Thursday, April 15, 2010

Differences? Not so big, anymore!

While I am sitting at my desk putting this post together, my thoughts are many miles away. This morning I received a reminder that EBUG is starting in just 2 weeks – by the time you read this, it may be as soon as next week – and it will only be the second time in more than a decade that I have missed this annual event. The picture at the top of this page is from last year’s event, the closing evening’s fun-filled outing in Prague – what a different atmosphere from that of the conference and exhibition! And good to see Keith enjoying himself!

For more years than I care to recount I have participated in user events. My first memories of any user event go back to March 1976, when I drove from Edmonton, Alberta, all the way down to Dallas, Texas, to participate in one of the earliest gatherings of the user event sponsored by Insyte Datacom Corporation (IDC), the company behind the Datacom/DB product. In 1979 I made the even longer trip from Sydney to Richmond, Virginia to attend the user event sponsored by The Computer Software Company (TCSC), the company behind the third-party IBM operating system, EDOS.

The user conferences in support of BASE24, especially those held in Europe, were always times when you had the chance to connect with “real” users; more so than at other platform events, with their focus on systems managers and developers, participants at the BASE24 conferences were usually drawn from the business units directly responsible for supporting revenue-generating operations. While I was working for Insession EBUG gave me great insight into the needs of these business users, and I certainly came to appreciate their needs and concerns.

In many respects, the same appreciation of the needs of the business user that we had at Insession lives on today, within Infrasoft. A start-up focused on infrastructure for the NonStop platform it seems appropriate to acknowledge the arrival of Infrasoft’s first product, uLinga as we talk of EBUG. uLinga provides the services and protocols needed to unplug from the need for SNA, and allows for a pretty easy switch to TCP/IP. Requiring only configuration changes on either platforms, it will give users a less expensive solution for connecting applications on NonStop with those on the IBM mainframe.

Although this post isn’t just about EBUG or Infrasoft, they do provide a good way to return to the topic of new development on NonStop. Recent exchanges posted to the LinkedIn group, Real Time View, on the discussion “NonStop; HP's ‘halo product’ ...” continues to highlight how much emotion persists across the community over how much better NonStop could be doing in the marketplace. Visit this discussion on LinkedIn, referenced above, and you will read where one commentator suggested that “it might be possible to get HPers to consider NonStop as a 'halo' product, but IMO it would require a smoother migration strategy.”

Smoother migration strategy? Now there’s a thought, and I’m sure it’s going to be the center of a lot of different points of view in the coming week. If EBUG events in Vienna and Prague are any guide to what may eventuate in Madrid, migration strategies will be a very popular topic, smooth or otherwise. There will also be many different opinions about the merits or otherwise, of leaving NonStop for something that is today probably one of best known symbols of all that’s considered legacy, the IBM mainframe. Just as there have been many different opinions over the years about leaving IBM’s mainframe in favor of NonStop.

NonStop provides a different architecture from anything else in the marketplace. This has helped NonStop systems to provide far better availability and (linear) scalability than anything else on offer. Much of this architecture, however, is being pushed into the background today and hidden behind middleware designed to better bring NonStop into the “modern world”, usually equated to the Unix-like or even Windows-like environments. Indeed, a later comment posted to the same LinkedIn group’s discussion noted how today, when it comes to open systems and NonStop, “I don't think the difference is as big as it used to be.”

To be more precise, the comment then went on to add “there is the target database, SQL/MX that is much closer to what is used elsewhere than SQL/MP used to be …
people nowadays know what a cluster is: it is very common in the industry. What (HP) NED does not stress enough, IMHO, is that the NonStop server is a cluster, but does not have the complexity of managing one.”

Clusters, without the complexity? How did I miss that! With this simple observation, I realized that, given my own background and all the time I have spent working with mainframes, perhaps I was barking up the wrong tree! Perhaps the success of NonStop has no connection whatsoever to mainframes. Perhaps I needed to move on and look at NonStop differently!

New solutions are coming to NonStop, and they are coming from the Unix and Windows marketplace. Historically, NonStop has proven to be an excellent choice for offloading some mainframe functions and it never seriously challenged the mainframe outright, although today there’s nothing really prohibiting consideration of that possibility. However, a modern NonStop Blades System can certainly displace the need for a Unix or Windows server – particularly when applications reach a size (and visibility) where clusters become a consideration.

When I wrote the post “Game changers!” in February 2009, I reviewed the work Modius was doing in porting a data center environment monitoring application to NonStop. The team at Modius had been deploying their product on low cost commodity servers, but over time they had found it necessary to utilize clusters. The complexity this added to the application was considerable, so much so that I was told that a NonStop’s architecture became beneficial to Modius. “No more issues with configuring a network of servers to ensure high availability – the inherent N+1 redundant architecture at the core of NonStop makes a very persuasive argument – an absolutely mandatory requirement where the ‘availability’ of the whole data center is at stake.”

In the past couple of weeks I have written about AJB Canada, Opus, and Lusis who have all successfully migrated their product offerings to NonStop, and even though it is still early days for each of them, Opus already has successfully deployed its electraSWITCH product on NonStop servers for the State Bank of India! Looking further ahead, I can only see more successful deployments ahead for all three. and each has committed sizeable resources behind these projects. It is not just the migration of applications that we see of late and referenced above, as Infrasoft has built its offering from scratch on the NonStop platform, but these latest application migrations from “open systems” is beginning to put the NonStop server in a different light!

It should come as no surprise to anyone familiar with NonStop and with Unix, and indeed with what has been happening all over the Internet, to read the latest comments posted to the LinkedIn group on Real Time View. “NonStop is a self-managing cluster (albeit with a few unusual interfaces), and should be positioned as such. As I recollect, most of the positioning that was done with NonStop was in terms of comparisons with IBM mainframes, which I think just aren't really that likely to be sources of new customers.”

To be fair, some 35 years ago, when NonStop was born, nobody viewed Unix as a candidate for the deployment of a business application: these were servers mostly used within the academic world - inexpensive, not secure, and not stable enough to use for the serious stuff! Fair enough, but probably of more interest are the comments that immediately followed. “NonStop can compete favorably with UNIX clusters (the Google, Facebook and Yahoo models; loosely-coupled, cheapo, Linux servers are tougher to compete with price-wise, but it takes a lot of interns, late nights and non-standard programming (i.e. no sql) to make those things work at scale) … It's certainly possible to go from a cluster like this to a NonStop cluster …”

EBUG Madrid will generate a lot of discussions. Migrations will be among the most animated, I suspect – the clock continues to wind down to November 2011. From the time that I first heard of this date, during EBUG in Vienna, there are so many differences in opinion as to what should be done and yet, through it all, the choice for NonStop continues to prevail. A future life on the IBM mainframe? I don’t think so! Perhaps a potential move to a Unix cluster? Then think again!

Many within HP may remain perplexed about the future of NonStop (and perhaps they should spend more time at user gatherings), while others may be dismayed about focusing too much attention to how much better than traditional, packaged open systems, NonStop - the “cluster within the box” - has become. True, for many NonStop is still different, but the gap has narrowed appreciatively!

In a parting shot at Oracle and Sun, the exchange on LinkedIn closed with “the opportunity is not significantly different from when Bill Heil (a former head of NonStop under Compaq) said NonStop was a get-well hospital for successful Sun customers.” When balanced with the commentary from partners porting products to NonStop this is a vastly different perspective than had been developed early on, when NonStop was compared to the IBM mainframe. Indeed, when compared to platforms like Unix and Windows, and to paraphrase the earlier comment “the difference (isn’t) as big as it used to be!”

No comments: