Network Attached Processing - the next big thing for Java ?

Will NAP have the same all-pervading presence that NAS gained? It's not often that a new piece of technology comes along where people are tempted to say "why didn't they think of that before?" Yet I think I have found just such a technology. And if I am right, you will be hearing a lot more about it in 2006. The concept is called Network Attached Processing, and the company in question is Azul Sytems. When Java was first invented, it was decided that software code would not be compiled "natively" to any particular hardware or operating system. Instead, the compiled code (sometimes called "byte code") is designed to run in an environment called a "Java Virtual Machine" (or JVM). The JVM represents an imaginary machine architecture. In order to run Java applications on, say, Linux or Windows, it is necessary have a Java Runtime environment that maps the JVM to the specific target operating system architecture. Once the Java Runtime environment has been written, then all Java programs would be able to run on that target. This architectural approach lies at the heart of the "write once, run anywhere" mantra for Java. Obviously, if the underlying OS and architecture is relatively similar to the JVM, then the Java Runtime will be relatively easy to write, and should be very efficient. Ideally, then, an executing JVM should be located on an Operating System and Architecture that is architecturally similar to the JVM. On the other hand, there is a requirement to be able to run Java applications on Industry-standard Operating systems (e.g. to run Websphere or Weblogic or JBOSS on Solaris Linux). Azul Systems have provided a solution to this challenge. - Network Attached Processing, in the form of an Azul Compute Appliance. The Azul solution consists of a customised JVM or "VM Proxy". This VM Proxy receives incoming byte code for execution and forwards it directly along a network path to the Azul Compute Appliance, with executes it. On the Compute Appliance, the byte code is received by a "VM Engine", which then performs the Java compute operation, before returning control to the calling server's VM Proxy. One obvious advantage of this architecture is that the "client" Operating System still appears to be performing the same functionality as before, running the business Application. However, the actual CPU processing is being off-loaded to the Azul machine. Neat, or what? So what's the Azul Compute Appliance? In a nutshell, it is a 24-way (or up to 384-way!) multi-processing server which is specifically designed to run JVM environments very very efficiently. For example, it handles multi-threading very well (not surprising, if you could have 384 CPUs!), oodles of "heap" memory, highly efficient memory garbage collection etc. etc. So how about the best of both worlds ? Keep your existing application on your old server, hook in a Gigabit Ethernet card, and hang the Azul System off the other end. Better still, have multiple Servers being Compute Served by a single Azul machine. Sounds a bit like a Compute equivalent of NAS ? Yep - you've got it! Once the concept is grasped, all sorts of opportunities arise. Firstly, we are used to purchasing servers to host J2EE environments, based on their computing power. Instead, the host server becomes just a "mount point", a suitable O/S architecture for running the I/O and communications activity. The real processing is done in the "compute farm". What happens if you need additional CPU for this growing application ? No action required - just ensure that the compute farm is powerful enough ! The use of a "Compute Farm" suddenly changes the whole dynamics of Servers in the datacentre. Each Java server could just be a tiny blade (or a Virtualised server), providing it has the O/S and I/O capability for the application. Datacentre management of Servers would be massively simplified with NAP, just as it has been in the storage arena with NAS. Azul Systems web site is at http://www.azulsystems.com. I hear that they have plans to support .NET runtime as well in the near future. Definitely one to watch in 2006.

News Review: Ingres Open Source Buyout from CA

Yet another twist in the Ingres saga, as CA float it off independently. But have CA forgotten what they use Ingres for? It seems only yesterday when I was commenting on the decision by CA to Open Source the Ingres relational database (see Ingres Open Source = Graveyard or Second Life ), in July 2004. At the time, I considered that there was significant business logic in the decision. A genuine heavyweight Open Source DBMS would be able to compete with the likes of MySQL, with the ability to handle massive volumes of data. On the business side, revenue would come from commercial support, even though the software itself could be free. Now the world has changed again with the Management buyout of Ingres Corp from CA. The new Ingres Corp I suspect that the writing was on the wall for Ingres as a result of the reorganisation last April, when Ingres became part of the "others" division. Now, however, CA have sold the product and many of the staff to a Silicon Valley private equity firm, which itself has only been going for a year. However, one of the most impressive things about this transaction is the management team, which reads like a "who's who" of database heavyweights: * Dave Dargo (CTO): 15 years Oracle, including responsibility for the Oracle on (OpenSource) Linux programme. * Emma McGratten (Engineering): 11 years running the CA Ingres engineering team. * Andy Allbritten (Support Services): ex-Oracle VP for Support and Services. * Dev Mukherjee (Marketing): ex-Microsoft Servers General manager for Marketing. Ingres Corp is now a company of 100 employees (most of whom are ex-CA, who have moved across to the new organisation), with plans to grow it significantly. In addition the OpenRoad development tools (which are tightly coupled with Ingres itself) and other products like the Enterprise gateway are also included. Even the Follow the Sun support teams are moving. However, OpenRoad will not be Open Sourced yet (try saying that in a hurry!) We are promised that aggressive marketing will start soon. Coincidentally (?) the announcement was made on the same day that Microsoft released the latest version of SQL Server. That is called timing ! Reactions Looking at the Ingres Newsgroups, there is a mixed reaction. One DBA complained that his company was just in the middle of negotiating a move from the earlier closed-source versions of Ingres to the new OpenSource Ingres R3, just as the announcement was made. Others responded positively, since many had complained in the past that CA had not really marketed, or put RD into Ingres. With the new company there will have to be a strong focus. And existing customers should still be supported, often by the same people as before. CA's decision Whilst this appears, in my opinion, to be very good news for Ingres, I am slightly puzzled by why CA have agreed to it: Don't they realise that a huge percentage of their software is deeply reliant on the Ingres database ? Unicenter R11 is due to be launched, which uses Ingres as it's only relational database. Won't the marketing people at CA feel a bit exposed at seeing their products dependent upon a non-CA database? Challenges In the past, one of significant aspects of the Ingres sales proposition was that it was being actively used and promoted within CA itself. Therefore, anyone who bought Ingres would be purchasing a product with a significant locked-in installed base, which would guarantee longevity. Now, things have changed. The "used by CA" tag no longer has the same credibility. Ingres Corp is out on it's own in the wide world, and will have to fight on it's own against the likes of MySQL. The competition will be fierce. Oracle have just recently announced a free (yes - free!) version of Oracle Express for Linux. SQL Server has just seen a major enhancement. And MySQL is beginning to move into the big league, with an improved optimizer, views and triggers. One big question for all "Ingressers" (if indeed that is the correct collective name) is whether the Ingres development programmers and support team will remain with the new company in the medium term. One thing is clear to me, Ingres Corp is unlikely to have the same marketing muscle as Computer Associates. On the other hand, it should have a more focused approach to selling the product. I wish them well.

News Review: Peregrine finds a home inside HP

The Prodigal Returns News that HP have agreed to purchase Peregrine must put some smiles on the faces of the existing 3,500 ServiceCenter customers. At last they can feel that the software house has a valid home, where it will hopefully get the investment and marketing effort it deserves. Peregrine has had an interesting story getting here. It was quite an acquisition-maker itself in the early part of the decade, including an interesting time "dating" Remedy (now part of BMC). Then it ended up filing under Chapter 11, and seemed to write itself out of the history books. The thing that seems to have saved it is that it has a reasonable product, at a time when every company is trying to get into the ITIL framework, by producing software offerings with the ITSM (or Service Management) strapline. HP's offering in the form of OpenView very much complements this, so I foresee a strong future for both products. The key factor for ITSM offerings is having a common configuration management database. This database should be able to tie together all the assets of the company (Servers, workstations, software licenses and installed applications), and cross-match them to the HelpDesk (so that incidents can be logged against them). This in turn means that Problem Management can drill down into root causes by looking at the Incident history. Then Changes and Releases can be implemented against these assets. So a common CMDB is vital - both for a good ITSM offering, and for a successful deployment of ITIL processes. Peregrine has potential to be such a product, particularly if it is well integrated with OpenView in future releases. One curious question: Why didn't IBM purchase Peregrine ? After all, IBM acts as the channel for a lot of the Peregrine products. And surely IBM would benefit from Peregrine's CMDB. These days, I don't hear much about IBM Tivoli. It used to be the market leader in management of mid-range systems and applications. Now we have BMC, Computer Associates and HP. Are IBM unconcerned about the ITSM market ? Or maybe they are just biding their time.

News Review: Oracle + Peoplesoft + Siebel

If you can't make it, buy it. The recent announcement that Oracle will be buying Siebel had been fairly widely predicted in some areas of the press. Oracle are paying $5.8 billion for 4,000 customers. I guess this is relatively small money, compared with the $10 billion they paid for PeopleSoft. However, this leaves Oracle with a massive workload to integrate and get value from all their many products and offerings. The Support Challenge If you look through Oracle's acquisitions during the last few years, there is a huge amount of CRM software which they have in their portfolio: * Peoplesoft * Vantive (part of PeopleSoft) * JD Edwards * plus Oracle's own offering * and now Siebel How on earth could Oracle support that many different code lines for just one functional requirement? Some clues about how Oracle might wish to do this are in a recent article in Oracle Scene, the UK Oracle User Group Journal, about Project Fusion. Fusion is the name for Oracle's new Service-Oriented-Architecture (SOA), which is a way of building software applications which promote connectivity between applications. Oracle will need it!. The article explained how this approach would bring together the best of Oracle, PeopleSoft and JD Edwards. Perhaps SOA-enabling all these tools will work. But there is still a lot of (redundant?) code to support. Motivation So why did Oracle buy Siebel? Basically, there are a number of reasons for buying a rival manufacturer in the industry: * Gain technology, to update or improve your own offering * Take out a competitor, to enable you to charge monopoly prices * Defensive action, to consolidate against another, larger rival. Somehow, I don't think Oracle have bought Siebel in order to gain some valuable piece of technology. Instead, I think the second and third reasons are more likely In the case of third reason, the big rival is, of course, SAP. However, Oracle may have a large percentage of the market share of CRM / ERP software, but their offering is fragmented, unfocused, and expensive to support. There is a battle between Oracle and SAP. And any general will tell you that the organization which is able to mass all it's forces against a single point of attack will win the battle. Oracle is in desperate need of a good coordinating strategy.