Previous month:
December 2011
Next month:
July 2012

January 2012

Cloud Transformatics - Rethink Your Storage Strategy (SAN)

One of the great things about the Cloud is the ability to stimulate new/different ways of thinking about your datacenter and IT operations in general. This is particularly valid in the areas of service delivery and quality management for an ever more demanding consumer population!

This is what I call Cloud Transformatics - the vectors and drivers of Cloud induced change in traditional IT operations and service delivery. I will be writing a mini-series of blogs on this theme, starting with SAN storage.

One of the tenants that the Cloud has really hit on the head, is the notion of massive old-fashioned monolithic arrays. The simple idea that using smaller lower-cost building blocks can actually reduce overall cost while giving more reliability, performance and ultimately service quality swept through the storage industry through the last decade.

Most storage manufacturers moved to disguising their systems as modular and scalable storage systems and only under innovation pressure from competitors started putting in features that cut margins and overall sales volume - think deduplication, virtual LUNs, compression, unified storage, storage tiering, thin storage etc.

Indeed even the very storage protocols became a religious debate. SAN FC vs iSCSI, SAN vs NAS, CIFS vs NFS, converged cabling etc. Many examples of this spring to mind, the NetApps and 3PARs of this world. Many of these were simply acquired by those companies that were their competitors. 3PAR went to HP, Equallogic to Dell, Isilon/Data Domain to EMC and Pillar Axiom to Oracle.

Curiously, in the IT market, good-innovative products don't always survive. Isilon was struggling in the market against massive competition - and was acquired and survived and thrived. That is great news - without the acquisition the industry may have lost the important innovations they bring to market. The same goes for 3PAR and Data Domain. 

I have been watching Pillar Axiom for some time. It struck me when verifying its architecture, market penetration, sentiment analysis and "cool" factor that this was a very promising technical architecture directly tackling issues typical in customer datacenters and storage services delivery in particular!

Firstly, there is the idea of SLAMMERS - the protocol and CPU heart of the Axiom Storage system. These are modular, built on standard industry CPUs (AMD) and are easily ramped up for roaring performance. A simple CPU upgrade from dual to quad core delivered vastly more vroom and correspondingly better service management. Imagine what an AMD Bulldozer (16 cores) or a corresponding Intel chip would deliver!

Secondly, there is the idea of BRICKS - which are dual RAID controller disk enclosures providing up to 24TB of raw storage using 12 disks, with a separate hot spare - all in a 2U package. This is what I call modular. These enclosures have their own cache and offload disk controller operations away from the slammers. Local hot spares all ensure rapid rebuild operations localised in the brick. Oracle claim 128 RAID controllers operating in parallel in 64 separate bricks per axiom storage system.

This is a great level of parallelism and modularity. Again, industry standard components are used, and scalability through component upgrades (CPU, RAM, backplane, interconnects, disks) can yield surprisingly large gains in performance and capacity!

Last, but not least, there is the PILOT - the hardware appliance providing out-of-band management features. This provides in a simple intuitive interface the management cockpit of the axiom storage system. It has features you would expect such as managing disk resources and pools. It also provides features that are definitely in the enterprise class league:

  1. Application aware storage profiles
  2. Policy based resource control
  3. Spindle striping levels
  4. IO Priority Queues
  5. Disk stroking (using inner or outer spindles for performance)
  6. Network bandwidth control
  7. Multi-tenancy with Quality of Service profiles
  8. Thin storage, storage tiering, distributed RAID for linear scaling

Pillar was typically put in the mid-range league, probably due to not supporting mainframes. However, the feature set is clearly enterprise worthy. Recently Oracle has started to add support for Hybrid Columnar Compression which further adds value to this device to provide SAN enabled storage support for Oracle 11g environments over and above what is available in any other high-end storage array currently.

Looking under the covers, and extrapolating somewhat, we have a SAN storage array with enterprise features. It has a modular slammer structure that corresponds to the EMC VMAX with its engines model. The Axiom also offers 8 engines, but provides, through its bricks, a linear scalability. If Oracle inserts Infiniband interconnects (40Gbps links) inside the axiom, then this will be a screamer of a system!

I have often seen clients who exhaust the high-end arrays in I/O well before the claimed 2000+ disks are reached. So top end capacity is nice on paper, but in reality it is rarely reached in intense environments.

The Axiom with its odd brick and slammer architecture addresses one critical concern that I have noticed with the traditional gamut of high-end arrays - namely the posturing that this is the last word in storage and why one would possibly need anything else! Well, Cloud Transformatics clearly shows that true modular systems are winning out in the real world and provide dramatically better performance and value.

If one needs more performance than available in a single Axiom system, then put a second Axiom system in place! That is how Cloud storage works - ramp up IO and Capacity through adding another module. That we can do this in a SAN is definitely worth taking a look at!  

The Race to Business Value

The Axiom already had many of the features associated with high-end SAN storage, but available in a pay-as-you grow model through adding additional modular slammers and bricks! This is remarkable value! Further, the Axiom also provides, through its architecture, NAS services. 

With the current shortfall of global disk production due to the tragic flooding in Thailand, every storage manufacturer is projecting steep price increases or delayed delivery. This is the perfect time to re-evaluate the SAN storage strategy. Oracle is providing substantial intellectual property integration into the platform to further deliver real Cloud derived value (CVD).

This platform in its architecture shares similarities with scale-out NAS systems such as Isilon, providing infrastructure for big data. It does not take a big leap in imagination to see such functionality coming to this important datacenter building block.

Why is this important for the CIO?

One would not typically think of Oracle as an enterprise SAN storage provider - but with the Pillar Axiom acquisition, they are squarely set to disrupt this space. Customers are using these systems to great effect. Axiom represents an inflexion point in how one things of SAN storage. That it can be cost effectively implemented, high end features provided and direct links to enterprise software stacks of database/ERP systems (SAP/Oracle) make this a potent platform at dramatically lower prices!

CIOs setting their IT agendas in 2012 should take into account that there is a new disruptive player in the datacenter offering storage services. Oracle can not really be disregarded in this respect. Time will tell how clients react to the exciting potential of the Pillar Axiom platform now that it has some enterprise backing through Oracle. 

CIOs looking to secure deep cost savings and efficiency should note the potential of this SAN storage platform. Clearly, if there is a use case in their organisations for enterprise software stacks (databases, messaging systems, ERP, CRM etc) then Oracle is a major player. Oracle is already probably used in most enterprise shops, and further deep practical value (not just paper savings) can be accrued from using the Axiom SAN storage platform.      



The opinions expressed here are my personal opinions. Content published here is not read or approved in advance by Oracle and does not necessarily reflect the views and opinions of Oracle.

The Practical Cloud - Road to Cloud Value Derivation (CVD)

After a change of residence and a change of job - it is high time to write another blog!

In talks with many customers and indeed feedback on this blog site, I receive a lot of indication that the "current Cloud" the pre-2011 industry was marketing was simply tackling the infrastructure side of things. Much of the focus was on consolidating the x86 server estate and delivering features such as migration of live virtual machines across physical servers.

While this is fine as the initial steps to deriving some form of value - it is typically too little. Many business leaders and IT managers indicate that "we have bought into the Cloud, we have virtualized and we can even offer a level of VM automation in terms of provisioning" - "so where is this taking us and where can I highlight the ongoing value to the business?".

This is a very valid line of questioning. The client has millions of $$bucks$$ of equipment sitting on the floor, they have attended training and have done everything they were told to do. The result - they can create a virtual machine with software in minutes as opposed to hours or days. Cool!

That is not a lot to show the business for all that investment and evangelism. This is typically (and incorrectly) lauded as a solution and a great win for all!

IaaS alone, in my opinion, was always too little value.  The approach of simply putting servers, storage, network with a thin veneer of hypervisor magic has limited value in itself. This was incidentally the main haunt till 2011 for mainstream hypervisor purveyors.

This type of datacenter transformation using pre-assembled hardware for the sole purpose of consolidating x86 is too simple and let's face it - too dumb. Clients are cleverer than that. Clients have persisted in following the virtualization wave, and that is good. They have somewhat resisted the Cloud marketing till now as it was simply focused on replacement of their existing hardware and hypervisor stack.

Towards the tail end of 2011 we started seeing a stronger focus on provisioning enterprise software and environments - DB as a Service (DBaaS) which was nothing more than installing a database instance on a virtual machine through a browser provisioning page. Well that is better - but still does not smack of value! Indeed, if you want many big instances of databases with say 64 virtual CPUs per VM you were out of luck! AND yes there are customers that do this!

In 2011, we started to see the emergence of the appliance. This was an entire hardware and software stack that was factory installed. In some cases, such as the EMC GreenPlum appliance, this was built using the components with functional tuning to undertake the task. Others such as Oracle with Exadata Database Machine (which has been around since 2008 incidentally - but first used Sun intellectual property acquired in 2010) not only took the idea of virtualization but actually embedded it into all the components in the stack.

Through innovation, integration, best-of-breed technology and the simple idea that a system should do what it is designed for to the best of its ability, Exadata represents, in my opinion, a new approach to transformation that makes real business impact.

I am sure that during 2012 we will see a move away from the generalized Cloud stacks, such as VCE VBlock, Dell prepackaged servers with VMware installed and something similar from HP Virtualsystem for VMware. These systems are all focused at helping the hypervisor - in this case VMware vSphere, perform its job well. However, the hypervisor only lets you manage virtual machines! It does not do anything else!

That is also the reason that I see the move away from expensive hypervisor software solutions towards open source solutions or systems having the hypervisor embedded as a functional technology to support an enterprise software stack - with no $$ for the hypervisor per se.  

The Race to Business Value

One of the issues that has been stagnating business value derivation through Cloud technologies has been the lack of business as a driving stakeholder. Business should be  driving the IT roadmap for an organisation. Business defines what it wants from developers in the form of functionality. Why not the same for IT infrastructure?

You see the value of Business is that it thinks differently. Business tends to think at levels of enterprise architecture holistically as a driver and motor for business value generation! They think frameworks and they think (with developers and architects) in terms of enabling software platforms upon which to further their unique selling points.

The real Cloud value to be derived in that case is based on the software Cloud platforms leveraged to facilitate global service/application delivery with quality of service baked in. These platforms in turn are used to create further value!  

The real business case for the Cloud comes in the form of Platform-as-a-Service (PaaS). I think that Exadata hits this nail on the head. I don't just want to be able to setup a virtual machine running the database inside, I want the functionality of the database itself! Exadata delivers just that through a clever blend of components!

Why is this important for the CIO?

CIOs have set the agenda for Cloud in 2010-2011. They have seen that it has an effect on the delivery of IT services - but not necessarily a direct impact in the culture of the business or indeed the value the business derives. The early gains have been achieved, and it is time to move on to business focused IT.

CIOs look beyond the mainstream hype. They verify through intensive research and peer-level networking the effect of IT strategies on business value. The CIO pioneers and sets the agenda for deep intelligent consolidation. Not just doing more with less - BUT gaining greater business insight and leverage with fewer more effective resources!

Exadata, and engineered systems of that ilk, with embedded technology are paving the way for scale-up/out with extremely high performance and gathering in the benefits/innovations of the IT industry over the last years e.g. unified networking with Infiniband, high performance SSD storage, deduplication, compression, tiered value-oriented storage, big data capable file systems and indeed open source.  

That is a very potent mix, and Oracle customers are actively leveraging this. They have been using Linux and Oracle Solaris 11 to support those enterprise workloads needing that level of reliability and speed. They have been consolidating hundreds of database and middleware servers - yes - hardware, mixed OSs, non-x86 systems, licenses, management tools, script frameworks and so forth. This is real consolidation!

Further, they have used the well respected Oracle 11g enterprise capable platform to power their Java applications, drive the backend of their middleware platforms, created new value by delivering through the Exadata platform applications to the mobile space (iPads, Androids, Browsers, OS independent applications). 

Indeed, if the Java virtual machine (JVM) is one of the ultimate forms of virtualization, it makes perfect sense that as a business which has elected to use that technology you create the underlying infrastructure AND platform ecosystem to support those efforts at scale.

The Corporate Cloud Strategy can be dramatically refreshed and aligned with the ability to deal with all data needs in a single well managed platform. Exadata provides in this case the ability to deal with all database needs that an organisation has from the smallest to the largest. It provides significant front-end direct value. 

Other Exasystems have started to arrive to deal with specific challenges such as big data and middleware. These use the same magic source of Exadata Database Machine, but are tuned/enhanced for their specific functions. Deep lasting transformation can be achieved and the very nature of these Exasystems means the Business must be included as a principal stakeholder - they can truly see what the value of extracting a business insight means in hard $$ terms!

Look out for these paradigms that directly affect business value and indeed allow new business insight to be gained by easily manipulating petabytes of information in near-realtime! They provide the ability for the business to rapidly come to market with new products, support directly application developers, are built on industry-proven technologies - and best of all - retain the key know-how of your developers and DBAs - they will be up and running with little change to their operational routine!    



The opinions expressed here are my personal opinions. Content published here is not read or approved in advance by Oracle and does not necessarily reflect the views and opinions of Oracle.