blog This might be just my general ignorance of the server market speaking, but it had been my impression that Intel’s x86 family of processors (you know, the ones found in virtually every PC today) had been gradually taking over the entire server market, particularly as the drive to blade servers and commodity cloud computing has accelerated. Commodity x86 servers, to my mind, had largely become the norm. Of course, we still see alternative platforms such as mainframes and Sun’s (now Oracle) SPARC chipsets out there in the wild, but it had been my impression that most new machines were x86. However, according to a media release issued by Oracle Australia this morning (there’s also an accompanying YouTube video), apparently SPARC is still in vogue. The media release tells us:
Defence Housing Australia (DHA) has selected Oracle’s SPARC T5 servers based on the SPARC T5, the world’s fastest microprocessor, to support a new financial management system (FMS) that will be deployed later this year. The new hardware will assist DHA in simplifying its IT systems and potentially reduce cost through the consolidation of its data centre footprint. In addition, the extreme performance expected of Oracle’s new SPARC servers with Oracle Solaris will help DHA gain faster and more reliable access to information, enabling business users to make quicker and more accurate decisions.
“We have been an Oracle SPARC customer since around 2000. We know the technology’s capabilities well and continue to be impressed by the performance and stability of new iterations. The SPARC T5s are going to help us consolidate more workloads onto fewer servers; and we are confident that we’re going to gain a significant performance boost from them; we are looking to achieve at least three to four times more performance/capacity than our SPARC T4s,” said Mike Hehir, deputy chief information officer, DHA.
“By consolidating servers, we can minimise the amount of machinery that we manage and free staff up to focus on value generating activities and new initiatives, rather than just keeping the lights on. We have already been able to increase our BI activities enabling the business to leap forward in regard to the analysis of our portfolio and providing greater business strategic recommendations to the Department of Defence as to the cost of housing and services that we provide, and where we should be moving forward in relation to providing those quality services,” said Shane Nielsen, chief information officer, DHA.
I’d honestly be interested to know where people still see the SPARC architecture being used. I know there used to be a lot of it in Australia’s financial services industry, as well as in government. What applications do people feel are better suited to SPARC, as opposed to commodity x86 platforms? Are there areas in which it just doesn’t make sense to deploy x86? What’s operating system support like on SPARC at the moment? Let me know your thoughts.
Image credit: Oracle, Creative Commons
I think in regulated industries and government which are unlikely to end up running in the cloud anytime soon with a strong UNIX datacenter heritage will be quite happy to buy more POWER / SPARC type hardware.
Its a case of new stuff being bought based on the skills of the IT organisation – clearly there may be a point where a serious consolidation occurs and they switch completely but that implies a big strategic change of the datacenter. Obviously most regulated industries and government are quite risk averse so its not a great surprise to see them buying more of the same.
It will be interesting to see how things change with the arrival IaaS / SaaS datacentres in this part of the world though….as this would be a big carrot to move off UNIX platforms to get onto Intel gear.
Of course, there’s nothing stopping you from running Solaris on x86 hardware.
A SPARC T5 chip has 16 cores, each with 8 threads. That’s 128 simultaneously executing threads on a single chip. The T5-8 server can have 8 of those, and 4TB RAM. That’s a really beefy server in only 8RU.
Which makes a great start for a powerful database server or even heavily threaded web app. Add on top of that Solaris Zones which provide a more lightweight form of virtualisation than what “cloud” offers then you have a reasonably compelling offer.
Sometimes you just want hardware that works and 1 T5 can do a lot of work. Some of us don’t really want to live in shop with hundreds or thousands of commodity boxes. What’s even more amazing is that a T5 can replace an E or F series box at a fraction of the cost.
Presumably DHA runs custom J2EE business systems on WebLogic or some other Oracle app server with perhaps Oracle databases, CRM, Oracle financials, etc. Being govt/corporate IT, they’d take a risk-averse stance and stick to what they know. That and it’d be much cheaper to upgrade standard tech than to migrate to another platform, Solaris or not.
I think that if you are running lots of Oracle databases then you are more likely to use SPARC – especially now that Oracle are pushing them. They are also extremely reliable. Solaris on x86 is no where near as reliable.
I would have thought this was an example of the ongoing death of SPARC . . . consolidating a whole pile of existing machines onto a “fewer servers” is mostly just about extending life of existing apps… hardly a win to one new app on …
I used to work for a large Defence contractor and we used Sparc/Solaris/Oracle for all critical ERP Database systems – only non-business critical stuff got put on to x86 RedHat based Oracle DB environments.
I am aware that SPARC is still used extensively in Defence environments … but are these consolidations of existing environments? Or are these actually new deployments? The point has been made a number of times in the comments here that it seems like SPARC deployments are primarily consolidating old SPARC deployments, rather than being new deployments wholesale.
While I was there (3 years ago) SPARC was seen as for mission critical systems and x86 for less important systems where cost was more of an issue – it didn’t matter if it was legacy or new systems.
The only reason people will be buying sparc hardware is for consolidation of existing legacy workloads. They will not be purchased for new workloads. This is not a sustainable business model.
This had been my opinion of the situation as well.
Obviously I should point out I am an Oracle employee (some bias will now be assigned) however I am not a Systems person, I work in the software business, my observation is that hardware is no longer about CPU design, this was only important when the microsoft/intel duopoly was in full swing and Windows OS on INTEL was the defacto U/I for all computing users. The WEB has changed that for ever our clients are showing less and less interest in the CPU / OS layer, beyond the TCO aspects. In fact if you remove the TCO factor from a decision they loose all interest in “what CPU your offering me”. Engineered systems do exactly that, they remove from your client the need to care about the insides of the box. They only care about the service level it delivers and at what cost to the business, it is under this kind of buyer dynamics that SPARC is prospering – specifically in our engineered systems or EXA products.
hey Rob, if you are saying that clients no longer care about what underlying CPU architecture is in their machines, then I would say this is inaccurate, as it really goes to the heart of the long-term future of their deployments. CIOs and IT managers try to plan for the long time, and there have been question marks about the long-term future of every other platform other than x86 for some time now — hence this article.
I agree that Oracle’s engineered systems are proving popular — but not as general solutions. I’m seeing them deployed as point solutions to solve a specific problem, because they offer an opportunity to generate unique outcomes in terms of performance relative to density. But will this represent the majority of the market? No, I think it will represent a tiny minority.
There are still some big overseas customers out there that don’t trust x86 for jobs they perceive as mission critical or ‘for big iron only’. You can’t convince them otherwise, either. So you have to buy and build if not test on SPARC h/w in Oz.
Apart from that my understanding is engineered big db appliances and legacy apps eg scientific sims etc form most of the rest of the market.
Sun h/w has always been nicely engineered IMO.
There are still some big overseas customers out there that don’t trust x86 for jobs they perceive as mission critical or ‘for big iron only’. You can’t convince them otherwise, either. So you have to buy and build if not test on SPARC h/w in Oz.
Apart from that my understanding is engineered big db appliances and legacy apps eg scientific sims form most of the rest of the market.
Sun h/w has always been nicely engineered IMO.
There are still some big overseas customers out there that don’t trust x86 for jobs they perceive as mission critical or ‘for big iron only’. You can’t convince them otherwise, either. So you have to buy and build if not test on SPARC h/w in Oz.
Sorry Renai, first comment and I’ve fluffed it.
Error messages saying ‘post not accepted’ from Chrome under iOS, hence the resubmits.
Hopefully you can scrub the dupes and forgive me :-/
Comments are closed.