IBM Launches PureSystems – An Integrated Compute, Storage, Networking and Middleware Approach

IBM PureSystems Highlights

• Launches PureFlex and PureApplication – the first of a family of expert integrated systems accommodating compute, storage, networking and middleware
• A physical manifestation of its Cloud reference architecture
• Designed for ease-of-use and rapid speed-of-deployment
• Introduces Patterns for capturing expertise in building infrastructure, platforms and applications
• Launches the PureSystems Centre – quasi-appstore for Pattern cataloguing and distribution
• Has certified 150 applications from 126 ISVs and launch among 600 available
• Has an infrastructure-in-a-box approach, but is also highly pluggable through the inclusion of Storwize V7000
• ISVs, SPs And Enterprises lacking the budget, energy or resources for highly active infrastructure maintenance likely to be early adopters
• Demonstrates how far we’ve moved from the pick ‘n’ mix horizontal market of the past

We should have seen this one coming as it’s a natural extension of IBM’s server approach which we have covered extensively here – see its Xeon E5, its zBX mainframe, its relationship with VMware, its Cloud Computing reference architecture, its building block services strategy, its relationship with Red Hat, its acquisition of Blade Networks and its introduction of its Smarter Computing concepts. However IBM managed to keep its announcement of PureSystems on the 11^th of April such a closely guarded secret that it took us somewhat by surprise. Even the ‘Expert Integrated Systems’ title of its event near Tower Bridge in the City of London gave little away. You’ll want to learn about why this is such an important announcement.

What Has IBM Announced?

IBM claims to have spent around $2 billion over nearly 4 years involving 27 labs in 17 countries leading up to the introduction of PureSystems. Initially IBM will sell 2 systems. In particular:

• PureFlex is an ‘infrastructure system’ initially available in Express (c. $100k), Standard (available at the end of May) and Enterprise (c. $1m) configurations (end of June)
• PureApplication is a platform system inheriting the essential features of PureFlex, with some narrowing of the middleware choices; its small (96 cores) (c. $2m), medium (192 cores) and large (384 cores) (upwards of c. $7m) versions will be generally available from the end of July; these systems are designed to run numerous easy-to-deploy ‘patterns’ (See below)

On one level PureSystems – which takes Power7 and x86 blades alongside V7000 storage and networking modules – is a new chassis which can serve IBM for the next 10 years: on another it represents the most advanced approach to integrated systems yet, surpassing those of Oracle/Sun, VCE vBlock and HP. In terms of its heritage it looks up to mainframes and back in time to the AS/400 and System/36.
The inclusion of essential middleware not only eases deployment and management, but is also a physical representation of its reference architecture, which in itself rightly excludes products.
PureSystems should not in itself be considered a new server (although the package is encased in a cabinet encasing its various rack-mounted modules): we’ll be counting the server, storage systems, networking and middleware components separately in their respected classes as before.
From a hardware point of view the new system has impressive potential capacity. The maximum capacity of the cabinet includes:

• 896 cores
• 43TB memory
• 480TB storage
• 26 million Input Output Instructions per Second (IOPS)
• 4 chassis

In some ways this is smaller to HP’s container approach: however in others very different – since rather than building a highly customised solution for each user, IBM has been brave enough to pick a set of standard infrastructure components it believes make an adequate and good platform for most customer deployments. Both approaches rely on building infrastructure in the vendor’s factory to save customers time and money.

IBM Introduces Patterns of Expertise And The PureSystems Centre Quasi-Appstore

Perhaps the most impressive part of the announcement is its ‘pattern’ approach, whereby the expertise needed to choose components and software has been captured in software. For instance:

• Its Infrastructure Pattern for instance includes automated ‘smart data’ handling placing relevant information into Solid State Disks (the provision for which is included as standard)
• Application Patterns will help with faster, easier deployments
• IBM itself has been busy in capturing its own expertise in patterns, claiming to have 4.5k already, while
• 3^rd parties will also play a part.

At launch PureSystems has 150 certified applications from 126 ISVs available; in addition users themselves will be able to simplify deployment on the new system by creating their own patterns. IBM has also launched PureSystems Centre, which is a (kind of) Enterprise appstore from which Infrastructure, Platform and Application patterns can be located: only ‘kind of’ because it doesn’t intend to sell applications there, preferring a ‘bring your own license’ approach.
The introduction of patterns reminds us of the deep knowledge IBM has in vertical markets. We think its application patterns can be extended to include industry sector specific implementations over time.

The Pluggability of PureSystems – Infrastructure In A Box

The new platforms do not just integrate compute, storage, networking and middleware as an implementation of its reference architecture, they also allow various other components to be plugged in. The inclusion of the Storwize V7000 means that its SVC storage hypervisor allows the attachment of other vendors’ arrays, although IBM charges higher licensing pricing for customer who don’t just want to plug them in for one way migration to IBM storage.
On the networking side IBM has worked closely with Emulex and Melanox, as well as including its own IBM Networking (BCN) top-of-rack switches. Its open approach here will lead to easier attachment of networking from Brocade and Juniper (which are OEM partners) and even Cisco (with which it has a distribution agreement). It supports Ethernet FCoE, Infiniband and other protocols. Cisco and HP through its acquisition of 3Com are probably less open in suggesting other vendors’ networking in their integrated approaches.
We think these systems will become the ‘infrastructure in a box’ for some customers, who may be able to use them as the heart of a more standard approach, plugging in relevant components as appropriate.

The Pros and Cons Of Ease-Of-Use And Speed-Of-Deployment

IBM’s recent research supports the long-held belief that IT shops spend the majority of their time maintaining data centre infrastructure (‘keeping the lights on’) rather than helping their businesses to be ‘innovative’: although the 75:25 split is wider than the 80:20 findings we remember from10 years ago.
The central issue here is the shift from a horizontal layered approach to the vertically integrated one exemplified by PureSystems. The horizontal market was a time of pick ‘n’ mix for IT managers, who could make their organisations significantly more productive by choosing specific competitive offerings in each of its various layers. In a sense computer systems had not been thoroughly invented 10 years ago. However things have become far harder for them with the influx of virtualisation (at least in the x86 server space) and choices have been limited.
Many users are exhausted by the mass of procurement relationships and applying scarce expertise in ‘what works with what’ issues. IBM’s bet is that the savings in terms of ease-of-use, speed-of-deployment and (most importantly) cost has led to a market in which more standardised solutions can prosper. No matter how well it has designed these systems and how they adhere to open standards, there will be an inevitable restriction of choice as a result of integration.
Early customer testing of PureSystems allows IBM to promote savings in systems, Op Ex, ‘management’, time to deployment and total IT cost categories. Particularly impressive is its demonstration of ‘4 clicks to Cloud’ enabled by the inclusion of SmartCloud Entry in the PureFlex system. The advantages of IBM’s approach was well summarised by one its partners at the event, who summarised it as a change from a cottage industry to the industrial age.

Some Conclusions – The Likely Early Purchasers Will Be ISVs, SPs And Enterprises

With the lowest entry price at $100k very few small companies will have the budget to buy these machines, although IBM will no doubt extend its integration approach downward in future. At the London launch were heard endorsements from Temenos, Dassault, Opennet, SunGuard and SAP who are among the 600 partners IBM has work with on the project – a good illustration of the types of company likely to buy. We think the systems will sell well to mature organisations looking to save on infrastructure maintenance, especially if they can redeploy the salary savings on doing things important to their business. We think ISVs, Service Providers, Managed Service Providers and Outsourcers will be among the early adopters, along side IBM Enterprise customers looking for an easier way to modernise their architecture without having to make a large number of intricate choices.
IBM’s mainframe users are unlikely to adopt these systems unless they are developing something big, different and new, since they already have the ability to couple Power and x86 processors tightly to their machines using its Universal Resource Manager (URM) and attached zBX rack and are encouraged to use cheaper mainframe Mips to run non-transactional workloads.
Those customers who still want to make detailed choices and have the experience to maintain their systems efficiently are unlikely to leap at the chance of becoming fully integrated in one jump. Customers in countries where technical expertise is hard to acquire are more likely to think about purchasing PureSystems. As is often the case IBM’s success with these products will be in its ability to sell these machines to medium or large non-customers; of all of these we expect ISVs in the US, Telcos acting as Service Providers/Aggregators in Europe and Enterprises lacking technical resources (perhaps mainly in Asia and growth countries) to be the good bets.