blog

Windows Server 2008R2 in Production

Wed, 07 Dec 2011 10:08:11 +1100

After several weeks of Beta testing, OrionVM is now unveiling our support for Windows Server 2008R2 on our platform.

Many of our customers have requested the support for Windows over the past few months and we have delivered!

Launching a Windows Server 2008R2 instance is as simple as launching a Linux instance. OrionVM provides per hour billing for Windows Licenses so you now truely have the flexibility of Cloud Computing within a Windows environment.

Pricing for resources is the same as Linux instances, with an additional $0.05 per hour charge to cover the Windows Server 2008R2 license.

If you are interested in testing our these Windows instances, please feel free to give us a call on 1300 569 952 or email us on sales@orionvm.com.au.

Anthill's Cool Company Awards

Thu, 24 Nov 2011 13:39:03 +1100

OrionVM has earned a place among Anthill Magazine’s Top 50 in its 2011 Annual Cool Company Awards, a national program developed in 2006 to recognise Australian companies that are doing things differently to bring about positive change.

Over 700 aspiring ‘cool companies’ were nominated for the 2011 awards, making ‘The Cools’ one of Australia’s largest business awards programs.

What are Cool Company Awards?

The Cool Company Awards was launched in 2006 as a way for Anthill Magazine to publicly acknowledge and celebrate Australian organisations that are doing things differently to bring about positive change. In its sixth year, the awards attracted nominations from over 700 organisations, making it among the largest business award programs in Australia. Judging criteria takes into account more than just business variables, such as revenue and wealth, to include a range of other qualities like culture of the organisation, through to the disruptive nature of the product or service.

We’re proud of our work and our work culture at Outware, so getting named as one of Australia’s ‘Coolest’ businesses in Anthill’s Cool Company Awards is great.

The winners will be announced Thursday 24 November 2011 and published on AnthillOnline.com.

OrionVM Welcomes Angel Investors to the Team

Tue, 15 Nov 2011 14:45:12 +1100

OrionVM is proud to welcome Gordon Bell and Stephen Baxter, two of the biggest names in the US and Australian tech space, on-board as Angel Investors.

OrionVM was founded in early 2010 and introduced their high performance cloud platform in April 2011. The team, a trio of university students (on leave) created CloudDC, an Infrastructure-as-a-Service platform, from the ground up in 15 months. CloudDC has been bench-marked against the leading Cloud platforms worldwide (Amazon’s EC2 and Rackspace’s Cloud Servers) and was independently verified as having the worlds fastest network backed storage performance. In recent months OrionVM has been featured in numerous publications including Channel 10, MSI Financial Review and APC magazine.

Gordon Bell is considered to be the father of the minicomputer. As head of R&D at Digital Equipment Corp, he was responsible for various PDP and VAX computers. He also was responsible for establishing Digital’s first office in Australia. Bell brings to the table a high level of understanding the IT industry in Australia and Silicon Valley as a resident of Sydney and San Francisco.

Along with numerous other Internet related businesses Stephen Baxter co-foundered PIPE Networks in 2001, and in 2005 the company was listed on the Australian Securities Exchange. In March 2010 TPG Group acquired PIPE Networks for over $370M. Whilst with PIPE he was CTO then took on a non executive director role whilst working with Google Inc in Mountain View in 2008 and 2009. Mr Baxter offers the company invaluable experience in the running of a large and rapidly expanding tech business.

OrionVM managing director, Sheng Yeo, says he is ‘proud and honored to have industry leaders like Gordon and Stephen show faith in OrionVM’. Yeo also noted the biggest gain is not just money but the vast breadth of knowledge and experience Bell and Baxter bring to the company in establishing this next generation of computing —we call the cloud. He says the funds will ‘allow OrionVM to double staff, triple CloudDC customer usage and explore an international presence while still delivering the great product and service that our customers have become accustomed to.’

OrionVM services several hundred customers ranging from ASX listed companies to small startups. It’s clients include Wikia.com, Quotify.com.au (owned by Sensis), Getflight.com.au and DesignCrowd.

Windows Server Released into Beta

Thu, 10 Nov 2011 20:41:17 +1100

For the past 6 months OrionVM has prided itself on the performance and simplicity of our Linux cloud platform. Many customers have come to us and asked us to provide a Windows template that performs equally as well.

Well the wait is finally over! After spending many weeks writing an update to our platform, we have now pushed a Windows Server 2008r2 Datacentre edition template into Beta.

The feedback we have received from our initial testers has been great, with performance equal to or better to a dedicated server with SAS disks.

We anticipate this Beta to last 2 weeks, in which we will heavily test the Windows deployment and patch any issues that crop up.

If you are interested in giving the Windows templates a test, please do not hesitate to give us a call on 1300 569 952 or email sales@orionvm.com.au.

We are happy to provide users with credit on their account to test out the Windows template and provide us with any feedback.

OrionVM in the News

Fri, 04 Nov 2011 10:22:59 +1100

It has been an interesting few months here at OrionVM as we continue to develop features for our Cloud Computing platform. We are in the process of writing Version 2 of our platform, and there are also many interesting things in the works at OrionVM HQ.

I can not wait until I am able to tell everyone about some of the great steps OrionVM is taking towards bringing you an even better product and increasing our company size. Over the next few months, we will be bringing on board several new staff, focusing on key strategies going forward, as well as releasing key features.

All the work we have been doing has attracted some press over the past few weeks, so I thought I would share them with you.

Channel 10 Interview

OrionVM was recently (2nd November) interviewed for Channel 10 News, which was a great experience.

&amp;amp;amp;amp;amp;lt;p&amp;amp;amp;amp;amp;gt;&amp;amp;amp;amp;amp;lt;br /&amp;amp;amp;amp;amp;gt;&amp;amp;amp;amp;amp;lt;/p&amp;amp;amp;amp;amp;gt; &amp;amp;amp;amp;amp;lt;p&amp;amp;amp;amp;amp;gt;&amp;amp;amp;amp;amp;lt;br /&amp;amp;amp;amp;amp;gt;&amp;amp;amp;amp;amp;lt;/p&amp;amp;amp;amp;amp;gt;3 &amp;amp;amp;amp;amp;amp;amp;amp;amp;lt;/p&amp;amp;amp;amp;amp;amp;amp;amp;amp;gt; &amp;amp;amp;amp;amp;amp;amp;amp;lt;/br&amp;amp;amp;amp;amp;amp;amp;amp;gt; &amp;amp;amp;amp;amp;amp;amp;amp;amp;lt;p&amp;amp;amp;amp;amp;amp;amp;amp;amp;gt;&amp;amp;amp;amp;amp;amp;amp;amp;amp;lt;br /&amp;amp;amp;amp;amp;amp;amp;amp;amp;gt;&amp;amp;amp;amp;amp;amp;amp;amp;amp;lt;/p&amp;amp;amp;amp;amp;amp;amp;amp;amp;gt; &amp;amp;amp;amp;amp;amp;amp;amp;amp;lt;p&amp;amp;amp;amp;amp;amp;amp;amp;amp;gt;&amp;amp;amp;amp;amp;amp;amp;amp;amp;lt;strong&amp;amp;amp;amp;amp;amp;amp;amp;amp;gt;Dell Use Case&amp;amp;amp;amp;amp;amp;amp;amp;amp;lt;/strong&amp;amp;amp;amp;amp;amp;amp;amp;amp;gt;&amp;amp;amp;amp;amp;amp;amp;amp;amp;lt;/p&amp;amp;amp;amp;amp;amp;amp;amp;amp;gt; &amp;amp;amp;amp;amp;amp;amp;amp;amp;lt;p&amp;amp;amp;amp;amp;amp;amp;amp;amp;gt;OrionVM recently assisted Dell with a use-case about our implementation of their C2100 line of Cloud servers.&amp;amp;amp;amp;amp;amp;amp;amp;amp;nbsp;&amp;amp;amp;amp;amp;amp;amp;amp;amp;lt;/p&amp;amp;amp;amp;amp;amp;amp;amp;amp;gt; &amp;amp;amp;amp;amp;amp;amp;amp;amp;lt;p&amp;amp;amp;amp;amp;amp;amp;amp;amp;gt;This document can be found on the Dell.com site or directly here: "&amp;amp;amp;amp;amp;amp;amp;amp;amp;lt;a title="Delivering High Performance Cloud Services" href="http://i.dell.com/sites/content/corporate/case-studies/en/Documents/2011-orionvm-10009933.pdf" mce_href="http://i.dell.com/sites/content/corporate/case-studies/en/Documents/2011-orionvm-10009933.pdf"&amp;amp;amp;amp;amp;amp;amp;amp;amp;gt;Delivering High Performance Cloud Services&amp;amp;amp;amp;amp;amp;amp;amp;amp;lt;/a&amp;amp;amp;amp;amp;amp;amp;amp;amp;gt;"&amp;amp;amp;amp;amp;amp;amp;amp;amp;lt;/p&amp;amp;amp;amp;amp;amp;amp;amp;amp;gt;

Dell Use-Case

OrionVM recently assisted Dell with a use-case about our implementation of their C2100 line of Cloud servers.

This document can be found on the Dell.com site or directly here: "Delivering High Performance Cloud Services"

The passing of a legend, genius and father of modern computing

Sat, 15 Oct 2011 02:36:55 +1100

On October 8th the world lost one of its most influential and revered geniuses. Denis Ritchie the creator of the C programming language and co-author of the UNIX operating system is no longer with us.

For many his passing will go un-noticed, but myself and many like me mourn the passing of one of computing’s foremost giants, for without Ritchie many of us wouldn't have lived the lives we have.

UNIX and C are close to my heart, having dabbled in them for almost as long as I can remember. What has always struck me is the understated power and simplicity of UNIX's construction and paradigms.
Ritchie leaves behind not only a legacy of many operating systems directly descendent from UNIX or even the vast array of software written in C - more so he leaves behind a philosophy he helped create and nurture.

The UNIX philosophy is embodied in the code of the UNIX kernel, the deliberately simple and precise style he authored his books in and the private and humble manner of the greatest coder ever.

Simply put he left behind the DNA of today’s most successful and software design and development methodologies and set the stage for an amazing 40 years of innovation.

Rob Pike has already done a great job of explaining the far reach of Ritchie’s inventions, but I feel that the gift of the UNIX philosophy and the building blocks of open-source were his greatest achievements.

Rest in peace Ritchie, the world is a better place for the time you spent with us.

Sincerely,

A hacker

OrionVM Service Level Guarantee

Fri, 30 Sep 2011 15:04:50 +1000

At OrionVM we pride ourselves on our quality and reliability of our service. We understand that you run mission critical and customer facing applications on our platform. Therefore should any down time occur, for whatever reason, it can have a large impact on your business.

For this reason we have implemented a Service Level Agreement to back the quality of our CloudDC solution and compensate you, in the highly unlikely event anything were to happen to our platform.

Our Service Level Agreement guarantees the CloudDC platform to be up and running a minimum of 99.95% of any billing period. Should an outage occur, at any time, the OrionVM team will have all of your virtual machines restored within a reasonable Time to Restore period. Should our restoration take long than this period (as defined in the SLA) we will provide service credits to the value of 10% of that periods bill.

This means if the platform experiences an outage of more than 0.05% within 1 calendar month, and is not completely restored within the Time to Restore period you will be compensated.

For more details see the Service Level Agreement page. If you have any questions, please do not hesitate to email us at support@orionvm.com.au.

At OrionVM we pride ourselves on our quality and efficiency of service. We understand that you run mission critical and customer facing applications on our platform. Therefore should any down time occur, for whatever reason, it can have a large impact on your business.

For this reason we have implemented a Service Level Agreement to guarantee the quality of our CloudDC solution and compensate you, in the highly unlikely event anything were to happen to our platform.

Our Service Level Agreement guarantees the CloudDC platform to be up and running a minimum of 99.95% of any billing period. Should an outage occur, at any time, the OrionVM team will have all of your virtual machines restored within a reasonable period. Should our restoration take long than is reasonably acceptable we will provide service credits to the value of 10% of that periods bill.

This means if the platform experiences an outage of more than 0.05% and is not completely restored within a reasonably acceptable period you will be compensated.

Is cloud a commodity ( or will it be? )

Tue, 27 Sep 2011 09:01:33 +1000

As it is often cited many people believe that cloud computing is another step in the commodisation of compute capacity and data storage. By what is commoditization, how does it relate to cloud computing and is cloud really a commodity? ( or will it be )

Firstly the term commodity is well described in this excerpt from Wikipedia:

"A commodity is a good for which there is demand, but which is supplied without qualitative differentiation across a market. A commodity has full or partial fungibility; that is, the market treats it as equivalent or nearly so no matter who produces it. Examples are petroleum and copper. The price of copper is universal, and fluctuates daily based on global supply and demand. Stereo systems, on the other hand, have many aspects of product differentiation, such as the brand, the user interface, the perceived quality etc. And, the more valuable a stereo is perceived to be, the more it will cost."

Now, the important aspect of this passage is that the wares offered, in this case - cloud computing, need to be standardized across all sources in order for a global commodity market to emerge. In it's current state without formal open standards or even agreed industry standards cloud computing is a far stretch from achieving this. There is however some initiatives like the Open Virtualization Format and the Open Cloud Computing Interface that are gaining steam but wide spread adoption has not yet taken place.

Because of this cloud computing is about as far from commodity as it can get currently, vendors offer many levels of differentiation - be it varying interfaces, performance, value-added products and support.

For the sake of argument and pondering of possibilities lets assume that cloud will be standardized (at the container and API layer) at some later stage. If this can be achieved nothing stands in the way of a global market for compute.

Another interesting fact about cloud computing as a commodity is it's more of a utility, in this way it is comparable to power and mobile phone networks for instance. Conversely it is also global, free of the geographical constraints of the two former examples, I can use OrionVM from Bangladesh for instance, but I probably can't use Telstra Next-G. :P

The consequences of this are enormous as it places it in the realm of natural resources like copper and coal - being global markets that intrinsically change in value according to demand, yet still retaining some requirement of locality due to the properties of latency.

But can this last gap really be achieved? Is it possible to standardize soft qualities like customer service and support? How about geographical locality what part would that play considering hot topics like data sovereignty, latency and local support?

I believe what represents the largest barriers to commoditization are what I like to refer to as "People Problems", more scientifically - political, emotional, sociological and business (read profit) problems, whether real or perceived.

Technology has already gotten us 80% of the way there, but the last 20% is going to take 80% of the time as per the omni-present 80/20 rule.

To summarize, cloud is not yet a commodity but it's quickly heading in that direction. Is that good or bad? Maybe the topic of another blog post...

Exciting times at OrionVM - An Update

Mon, 26 Sep 2011 09:25:58 +1000

It has been a busy few weeks here at OrionVM HQ! As many of our customers would know, we have migrated our infrastructure to a new Data Centre, increased the amount of connectivity to our platform and released a new website. Things are moving at such a pace that we often forget to update everyone on our progress and the things to come going forward. I thought I would put together a quick post to outline what has been done and what is on its way.

Over the past few weeks there have been several changes at OrionVM that are a glimpse into what is to come.

Larger Office - OrionVM HQ has been moved back into the heart of the Sydney CBD, on Bathurst Street next to Hyde Park. With the rate the company is growing, we have needed to hire more staff to keep up!
New Staff - The OrionVM Team has grown, making it look like we should have rented a larger office. As we continue to grow the company, there will be many more staff that join the family. A big welcome to Scott, Mary and Mike. We are always looking for great staff, so feel free to ping us if you are interested. We are looking for awesome JS and Python developers :-).
Private Cage in Equinix SY3 - We were quickly outgrowing our rack space in the Vocus DC and migrated all our infrastructure to the new Equinix Sydney3 facility. With our own private cage, we now have plenty of rack space to expand into, as well as increased transit connectivity.
20% Reduction in Bandwidth Prices - The data centre move gave us access to more transit and peering connectivity, allowing us to reduce our bandwidth prices by 20%. We will continue to try to reduce these rates over the next few months.
New Website - We have pushed out our new and improved website which will make it far easier for clients to determine just how much an instance will cost.
Ovm-ctl - Linux command line utilities released for the OrionVM CloudDC platform. This tool allows for the easy automation of infrastructure, and allows System Administrators to access the cloud from an environment they are used to. These can be found here: OrionVM Utilities
99.95% SLA - As part of the new website launch, OrionVM has also published our official SLA document. We back a 99.95% SLA on our Cloud Platform, giving you peace of mind. The SLA can be found here: Service Level Agreement

So what have the team here been busy with?? Without giving away too much, here is a brief list...

Cloud Storage - The team have written the OrionVM equivalent of the popular Amazon S3 bulk object storage, we are awaiting the final hardware (raw disks + a few chassis boxes) before we release this into beta.
Windows Support - The OrionVM CloudDC platform will soon support Windows templates once we have fully tested them. This will allow us to deliver our great Storage I/O performance to Windows servers. Let me just say that these Windows servers absolutely fly!!
API Bindings - We are building API bindings for several of the popular languages to allow you to integrate with the platform with ease.
Many More!

Well that is a quick update from all of us here at OrionVM HQ. If you have any feature requests or suggestions, please do not hesitate to contact us. There is also another big announcement on its way soon, so stay tuned!

Transit Prices Reduced by 20%

Thu, 22 Sep 2011 16:41:59 +1000

Starting 1st October 2011 OrionVM will be lowering our transit prices by 20% to $0.80 per GB.

Earlier this month we transitioned our Cloud Platform (CloudDC) into the new Equinix SY3 data centre in Sydney. SY3 is one of the most advanced, state-of-the-art data centres on offer in the Asia-Pacific region, offering us far better connectivity and peering options than our previous facility. OrionVM has increased the number of quality upstream providers and committed capacity into our platform, and we would like to pass on this benefit to our customers.

We understand that one of the biggest barriers to adoption of Cloud Computing in Australia is the price of ingress and egress of your data, hence we are offering a price reduction of 20% on all transit costs.

We are doing all we can to reduce these prices, and will continue to introduce more peering arrangements going forward. If you are a heavy bandwidth user, please feel free to contact us, as we are able to work with our upstream providers to achieve better rates going forward.

I would like to thank everyone for their support over the past few months. We are working on some great features within the OrionVM offices, and I look forward to releasing them to you.

Many Thanks,

Sheng Yeo

Managing Director

State of Xen and its bright future

Tue, 09 Aug 2011 18:55:09 +1000

For a while there things were looking grim for the Xen hypervisor, disadvantaged by being out of tree and having a large and unwieldy patchset it looked as though the war was lost to KVM. Looking at the current state of Xen development one would never imagine that such troubled times were only maybe a year prior.

Today Xen dom0 support is almost entirely upstreamed into the Linux kernel and development is busy with contributors from all around the world, even big hardware heavy-weights like Intel and AMD are behind the effort to ensure Xen's continued existence. The xen-devel list where day to day development takes place is the busiest I have seen it in years.

Though other hypervisors have their place Xen has for the last decade reined supreme in performance reliability and security and things haven't changed much on those fronts. But finally Xen is starting to tackle the problems that have hindered it's mass adoption for so long, mainly upstream support and a unified toolstack. The amount of expertise both in Linux systems and Xen itself required to obtain a function Xen installation has always been its major downfall and cited as such.

One of the most exciting developments of the recently released Xen 4.1 version was the move to XL and libxenlight as the prefered toolstack, deprecating Xend and XM. I think this is one of the most important steps in ensuring that Xen stays relevant outside of huge hosting and cloud computing plays as the previous Xend stack was a nightmare for less experienced users. This move simplifies many tasks, cuts out complications induced by Xend misconfiguration, removes network bridge configuration from Xen and hands it back to the distribution etc. Behind the unified libxenlight Xen APIs will hopefully become much more standardized and easy to develop against.

Still the future holds better yet, with the upstream effort moving along as quickly as it is the next generation of Linux distributions will all support Xen out of the box, removing the largest barrier to users - having to use a modified kernel. While Linux domU PV support was added in 2.6.36 and PV on HVM along with preliminary dom0 support in 2.6.38. However the required device backends weren't present and as such a fully usable dom0 didn't make the 2.6.38 merge window. Just a few days ago however blkback (the paravirtualized block device backend) was submitted to the kernel community for review and is looking like it will procede to be pulled into Linus's branch, heralding the dawn of a new age for Xen. An age where Xen is now one of the most accessible virtualisation technologies and no longer an arcane technology only wielded by infamous Xen masters. :P ( a little over the top there but a little drama never hurts, work with me here! )

So what is on the horizon? Well for starters you can expect Xen domU support to be available in all of your recent distributions and all next-gen releases will probably be sporting dom0 support out of the box. Libxenlight should enable the development of a whole new class of Xen management utilities now that the obscure Xend-API and associated are now deprecated. Documentation should get increasingly better as Xen becomes more new user friendly. With AMD and Intel putting their weight behind the project, awesome GSoC (Google Summer of Code) projects and other interesting developments the stack will continue to be the most cutting edge hypervisor available.

At OrionVM we are a massive fan of the Xen hypervisor and it's dedicated community of developers. We use it exclusively to power our high performance infrastructure and have drawn on the experience of over a decade of Xen users. In the coming months I hope to contribute as much as I can back to the project in the form of documentation and user guidance on the xen-users mailing list. Xen is the best, used by the best, lets make it the best choice for new users to server virtualisation. :)

EOFY.. the close of a great year at OrionVM!

Wed, 06 Jul 2011 12:21:55 +1000

30th June has come and gone, and another financial year has come to a close. This has been one big year for OrionVM, and as many of you would know, it has been one filled with many milestones for the team.

Not only have we built and released a product, we have also managed to gain a great portfolio of clients! From everyone at OrionVM, we are grateful for all the support our clients have shown us. We couldn't have asked for a better group of people!

I am in the mood for reminiscing and so I thought I would recap some of the achievements that everyone in this team has worked so hard to achieve. I personally couldn't have done it without the help of everyone here (thanks especially to Joseph, Alex and Mike!).

We took what was an idea and grew it into reality. The dream to build Australia's fastest and most reliable Cloud Computing platform has been realised, with our platform recently being reviewed by Cloud Harmony as having the fastest networked back storage in the globe!

Looking forward, many people have asked us what is in-store for the year ahead. If the year that has just passed is anything to go by, then we are all in for a great year! We are consistently working towards delivering you (our clients and potential clients) a better product and more features. From day one we have focused on delivering an easy to use product that performs well and is reliable, and we will continue to do so moving forward.

Uptake of the OrionVM Cloud Platform has been great over the past few months, so much so that we have doubled the size of our production cluster several times. We now have several hundred happy clients to show for, and this uptake isn't slowing down!

Without giving away too much, expect to see a few key developments and updates within OrionVM over the next few months as we grow both the company, and the product. I look forward to announcing these milestones as they happen.

I truly appreciate all the help and support provided by both the OrionVM team and our customers, and here is to a great year ahead!

OrionVM - Finalist CeBIT 2011 Service Excellence Award

Wed, 01 Jun 2011 12:40:55 +1000

The team here at OrionVM is proud to be one of the three finalists for the CeBIT 2011 Service Excellence Award for providing a quality IaaS platform.

As many of you would know, we are exhibiting at CeBIT at stand H30, so if you are interested in the Cloud, please come past and chat to our team! Exhibiting at CeBIT has been great so far. We have met a huge number of great people, so hopefully things carry on the same way for the next day and a half.

Over the past few weeks, there have been some interesting developments regarding our platform, including a new upgrade to V1.1 of the OrionVM Cloud Platform to deliver a handful of new features (including private networking), as well as one of our clients (getflight.com.au) getting published on news.com.au and receiving hundreds of thousands of page views each day.

GetFlight is a great example of the power and usibility of the Cloud, in which we were able to scale up their server to deal with the load. I will release a blog post about this shortly in conjunction with GetFlight to demonstrate the benefit of using the Cloud for your business.

Lastly, for those who can not make it to CeBIT this week, here is a photo of some of the OrionVM team infront of our booth (complete with OrionVM shirts!).

OrionVM Exhibiting at CeBIT 2011

Sun, 15 May 2011 06:44:51 +1000

The OrionVM Team will be exhibiting at CeBIT Sydney 2011. It will be a great opportunity for us to meet all our customers in person, both new and old.

The event is running between the 31st of May and the 2nd of June 2011, in Darling Harbour Sydney.

OrionVM will be located at Stand Number H30, so please drop by and have a chat with the team. If you would like us to send you an invitation with a promotional code to gain free entry, please contact sales@orionvm.com.au and ask for an invitation to be sent to you.

We are looking forward to seeing you there.

Cheers,

The OrionVM Team

From Login to Shell in < 1.5 Minutes

Wed, 11 May 2011 10:47:56 +1000

At OrionVM, we focus on delivering High Performance, Reliable and Easy to Use cloud infrastructure. We often have customers asking us what is involved in creating an instance on the OrionVM Cloud platform.. Is it difficult? Does it take very long?

The easiest way to demonstrate this is to show you just how simple and quick it is to create an instance on our platform.

In this video we demonstrate how to create an Instance on the OrionVM Cloud Platform using the simple 'Instance Wizard'. We have accepted all the default options in this video as a quick demonstration. More detailed videos of all the OrionVM features will be released shortly.

&amp;amp;amp;amp;lt;/p&amp;amp;amp;amp;gt; &amp;amp;amp;amp;lt;p&amp;amp;amp;amp;gt;&amp;amp;amp;amp;lt;br /&amp;amp;amp;amp;gt;&amp;amp;amp;amp;lt;/p&amp;amp;amp;amp;gt; &amp;amp;amp;amp;lt;p&amp;amp;amp;amp;gt;&amp;amp;amp;amp;lt;br /&amp;amp;amp;amp;gt;&amp;amp;amp;amp;lt;/p&amp;amp;amp;amp;gt;

Cloud security, more so or less so?

Sun, 08 May 2011 10:58:02 +1000

For a great time the defining debate of cloud computing's viability has always been centered around security, whether or not applications are of more or less risk "in the cloud".

So, to further analyze this - What is different about running an application on cloud hosting versus say.. in house infrastructure, co-located infrastructure or traditional virtual machine hosting? The difference is where your control ends and that of your service provider begins or more importantly for some - where the liability lies.

In a traditional in-house hosted solution you are responsible for the entirety of your application hosting environments well being. This includes such utilities as power, cooling and the physical facility. Whilst this gives you the absolute level of control it comes with trade-offs, such as required the staff and expertise to manage all of these requirements and or consultants to help design the environment. Physical servers are no silver-bullet in terms of security though, you may control whoever physically access your machines but most threats come in the form of remote vulnerabilities.

The answer to the above downfalls comes in the form of co-location, placing ones physical equipment in a datacenter owned and managed by a service provider. This alleviates you of having both the burden of physical security and the technical expertise required to build your own facility. You do however, give up absolute control of physical security for this luxury. In this case your service provider now becomes responsible for ensuring the safety of your equipment and restricting access to it from third parties.

Virtual machine hosting represents the next level in loss of control, whilst in a co-located environment you were still responsible for the administration of your own physical hardware including machines, switches and storage devices. Virtual hosting, except in the case of virtual dedicated servers a primarily multi-tenant. This introduces a number of additional risks involving hypervisor security, storage segregation etc. Attacks such as side-channel attacks etc have been successfully demonstrated against a number of hypervisors, storage products and networking setups. These environment factors are only of a concern in a multi-tenant situation however, hosting your own virtual infrastructure is not prone to these attack vectors.

Cloud computing is somewhat of a special case of the virtual hosting paradigm, cloud computing may or may-not employ virtualisation at the computational layer but it must at some level to allow for rapid provisioning of resources. The point at which cloud computing becomes different from a security standpoint is now the additional layer of infrastructure logic that drives it's automated operation. This software, generally called cloud orchestration software; is the only real difference architecturally and technically from standard virtual hosting.

It is worth noting however, loss of control is in some ways also transfer of liability and responsibility. You entrust your service provider to manage physical infrastructure better than you can, or else why would you outsource it? Typically most organisation lack the skills to build their own datacentres, whilst a lot have enough skills to run applications on dedicated servers or setup virtual environments this number is still minimal. The majority of medium scale websites and applications run on virtual private servers in multi-tenant scenarios or now more recently - cloud hosting.

Included in this relationship is the understanding that your cloud provider is significantly better equipped to deal with infrastructure security concerns, as this is their core business. Security experts are few and far between in the usual enterprise IT department but make themselves at home in the operations team of a cloud computing provider. However, your cloud service provider can't protect you within the realm of your control. Most infrastructure clouds allow you full control over guest operating systems, bringing the end to your protection. You are responsible for all security concerns above this layer as you normally would be in any other scenario.

In most cases I think it would be fair to say cloud computing is on par with standard virtual hosting in terms of technical security, except in the special case of cloud orchestration subversion or malfunction. Cloud infrastructure is housed in secure datacentres, just as you would co-locate any equipment and as such offers similar levels of physical security.

My verdict is that cloud computing is no less secure than what is already in use by the majority of enterprises. In some cases it can provide additional security, by bringing expertise to the table that was previously beyond some companies reach.

Cloud computing isn't really all that different under the hood, it's all about a change in paradigm, increased flexibility and agility - in both resources and billing.

On Abstractions and the Magical Network Backed Disk

Sun, 24 Apr 2011 09:26:27 +1000

In the recent Amazon outage a lot of doubt has been raised about the reliability of cloud solutions and in particular block based cloud storage.

The US East Amazon region suffered a critical failure emerging from it's Elastic Block Storage (EBS) subsystem, overwhelming all 3 Availability Zones in the region. Taking down Amazon RDS with it, for a period extending over 24 hours, it is the worst failure of it's kind to strike Amazon.

Through the ongoing aftermath, a number of cloud experts have shared their views on EBS. One of which I thought is worth noting is that of Mark Joyent. You can read the post on his blog about EBS here: http://joyeur.com/2011/04/24/magical-block-store-when-abstractions-fail-us/

Mark talks of how it's abstractions and lack of transparency both instills blind trust and wariness into developers, sysadmins and DBA's alike when dealing with these "magical block storage" systems. The point he is making is quite correct and the story he tells of Joyent Clouds interaction with major vendors of such solutions rings all too true with most cloud providers - including ourselves.

Though the Amazon outage is the one to receive all of the press due to the enormous client base and high profile websites hosted on it, this scenario is far from uncommon. SAN and other network backed storage systems are notorious for causing widespread failures in cloud environments. Even in our own backyard, local provider Ninefold suffered a very unfortunate failure of their SAN backed storage in their first month of operation. Only weeks prior to that, the Relia cloud was bought down by a SAN malfunction... I think I can reasonably say that more than 50% of cloud failures are due to storage arrays malfunctioning, be they homegrown DRBD backed storage as Mark mentions or enterprise class SANs from Netapp or EMC. To be fair I would like to say it's often not the cloud providers fault as they are usually employing "black boxes" supplied by their vendor to deliver their flavor of "magic block storage".

So is it a problem with an abstraction or how it is implemented?

This is a problem we at OrionVM struggled with for a while during the early months of our existence, we had set out to build the "Worlds fastest cloud" but not sacrifice any element of scalability or reliability. Early on in our company's life, we built solutions on open source software such as GlusterFS, ZFS, DRBD, iSCSI, etc etc. Even prior to this we discarded expensive SAN storage as not being a viable solution due to it's prohibitive costs and lack of scalability. To fully understand the requirements of cloud storage we dissected what the cloud is really used for and went about learning where we should focus our efforts. In the end it was all too obvious that object storage and other abstractions were only good for a subset of use cases and the reality was that "magic block storage" is what people wanted, regardless of how hard it was to bring to the table. Databases are the very core of the dynamic Internet and they are very reliant on the POSIX style file system which is best deployed on fast block storage.

But that highlights the prior question, is it a problem with this abstraction that we should try to fix or can we cure the implementation?

Ultimately I feel it is the latter, sure the problem is hard but it's by no means impossible. The storage system in use at OrionVM is far from what anyone has ever seen before in both architecture and performance. We encountered the same issues as Joyent, heavy IO loads would outpace replication when using standard networking equipment, so we searched for and found a solution, Infiniband. One could say that Infiniband is at the core of our secret sauce. It powers our extreme point to point networking performance that allows us to replicate and access data in real time anywhere within our system, providing that "holy grail" Mark mentioned. Other issues such as "hot" or heavily used volumes impacting performance and having to be moved to dedicated filers doesn't affect us as much as we are able to remap "disks" in real-time automatically.

Whilst we could argue a problem with the abstraction, it then presents the problem of what to replace it with.. A question we and the industry are not yet ready to answer. At least for low latency and high throughput solutions at this point in time.

The power of OrionVM's architecture is derived from learning the lessons of those that have come before us, standing on the shoulders of giants. While Amazon is in some ways a competitor we praise them for what they have achieved. Like ourselves, Amazon designs and builds their own stack - an admirable feat for a company that was the forerunner of our industry.

EBS is a dated technology but was amazing for it's time. I think it would be fair to say it is the biggest implementation of network backed storage in the world. Taking away the criticisms of EBS and other SAN backed systems, we have engineered what we believe is the solution to EBS's problems, both from a reliability and performance perspective.

But how is it that we can be fully transparent without losing our competitive advantage?

Well we can't, but we do seek to have complete transparency of failures within our system. We explain as much of our architecture to our customers as we can afford to, so they can trust that their data is safe and that we are not at risk to the failure modes of EBS or any other current SAN solution for that matter. We do have failure modes, but we have engineered our system as such that they are all steady state and localised. To fully understand why we can not experience the same failures as EBS or a SAN is out of the scope of this blog post, but it is enough to say that we have put a lot of thought into it. For those interested you can find out more about OrionVM's infrastructure on our Technology page.

In the end, "magical block storage" is here to stay until a new abstraction which provides similar characteristics is available (low latency, high throughput, guaranteed consistency). My bet is it will come in the form of a next generation object storage system, something that the OrionVM team are hard at work exploring currently. :)

18 Days Since Launch.. An Update

Mon, 18 Apr 2011 09:26:24 +1000

So, it has been 18 days since the OrionVM Cloud Platform has gone live, and things have been progressing quite nicely. After 12 months of solid work to develop this platform, it is great to finally see people using the platform and loving it.

We have been in contact with the majority of our customers over the past few weeks, and the feedback has been nothing but assuring.

From the feedback we have received, our users have selected our platform to drive their applications because of the terrible performance issues they have experienced on many of the other cloud platforms.

The team at OrionVM is working hard to bring the next release of features to the market, so stay tuned!

3rd Party Performance Benchmarks

Tue, 29 Mar 2011 09:26:22 +1100

At the end of our last article ("Cloud Performance - It's all about storage") about the storage performance of the OrionVM Cloud, we mentioned that we would be publishing performance statistics that encompass the all round performance of the instances.

Over the past week, our platform has been benchmarked by 3rd Party Cloud Benchmarking site CloudHarmony.com. Cloud Harmony is a great resource that allows you to compare the performance of many cloud vendors globally, based upon 60 or so performance metrics. We have requested the permission of Cloud Harmony to publish some of these benchmarks here, as it gives you an overall overview of how some of the Australian vendors fair against their US competition.

For this blog, we have decided to include several Australian vendors (Cloud Central, Ninefold, OrionVM) and also two highly popular US providers (Amazon EC2 and Rackspace). One small note to make is that all of these vendors, with the exception of Rackspace, use network backed storage for redundancy. Rackspace utilises local disks, and hence have no redundancy against physical node failures. This also makes it easier for them to achieve higher disk speeds, due to the lack of network overheads. Rackspace were included in this article due to their popularity and widespread use.

We selected the similar instances on each vendor's cloud to do these comparisons (14-16Gb RAM). A notable difference is the inclusion of the Amazon Cluster Compute instance, which is in effect a High Speed Dedicated Server (Single Instance, 16 Cores, 23.5Gb of RAM). This instance was included as it contains Amazon's high speed 10GigE backed EBS volumes, allowing us to compare our storage engine against the fastest storage Amazon has to offer. It should be noted that a comparison between this Amazon Cluster Compute instance and the other instances on CPU is not a fair comparison, as all the other instances are multi-tenanted, while the Amazon CC instance has a single instance on the hardware, with access to all the CPU, Memory and Storage resources on that physical box (effectively a dedicated box).

The table below details the specification of each of these instances:

Server Name	Provider	Data Centre	Instance Size	Notes
ec2-us-east.linux.cc1.4xlarge	Amazon EC2	VA, US	Cluster Compute - 23.5Gb RAM	Fastest Instance, Single Instance, Included for IO Comparison
ec2-us-east.linux.m1.xlarge	Amazon EC2	CA, US	Extra Large - 15Gb RAM	EC2 Standard Instance - EBS Roots
cc-huge (AU)	Cloud Central Servers	AU	Huge - 15.5Gb RAM	Australian Cloud - iSCSI SAN Storage
nf-xlarge (AU)	Ninefold Virtual Servers	AU	Compute Xlarge - 13.6Gb RAM	Australian Cloud - iSCSI SAN Storage
or-16gb (AU)	OrionVM	AU	16Gb RAM Instance	Australian Cloud - Infiniband Custom Distributed Storage
rs-16gb-il	Rackspace Cloud Servers	IL,US	15.8Gb RAM Instance	Local Disk Array

Disk IO Aggregated Performance Benchmarks

The Disk IO test as performed by Cloud Harmony is an aggregate benchmark containing weighted scores from several disk benchmarking utilities (such as IOZone, Bonnie++ and Geekbench). You can read all about how the benchmark is performed on their blog: Cloud IO Performance. To summarise the article, the aggregate score of all the benchmarks are used to create an IO Performance Score (IOP).

To achieve a baseline, a dedicated server with dual quad cores (E5506), and 4 * 15,000 RPM SAS disks in RAID 10 has a performance baseline of 100 IOP. In this test, the OrionVM 16G instance returned with a score of 160, which makes it 60% faster than this dedicated server with 4 * 15k SAS disks. This is very good results for a redundant, network backed storage array and shows the strengths of our architecture and interconnects (40gb/s Infiniband). It should also be noted that the OrionVM instance out performed the Amazon Cluster Compute instance which is backed to EBS over 10GigE, and has the fastest off-instance storage IO available on Amazon EC2.

A slight note that we would like to make is that this test was done using CentOS 5.5, which utilises a 2.6.18 kernel, which is quite an old kernel. The IO schedulers in this kernel are slower than those as included in the newer 2.6.3x kernels. This accounts for the difference between the results of this test and those performed using a Ubuntu Server 10.04 instance, with a 2.6.32 kernel, which showed even greater speeds in our testing.

Database Aggregated Performance Benchmark

This database test comprises of several database testing suites (such as mysql-bench and pgbench) which produces an aggregate result as can be seen below. As you can see, the majority of network backed storage suffers in this test due to the latency of the network interconnect. The RackSpace instance excels because it is backed to local storage, with lower latency than accessing storage over a network.

The latency on the infiniband interconnects used by OrionVM is far lower than Ethernet, allowing us to deliver a higher number of IOPS to the instance.

Memory IO Aggregated Performance Benchmark

The Amazon Cluster Compute instance excels in the Memory IO performance test, as it has far more memory (23.5Gb) than the rest of the instances. Also, as it is the only instance on that server it is essentially a dedicated server and as such has uncontended access to memory resources. As OrionVM is multi-tenanted environment, with multiple instances sharing the same box, this isn't a fair comparison but is included for the sake of completeness.

Web Application Aggregated Performance Benchmarks

This benchmark compares performance using web server related benchmarks (such as blogbench, phpbench and postmark) and is affected by CPU speed and disk access speeds. The high speed storage array of OrionVM allows us to excel in this benchmark. This benchmark demonstrates the relationship between storage IO, and the performance of your web application.

The LAMP performance metric measures performance of a Linux Apache MySQL and PHP stack. Some of the tests performed include mysql-bench and phpbench. This graph also shows the benefits that high disk IO speeds on an instance can deliver to your application performance.

The Java, Python, Ruby and PHP performance benchmark is a performance metric made up of 4 interpreted programming language benchmarks. As these tests are CPU heavy, it demonstrates the CPU performance of each provider. As the Amazon Cluster Compute instance has uncontended access to the Dual X5570 Quad Cores processors, they exceed the rest of the pack in terms of CPU performance. The Australian provider Ninefold utilises X5660 (2.8Ghz Hex Cores) processors in their servers and they come in second in this benchmark. OrionVM's next hardware round will be packing hex core CPUs, so you should see an increase in CPU performance once we upgrade.

It should be stated once again that this comparison is skewed towards the Amazon Cluster Compute instance, because it is essentially a dedicated box, with uncontended access to CPU resources.

CPU Aggregated Performance Benchmarks

The CPU benchmark is an aggregation of several CPU benchmarking utilities (such as Unixbench and Geekbench). As you can see, Amazon EC2, Ninefold and OrionVM lead the pack when it comes to CPU speeds. As mentioned above, the Amazon Cluster Compute instance has uncontended access to the Dual X5570 Quad Cores, and hence they exceed the rest of the pack in terms of CPU performance and Ninefold utilises X5660 Hex Core CPUs. OrionVM currently utilises quad core CPUs in our build, but comes up closely behind Ninefold due to overhead of using HVM, while OrionVM utilises paravirtualisation. We should see a performance increase once we upgrade our next build to dual hex core CPUs.

Conclusion

In closing, the aggregated benchmarks above demonstrate that the new OrionVM storage architecture delivers on the performance claims we have been promising. The importance of IO throughput on the performance of a web application is often overlooked. As the majority of web applications rely upon a database, the storage speed of a cloud instance often has great impacts upon the speed of an application. This becomes evident when you look at the aggregated web performance benchmarks.

While we do not perform as well on the CPU front, we have conquered the harder challenge of delivering faster disk IO to instances. Very few users will be CPU bound on our platform, and hence having faster disk IO is more important than higher performing CPUs. We will be able to easily increase our CPU performance on the next hardware upgrade by buying faster CPUs.

With 3 days and counting until our production release (1st April 2011), it should be an interesting month ahead! I hope this article has been of interest, and has shed some light on the importance of storage performance on your application.

Flexibility - What you need in a Cloud

Thu, 24 Mar 2011 09:13:18 +1100

One of the major drivers behind a movement to the use of Cloud Infrastructure is the ability to only pay for the resources that you require as well as far lower provisioning times and greater reliability. This is a great asset for developers, start-ups and enterprises as it allows them to scale resources with their requirements rather than over provisioning on day one.

When we were deciding on the billing structure for OrionVM, we had a look at what everyone else was doing and dived into the depths of many pricing models. What we found, was that quite a few providers were very lacking in information or flexibility. Often you were presented with a number of predefined plans, with an inability to add specific resources as required. The main thought that came to our mind was "Why should I have to pay for an extra 100GB of storage space, when all I really want is an additional Gigabyte of RAM?".

This made absolutely no sense to us, so we decided to get rid of fixed plans, and allow you to purchase the required amount of each resource. To facilitate this, we have allocated a price for each resource (RAM, Storage, Transit and IP Addresses) which allows you to very simply pick and choose amounts as required. In addition to this, we also do the following things to make your life easier:

Per Second Billing - we bill for your usage in per second intervals, so you truly only pay for what you use. Unlike many of our competitors, we do not round up to the nearest hour.
No Penalty Rates - we do not charge you more, simply because you use more resources than your allotted plan. This is in many ways a marketing/pricing gimmick to force you to choose a higher plan than required, to remove the risk of going over your resource allocation and paying for resources at a far higher rate. With OrionVM, you will simply pay for more resources at your normal rates.
Single Pane Instance Creation Wizard - the majority of users will not use the majority features offered by many at the time of instance creation, hence allow for both a simple and advanced instance creation wizard. People have given us great feedback regarding the ease of use and speed of this wizard. As you can see below, it is simply a matter of entering a name, choosing your distro and the amount of resources on a sliding scale and pressing create.

The main reasoning behind our actions is to allow for end users to access the degree of flexibility the Cloud promises them, but without the complex pricing structures and penalty rates that prevent them from considering this flexibility. All in all, we are aiming for simplicity in our pricing!