massive

The Cloud exhibits truly massive behaviour and this is required to handle the exponential increase in data – files, databases, personalisation data etc. Hadoop and HBase are awesome examples of people resorting to Online (HBase) and Offline (Hadoop) methods of dealing with Truly Massive data. S3 is and Azure Blob Storage are examples a file level.

The Politics of The Sneaker Cloud Pimps

I think there is a real danger in people thinking that ‘Enterprise 2.0’ is the Cloud. You also see the muddying of the Cloud water with the term ‘Private Clouds’. So you re-badged your CIFS storage as a Private Cloud – well done you’re a winner. And virtualization isn’t ‘Cloud’ either (supports it Yes). Eucalyptus on the other hand is a truly awesome initiative that everyone should get behind. It took what Amazon had done and extended the nomenclature so that everyone could implement a Cloud.

I suppose what I am getting at is Cloud exhibits a series of characteristics that when they come together solve problems. You can’t take the file storage without the API or the replication (it’s Cloud!). You could do elastic virtualisation but how do I process this dataset in minutes rather than hours? Or how do I store a couple of terabytes of data using this elastically virtualised platform? That’s why I like Eucalyptus, it has the virtualization (XEN/KVM), the S3 compatible storage and also Elastic Block Storage.
Anyway that was my opening gambit so that I can get @cloudpimps off my back and say that I started blogging. My focus on this blog will be alot more on the kinds of characteristics in this post. Investigating the different areas of Cloud, adding to them, finding cool new ways to solve problems – and stay well and truly out of the Enterprise (it’s a dinosaur you know?) 😉

Coming soon to a Thin Client near you… Wyse V10L (WTOS) multi-broker options!

A lot of customers that are evaluating VDI are probably looking at the Wyse (WTOS) V10L series of Thin Client. Wyse have a kick-ass little unit in these devices, and I am sure their sales are ticking along, as the ease of management for these puppies is well…. easy!

Getting back on point, The V10L does have the odd limitation, and one is that it’s not using a full VMware View client, its their port / interpretation of what features they think customers will want the most. And thus, there was only an option for a single broker VDIBroker= <set in your WNOS.INI>.

While this is fine for most customers, its not for those planning mass VDI deployments with multiple brokers internally. Brokers might be segregated by region, business unit, etc etc… and previously this would have put the poor lil V10L out of a large sale. Customers needed to use a VMware View client (Win32 or Linux variants) to get a multi-broker capable client.

But…. Wyse are listening to their customers and partners ! and just a few weeks back after sending in a Feature Request for a multi-broker supported firmware, they sent me a working beta of exactly what is needed. While it was only limited to 5 brokers initially, this should get upped shortly, and hopefully permanent baked into new V10L firmware releases. This is another tick in the box for the V10L, where it previously had an X.

There are a few more features coming to the V10L firmware that I’m happy about, but I’ll talk about them another day. But I guess what really pleases me is that Wyse take their feature requests seriously, and their turn around times from request to private beta has been outstanding, sometimes in less than 48hrs!

Lets hope Wyse keep up the development for these little units !

VMware View “Broken pipe” error

This is a quick FYI for those running VMware View Connection Servers(Brokers/Managers). Regardless of whether you are using ‘Direct’ or ‘Tunnelled’ connections you may see these messages in your logs similar to the ones below with sequential request numbers.

(Request112) Request failed: Failed whilst returning body: Broken pipe

(Request111) Request failed: Failed whilst returning body: Broken pipe

The error messages are generally caused by an ADC or Load Balancer that is polling the Connection Servers web server, and forcing a close of the client connection. This is not something that the View Connection server is expecting to handle, so it dumps this message to the logs.

Essentially, its a false-positive in most cases, and I have heard on good authority that the VMware developers will change the log messages to something a little more friendly in future (~View 4.0).

If you’re seeing loads of these errors and your not using a Load Balancer, then you could have some clients out there that are pushing load generator scripts at your brokers, or potentially some idiot trying out his new script kiddie util. Check out vAudit from Richard Garsthagens site here — if you want more detail around successful and unsuccessful user logins. Its a great utility for those running View 3.x environments.



Intro’s and what not…

Welcome to the 1st post from the Cloud Pimps…

Initially I’d like to start by covering off what we’re all about… and that is T&T. Tits & Technology… yeah it may be a bit sexist, but that was our mantra over a decade ago… and while it still has some element of truth, we lean more towards the technology side these days.

Going back in time, the two of us created a very successful site ( about the former topic I mention above which has now gone by the wayside – if you’re interested to see our efforts, check it out via the waybackmachine, most of the links are probably broken now, but anyway — that was the past. This site will be about the later topic, but you might hear about both intermixed in our posts depending on our mood.

We aim to cover many topics, but we are primarily focussed on Cloud computing, Virtualisation and all those sticky grey matter areas in between.

If you’re keen to follow us, you can catch us both on twitter: @cloudpimps & @cloudjunky !


Like a rhino…..