Big Irons or blades, Itaniums or Xeons; A mind-boggling array of choices confronts CIOs in the server space. And on the business side, will the market grow by 8 percent as predicted by IDC or will it stay flatter than a can of Coke opened an hour ago? Prashant L Rao tries to decipher the riddle wrapped within an enigma that is the great Indian server bazaar

Servers: Top Trends

• Itanium won’t have much of an impact in 2002-03. AMD’s Hammer might shake up things a bit.
• Intel’s 32-bit platform will continue to thrive as Intel server revenue lead over RISC expected to continue.
• Server consolidation might take place-telecom market may lead this trend.
• HP-Compaq’s possible merger might make things difficult for Big Blue and Sun.

After a horrible year in which the server market shrank, it looks like we’ve finally hit the bottom.

That’s cause for some cheer in the Indian server market, even though no one is expecting a quick recovery. The climb, if it takes place, will be gradual, meaning 2002-03 will be the year of growing slowly. IDC predicts a growth of 8 percent, but vendors seem far more pessimistic they expect a flat market.

Despite all this, there is a lot of activity in terms of technology and platforms as vendors come out with diverse offerings hoping to steal whatever market share they can from each other in a flat market.

Gartner believes that open standards in hardware and operating systems will lower entry barriers, while server appliances, especially caching servers, will provide low-cost ease-of-use and capabilities benefits, forcing traditional server vendors into a defensive stance.

The processor front will see a lot of activity: the long and winding road toward Itanium will continue with furious development activity from the likes of Microsoft, Compaq and HP and of course, Intel. Foster and Xeon with hyperthreading will be launched. AMD is the joker in the pack. Presently the company has a negligible presence in the server market. Its upcoming 64-bit chip, Hammer, could change all that.

The RISC market is no longer the 800-pound gorilla of the server market. It is playing second fiddle to SIAS (Standard Industry Architecture Server) which has the lion’s share of revenues and units. That said, competition is still fierce. Sun leads but IBM is mounting a very aggressive challenge. The HP-Compaq merger is the wild card here. If everything works out on that front, both IBM and Sun could be in for some tough competition.

2001-02 saw vendors touting server consolidation as the next big thing. With the recent launch of blade servers, taking rack density to new heights, it remains to be seen if corporates will go in for centralised set ups with hundreds of rack servers consolidating their existing hardware, or if they will prefer to go in for mainframes or Unix SMP boxes.

Here are what we believe will be the defining trends in Indian server computing in 2002-03.

* 64-bit x86 will be a no-show for the greater part, with some activity towards the end of the fiscal. Itanium will be a no-show in 2002-03. Hammer is a question mark. It could see AMD entering the server segment in a significant manner for the first time.

Intel’s IA-64 architecture-based Itanium processor shipped late, and early sales have been scanty, to say the least. That’s understandable, however, when you consider that Itanium Mark I was just the first of many processors to come. Think of Itanium not as a production CPU but a development platform for independent software vendors (Microsoft and others) to get their products optimised and fine-tuned for the Itanium processor family.

Itanium is potentially the ace that could get Intel the kind of dominance it enjoys in 32-bit computing in the rarefied world of 64-bit. However, there are still many ifs and buts hovering around the Itanium family.

The problem lies in the fact that IA-64 is not backwardly compatible. Enterprise applications will have to be recompiled. Therein lies the rub since Itanium’s IA-64 architecture works very differently from the 32-bit x86 line that made Intel a household name. Itanium is not about raw processor speed but about VLIW (Very Long Instruction Word), which basically means that the compiler needs to be optimised to keep the CPU’s multiple pipelines filled with instructions. While a very rudimentary version of pipelining was there in the Pentium 4 and Xeon, Itanium takes this technology to a new level, one which software developers are still struggling to understand.

“IA-64 is in development mode. People looking at IA-64 are looking at developing applications on the platform this year. It will go mainstream with McKinley and 64-bit Windows and this will gain acceptability in 2003,” says Pallab Talukdar, director-enterprise products, Compaq Computer India.

That said, Microsoft, Intel, Compaq and many others are working hard to develop compilers and OSs that will take full advantage of Itanium. Analysts seem to see Itanium doing well in the medium term. Itanium will replace all but two RISC CPU families by 2006 Sparc and Power4. Barring Sun and IBM, the other RISC vendors will move to Itanium. Gartner has forecasted that midrange and high-end Windows servers will increasingly be built on the Intel processor families (IPF). Itanium technology is expected to trickle down to commoditised markets as prices drop. Gartner predicts that by 2006, IPF-based servers will have a 20 percent share of the overall server market by revenue. Add Intel’s IA-32 (Xeons) and you could see the vendor have a staggering two-thirds of the server market by that date. Revenue for new 64-bit 8-way Intel servers will exceed that of new 32-bit 8-way Intel servers by 2004.

HP’s top-of-the-line Superdome is already Itanium-ready. Gartner believes that by 2003, HP will offer the option of IA-64 partitions along with PA-RISC partitions for customers wanting to start migration. HP will continue to deliver PA-RISC enhancements until 1H04, after which it will be Itanium all the way. What’s in it for HP? The Itanium family will let HP market Superdome to high-end Windows data centre server buyers.

Itanium is essentially aimed at becoming a replacement for Sun Sparc and IBM Power in markets such as technical and financial workstations and rack-dense servers for telcos and ISPs.

* 32-bit computing will continue to thrive. SIAS server revenues have already beaten Unix revenues. That should remain so in 2002-03. Xeon will continue to do well.

While Itanium looks like it will give Intel a foot in the enterprise door, what of Xeon and Intel’s 32-bit line? Xeon will continue to do well with the high cost of Itanium systems and lack of enterprise software (at least initially) leaving it with a huge chunk of the Intel server market for the next three to four years. According to Gartner’s John Enck in “The Future of IA-32 Computing” IA-32 (Intel’s 32-bit architecture) still has plenty of life left. With desktop applications mostly incompatible with Itanium, and IA-32 chips running cooler (a bit plus in high-density rack/blade environments) and a clear roadmap from Intel to keep 32-bit processors going till 2006, IA-32 will continue to be the most successful low-end server platform. Intel’s pulling some interesting tweaks such as hyperthreading in Foster that should keep Xeon an evolving platform for the near future.

* Server consolidation, Blades or big irons?

Almost every server vendor we talked to spoke about consolidation either of servers or of storage or both. The million-dollar question is whether their customers are convinced about the idea of replacing a swarm of older, diverse machines with a few big irons or racks of blades? In the present environment where businesses are trying to bring down costs, consolidation makes sense on a broad level. It’s up to the vendors to reduce disruption during the process.

That said, India is still very much a country of low-end departmental servers doing simple jobs like file, print and, of late, messaging. Barring a few IT-savvy organisations that have implemented Notes or Exchange at the national level, most organisations are still doing branch automation. Mind you, this is not just in banking, but it is also true of the government. Talk of e-governance has been more hype than substance.

Till last year, the software sector was the silver lining for the RISC/mainframe crowd, but with software reeling and struggling to recapture its glory days, sales of big irons have virtually ceased, say industry sources.

Most buying is taking place in the financial sector, government, telecom and the like. Of these, a huge number of sales to banks are for branch automation where servers are used to hook the dozen or so PCs in a bank branch. This is where HP and Acer have played to their advantage. The government as we all know is notorious for buying hardware and then leaving it to collect dust, barring a few organisations like the Railways which have done amazing things with IT. So, who’s going to do the consolidation? Large enterprises running enterprise apps would already be running SAP (fill in your favourite enterprise app’s name there) on high-end systems. Software companies are unlikely to make large purchases this year.

That leaves telecom, which is where a lot of action is happening. Compaq is very strong here, as is Sun. But the telecom battle will be fought with rack servers and blades. Compaq appears stronger here and chances are good that Compaq will end up with most of the rack action.

India has never been mainframe country. Big irons sold here primarily as a development platform for software exports. So if server consolidation does take place it will be on Unix boxes, high-end SMP Intel machines or using racks of blade servers.

* HP+Compaq could give Big Blue and Sun a run for their money, possibly even overtake both as the top server vendor in both the Intel and Unix segment.

If the HP-Compaq deal comes off, it will create a merged entity that will have some synergies both are moving to Itanium in the medium term, for instance. Gartner Dataquest has estimated that HP+Compaq would have a worldwide server market share (based on total vendor-shipped revenue in 2000) of 28 percent vs. IBM’s 25 percent and Sun’s 18 percent. The downside is that HP and Compaq have produced similar products and technologies in server hardware (HP’s Superdome vs. Compaq’s AlphaServers), operating systems (HP-UX vs. Tru64) and system software (HP OpenView vs. Compaq’s Insight Manager). Gartner believes that some products and technologies will have to go and turf wars will erupt and has recommended that enterprises should carefully assess new-generation systems, such as blades, for upgrade and support continuity, and migrate OpenVMS and other legacy software without awaiting the outcome of the acquisition. This kind of caution could prevent the joint HP-Compaq army from unseating their rivals.

Despite predictions of an almost flat market, server vendors intend to launch a slew of products in 2002-03. While it is too much to expect that this would spur the market to earlier levels, it could help light a slow fuse that would lead to a better situation by Q4 2002-03 and create a decent foundation for the next fiscal. Yet, we should keep in mind that while the Indian market is battered the rest of the region is doing much worse. India is still the fastest growing country in ASEAN/South Asia, says IBM’s M Ganesh, vice president, Enterprise Server Group.

Server consolidation - more than just hype

Server consolidation is the new buzzword in the server market. Gaurav Dua, senior IT analyst-Technology Consulting Practice, Frost & Sullivan India says, “In developed countries this concept has already gained lot of acceptance amongst end users. However, in the case of India, the market is still in a nascent stage. The future outlook is positive. A beginning has already been made in the country and many organisations have already embraced this new concept. Undoubtedly, this concept offers lot of benefits to organisations, and over the next 2 to 3 years we will witness many customers going in for this technology to optimise and simplify the entire existing IT infrastructure of the company.”

Many enterprises in India are considering the deployment of enterprise solutions such as ERP, SCM and CRM in order to streamline their business processes and increase efficiencies. Here, server consolidation makes sense.

There are a large number of companies (both MNC and domestic) in the country who have their branches/offices spread across the country, especially in the banking and FMCG sector. Putting all servers in one server, or at most a few, dramatically reduces the total cost of ownership (TCO). This is primarily so, because if one takes into consideration the administrative cost, power consumption, floor space, cost of acquisition of new servers, installation and maintenance of software and files, it not only leads to duplication of effort but also turns out to be an expensive proposition.

“Some companies in India who have already taken a lead in this regard are the ICICI group (ICICI and ICICI Bank), Telco and ANZ Grindlays. ICICI, for instance had acquired more than 130 servers over the years, which were spread across the country. With the help of server consolidation methodology, it plans to reduce the number of servers to just four. Thus, it is very evident that server consolidation in India is no more a hype but is quickly turning into a reality,” adds Dua. The adoption rate has been slow in the country primarily because it takes time for any new concept or technology to gain acceptance in India.

Most Indian companies are caught in a vicious circle should they go in for decentralisation or centralisation. Due to the high cost involved in buying mainframes, a large number of companies in India directly went in for the client-server model. Changing that back to a centralised environment would mean additional investment in hardware and software, and many companies in India would shy away from making a huge investment in high-end servers, Internet connectivity and storage solutions.

But on the other hand, gradually companies are realising the benefits accruing to them as a result of server consolidation in terms of data or application integration, neutralisation or a consolidated environment.