Insight Community

This white paper leads with uncited assumptions and then makes a set of vague claims.  The uncited assumptions immediatley prompt the reader to seek a second opinion, and the claims made do not have enough details in the cost projections to support the assertions made in the summary of the claims.

Here's what I would do to this paper to make it more valuable:

1. Cite sources for claims such as "most companies keep their servers for three to five years".  From my ~15 years of experience, this is completely false; I'd assert that most companies keep servers as long as humanly possible, with 10+years of active production service being fairly common.

2. The consolidation savings asserted in the paper are vague and they do not cite figures related to the savings yielded from switching the actual hardware out, only from the implied reduction in workforce.

First, the claim that 100 servers require a staff of 10 is completely absurd; I know of datacenters with hundreds of racks, thousands of servers that have fewer than 10 admins.  100 servers may require 10 different developers for the different applications hosted on those servers, but swapping out hardware will not affect the software development demands, thus you will still need those same 10 developers.

If the only savings from swapping out the hardware is in the I.T. labor costs, consider that if I am hiring virtualized overseas admins, I am definitely not paying "between US$52,750 and US$82,500."  In fact, the initial hardware outlay may actually be cost prohibitive depending on how cheaply I can get admins.  So I would have to show me more figures on how the hardware actually saves money vs. a similar older configuration. 

The good news is that if I can actually show those figures, I can abandon the "3-5 year" argument in favor of a general switch.  I should be able to demonstrate that the new hardware saves so much money, that companies need to switch regardless of how old their current servers are. If I can't show that, then I am not being specific enough about the savings.

3. The conclusion is way too fluffy.  I would let the facts and figures lead the reader to draw the obvious conclusion... that this is a good investment.

Reassessing Server Costs is generally well written and fairly convincing.   This was refreshing because often technical writers assume the reader has background information not provided in the paper, quickly loosing the reader in a mess of detail.   After a slow start, reassessing gets to the point and provides the relevant details and information without clutter.  

However I do think the very first portion of the paper should provide a more cost savings detail and less Oracle "pitch".  The style is great but a cursory read of the introduction might lead a casual reader to think this is a marketing puff piece rather than the cool appeal to reason, quality, and ROI that a white paper should project, (which this paper does well in the detail sections but not the intro and pages 1-3.    No need to outline the obvious - that cost savings are very desirable and increasingly critical to operational success.     More important to focus on page 4 and 5's excellent tables which offer the key "meat" in this paper - support for the idea that an upgrade will save money through reduction in servers, energy, and labor.    For example the 9 to 1 server consolidation item is very compelling and intuitively appealing.   This kind of move - implemented successfully -  can make the CTO a cost savings rock star.  

Also I think more should be made of how this approach can be used not just to "cut costs", but as an approach to successful companies scaling up their infrastructure and those who have already decided to upgrade the datacenter.    Although more research would be needed to determine if my guess is correct, I think that even when the consolidation approach would be smart for the company, Oracle is far more likely to sell to growing companies than those facing severe economic pressures.    Is this paper too focused on the bad news part of the equation?

Recommendation:   The Introduction (and portions of the first few pages) of the white paper might read more along the following lines, noting the cost savings first and challenging the reader to continue reading about the details of the promised cost savings. 

This white paper details how many organizations can significantly reduce server costs while a the same time improving performance and quality.     Significant recent improvements in server technology such as the Oracle Sun servers with Intel Xeon processors mean that simply working to extend server life in your datacenter may be counterproductive, leading to higher costs and lower performance.  

Cost savings through virtualization, energy efficiency, and reduced management costs mean that the shortest path to lower datacenter costs may be a server upgrade to Oracle Sun servers.

The "Best Practices" white paper left me feeling frustrated at Oracle for failing to deliver any relevant case studies and very little meaty information.    Although the term "White Paper" is now used to include commercialized pitch work, I think most IT pros expect a White Paper to provide high value, actionable information about a topic.    At the very least the white paper should identify several specific items relating to the topic and offer references to other sources.   This paper seems more like it was simply a sales puff piece built around the general concept of datacenter best practices where the author forgot he was talking about "best practices"!   

"Best Practices..." does offer a bit of good specific advice regarding storage integration and the "lights out management" utility, but I think at the very least this paper owes the reader more references to case study information where a datacenter manager reduced costs by $X through implementation of  a specified approach.   The table of "best practices at a glance" is very general - I'm not convinced these ideas are going to help an IT manager, whereas references to specific implementations of these best practices, ideally with enough information for him/her to contact others struggling with the same issues, could be priceless.    For example referencing several high value detailed Datacenter IT papers like this would be helpful:   Datacenter Best Practices from Lawrence Berkely National Labs.      Here, we quickly learn that energy use has become a critical concern in the datacenter and we are presented with data and alternatives to address this.    Obviously the Oracle "Best Practices" paper is not attempting this level of detail, but it's also obvious this paper is more interested in selling a server than serving the reader.

Although clearly Oracle needs to sell Oracle servers, I think the best approach in the modern IT / socially robust online environment is to write to the topic rather than to the marketing objective. This builds credibility and will lead to a stronger long term relationship with the customer, as well as create a better experience for those who interact with the material at other levels.

I reviewed the "Reassessing Server Costs for Midsize Companies" whitepaper, because I thought it might help me understand how to build a small or mediumsized datacenter, or decide to outsource. But all I got was a lot of unsubstantiated claims about how these servers could lower costs. Yes, I can buy that compared to a five years old server, the capacity is nine times higher. But there are no examples that let me understand how it actually would save money, and there are a few very dodgy statements - for instance, that management costs would shrink, which is a function of the operating system and applications rather than the server architecture; and that operating system licensing costs would shrink, which of course is bogus if you use Linux.

In conclusion, does not help me at all, and makes me wonder if the hardware division of Oracle is going to be as bad as the software.

I reviewed the "Best Practices for Managing Datacenter Costs via Application and Server Consolidation" white paper. It starts with making an erroneous statement: That "best practices" should be a buzzword. Actually, best practicies are a well established method of spreading successful practices (even to the extent that they are standardized in the IETF).

Unfortunately, the rest of the paper is much like that. It makes a lot of loose statements mixed with things that are widely accepted, but does not make any attempts to substantiate these (apart from measurements from Intel showing that their own processors draw less power when idle).

I would have expected something that showed how you actually can save energy by resettning your energy consumption, instead of "administrators can customize its settings (using monitoring and management tools built into Oracle Solaris) to best meet their IT needs and to ensure the optimal balance of power consumption and performance. Automating these controls helps simplify the IT environment, freeing IT staff to work on more business-critical tasks." Yeah? How? Which settings are most effective - and hence should be best practice? And does it really help me to know that the Oracle Ligthts-Out Manager is included without cost? How hard is it to use, and does it actually save resources and help me run the data center more efficiently? The paper is all like that.

Sorry, this was not particularly useful. Must be written by a marketing person for another marketing person.

I reviewed the "Why Solid-State Drives Usage Scenarios Are Expanding for the Datacenter" white paper. Here I finally got some actual technical facts about how these can help, although I do not know if the increased shock resistance is so useful in a server deployment.

In this white paper, the energy savings from SSD vs HDD could be lifted, as well as the integration of ZFS with the SSD. But do they really give better I/O performance over the HDDs? I would have liked to see some detailed discussion on this, and also on how using them as caches can improve performance.

This white paper was rather useful, but it could be much more useful if Oracle lost the marketing statements and focused on telling how the tools they provide work (for instance, using the SSD as a cache is not automatic in Solaris, is it? The white paper seems to contradict this). So more consistency and focus on the advantages would be helpful. Otherwise it will be hard to motivate the additional cost.

I reviewed the "The New Economics of Midsize Enterprise Computing" white paper. This paper wants to make the claim that by exchanging the servers for the new Sun servers you can save a lot of money and make your data center more efficient.

Well, there is not nearly enough proof. Most of the problems the paper purport to solve could actually besolved by virtualization and more efficient application management (perhaps it could be useful to say something about that?). There is also no proof that it is the Sun servers, rather than the Intel processors, which make for the savings.

It would have been nice to see some detailed comparisons instead of generic statements, and also something which showed that the Sun part matters - why is it important that the "lights out manager" is included without cost? What does it actually do, and how do you use it?

There are lots of potentially good statements, but they would be a lot more credible if measured by some independent entity. And of course, it would help if the writers of the paper had talked to some people who run data centers.

I reviewed the paper "New Blades and Networking Solutions Ensure Solid Return on Investment" and it starts out really badly. Really, saying that "text and binary-code-based information evolved to multimedia files, high-resolution graphics, and massive relational databases" does not give you a very high level of confidence in the understanding of the evolution of storage and processing needs on behalf of the writers.

The rest of the paper is slightly better, but I would dispute that you usually have to have one NIC per blade. Nor is there any serious comparison against non-blade deployments; or any discussion on how the system actually improves the performance and decreases resource usage. This would have been easy to do: If this is really a new feature, just take an old version, run the same software on them, and show the result. But no, just a lot of statements on that it is useful, nothing about how it actually works, or how much it helps.

I reviewed the "Oracle Solaris Operating System—Optimized for the Intel® Xeon® Processor
5600 and 7500 series" whitepaper. This was finally useful! I got to find out a lot of how this optimization is done (although there could have been more), and I learned how the power saving works. I would have liked to know more about the integration of applications with virtualization, though.

The security part is interesting but could have been a whitepaper of its own, it is important and useful. There could also have been a different structure, for instance the references both in the middle of the text and at the end feels repetitive.

Keep this whitepaper and throw away the others. That is my advice.

The whitepaper reads well and highlights some of the key reasons why companies suffer from server sprawl.

Having just completed a review of the application and server blueprint for a midsize utility company in New Zealand the reasons from the server proliferation were very much as depcited here, each new application resulted in a new server and the SDLC in use suggested comparable environments as the application move through the design, test, production phases.

Another dimension mentioned here is the users desire for services from IT.  These usually become gold plated as standard with everything either being perceived as a priority one service or being hosted on priority one designed infrastructure.

The secret to unlocking some of the latent value of the server room came in three parts:

1.  A complete review of the application to server mapping through a service lens.  This promoted a rethink of how services were grouped together. This then allowed us to host applications with similar service levels on servers setup for that degree of uptime and reliability.  The first step in modelling the clsutering was to look at the type of service, the charateristics in terms of CPU and Memory utilisation, and the profile by way of platform requirements.

2. A new approach to service clustering allowed a more pragmatic approach to the use of virtualisation to get around some of the variety in platform needs.  Typically the base was Microsoft Server to prudent use of virtualisation software allowed layering of mutiple platforms on a standard core server.

3.  Change the infrastructure support provider.  The other major reason, not mentioned in the white paper, for server proliferation was the role that the vendor took.  Providing both outsourced support and maintenance as well as consultancy for new applications the contract had no clause for productivity gains.  In fact the opposite was true that the vendor was incentivised for adding servers rather than taking them away.  For companies finding themselves in this situation it is worth looking at the term of the contract and seeing if you can find a better deal in the market today.  We helped the company retender and negoitate a new outsourcing contract and built in a productivity lever to gain increasing returns over the five year life of the deal.  This then incentivises the vendor to maximise the use of conslidation and virtualisation to reduce the size of the data centre, thereby reducing operating costs over time.

To this end I the white paper deals with the first two components well.  However as a sales pitch my fear is that it actually encourages the third component.  My dialling back the sales element and highlighting the softer causes of the problem, insufficient IT governance driven by poorly incentivised outsourcing agreements, it would strengthen the key mesaage.

The first impact from this white paper is the amount of white space in the opening pages, for me this was distracting from the message here.

From a technical perspective the paper is strong, assuming that the intended audience has some knowledge of the issues at hand.

After reading through this a couple of times the message of performance of SSD versus traditional HDD's comes through strongly.  However, the unanswered question is the cost comparison between SSD and HDD.

What would strengthen the paper in ths regard is the cost vs performance comparison.  The price of SSD's is dropping from where they were in the early days, which would be prohibitive for enterprise usage scenarios.  A graph that shows how this price point has moved in recent years and how the need for IO rates has increased over the same time would allow

A. datacentre managers to calculate the cost of IO needs under SSD usage; and

B.  help managers and professionals extrapolate the trends of where the price point and IO rates will be in the near future to help them make a more informed decision of when to adopt SSD in lieu of HDD storage.

Best Practice is an aspiration and can never be reached, but organisations should try.  This white paper goes some way to lay out some of the approaches available to an IT organisation.  It is true that there are a number of best practice frameworks out there, including ITIL.  The problem that this paper tries to address is the reality of soft processes, such as the SDLC chosen by the company, on the datacentre size, operational complexity, and cost.

From a technical perspective the paper has the right tone and content, in terms of consolidation and the use of virtualization.  The piece that is missing for me is the role that IT governance and the visibility of the services offered by IT for the business.  The cause highlighted for the symptom noted here is the one application = one server formula.  If the business knew the true cost of the service, regardless of charge back, this server proliferation would be slowed down.

Oracle themselves has some products in this space, a suite of project portfolio management applications, that would be a good addition to the server sales pitch.

The Oracle Solaris Review paper offered excellent insight and stands as a great example of how a technical white paper should read.   The style is readable while the paper offers detailed introductions to most key aspects of the Solaris operating environment.    Rather than feeling "pitched to" in this paper, I felt the author was dedicated to helping readers understand how to best use the Oracle Solaris / Intel environments in real world implementations.  

My favorite part of the Solaris Review was the very effective use  of the many external references and a robust technical appendix, allowing the reader to click out to more details on several topics, organized according to the chapters of the white paper, and also to dig deeper into technical issues that may be of interest to the reader, especially server administrators.

This feature alone would have made this a smart paper - ie a simple outlin of the key points plus links to each resource.  However in the Solaris review the author provides a quality introduction to each optimization point, and then provides the external "for more information" links.  Excellent use of the online environment to expand the paper. 

Clearly in this case the author was not only extensively familiar with the Solaris / Zeon environment and implementation, he/she seemed to anticipate several questions and concerns of the reader and provide both quality introductions to the topics and follow up information resources.    

A+

Although "The New Economics" reads well and delivers some marginally helpful insight, I could not help feeling throughout much of the paper that this was written by a marketing person who was unfamiliar with most of the technical issues and -more importantly- the key economic issues of Midsize Enterprise computing.    A quality white paper in my view would discuss details of midsize enterprise IT budgeting, ideally with case studies in cost savings that the reader could apply to their own experiences.    This is an area where data and case studies are easy to find with a few phone calls and online research  so there is no excuse for simply asserting the remarkable ratio of 9 to 1 as a server consolidation estimate and the spectacular 8 month recovery of costs estimate.      Readers who *believe* these assertions are likely to become buyers, so it's very important to outline how this number was obtained, how realistic this is given other contraints, and how much actual cost saving per server can be expected.   Although estimates rather than case studies are acceptable, real numbers are essential here.

Less explanation is needed about why cost savings are desirable - nobody needs to be convinced of that.   However more detail is needed about the math of the "New Economics".   Does it require new equipment and if so how much will that equipment cost?    Even a representative company IT budget example would be fine - say a company with 100 old school servers vs a "New Economics" company with 12 optimized, virtualized Oracle servers.  

The key piece of meaty information in the paper does not come up until near the end, and does not really hammer the point home with any examples or cost estimates.

These models are designed to drive virtualization and consolidation initiatives that will reduce the total number of servers in the datacenter, which can help reduce other key cost components such as administration, software, backup, and maintenance agreements. And while these new servers are designed with virtualization in mind, their raw performance provides huge consolidation gains. In fact, some customers have seen 8:1 or even 10:1 server count reductions. 6

The footnote simply references some customer feedback at the Oracle website and provides little to support the idea that a server upgrade is needed to obtain this level of performance enhancement.    For example could you simply use better virtualization and consolidation with your current servers?    Even though the answer to that may be yes, and may not help Oracle's sales in this particular case, good social media marketing requires that this point be addressed at least in a cursory fashion.    White papers are a good tool for Oracle to gain trust, customer credibility, and enhance online authenticity.   To accomplish that goal the object of the white paper should be robust, helpful, and mostly unbiased information.    Most readers understand that Oracle will of course recommend Oracle systems, but building the papers around that notion is arguably counterproductive.     

This paper is well-written and contains some valuable information but is outdated and not specific enough to Oracle's products to be effective. The introduction is thoughtful, outlining many of the macro data center issues affecting IT, and it moves smoothly into a discussion of virtualization that is understandable and logical. However, the paper begins to get vague here, talking about virtualization ratios that seem out of line with current practice (5 to 1) and older data (see footnote 2).

The discussion of the Intel Xeon 5500 is valid, certainly, but non-specific and outdated. Most vendors, including Oracle, are now shipping Xeon 5600 processors, so the constant references make this paper seem an anachronism with a new vendor name attached.

More critically, however, the discussion supports a move to using Intel's CPUs rather than Oracle's servers. Any vendor could offer similar capabilities, and any reader would know this. Even the discussion of Oracle-specific features (Integrated Lights Out Manager and storage) seem strangely non-specific.

The eponymous Best Practices section is refreshingly strong, with a solid list of practices. I would have loved to see this be the focus of the paper rather than generalizations about virtualization and Intel's offerings. Sadly it falls at the end and ends abruptly with a short sales pitch.

In summary, although a reader unfamiliar with the range of IT infrastructure offerings might be impressed by this paper, these are not the target audience. Any potential buyer would be unimpressed by the paper generally and its coverage of Oracle's products in particular.

The New Economics of Midsize Enterprise Computing white paper presents a more-compelling case than the similar Best Practices for Managing Datacenter Costs via Application and Server Consolidation. It is a readable account, flowing naturally from the business issues to specific technology solutions. But it seems out of data with the current corporate ownership and product line.

This Midsize paper offers more specifics than the Best Practices paper, including a welcome discussion of SSD. But the ball is dropped in terms of differentiating these products from the competition: Although Oracle/Sun offer many compelling product benefits, none are called out here. There is no mention of the variety of CPU architectures offered, for example, or the unique capabilities of Sun's storage.

As we wind through the paper, the absence of Oracle's products is jarring. Mention is repeatedly made of MySQL but the namesake database of the new corporate parent is glaringly absent. It is obvious that this paper was written for Sun and updated with the Oracle name only. Undoubtedly, a writer could easily refresh it to be more current. It is also odd when reading to stumble across a mix of occurrences of the "Sun" and "Oracle" names with little logic to their placement.

The focus on the Xeon 5500 series is also an obvious anachronism, with Oracle now offering M2 servers with the new Xeon 5600. The footnote on page 5 should also be re-evaluated. If Turbo Boost is one of the major features of the Nehalem CPU line, why undermine it with a "Warning"?

It is nice to see some graphics in this paper as well. Whitepapers should not be all black text on white paper!

Finally, the "Why Oracle and Intel?" section is woefully short and unsatisfying. Again, there are many reasons to select Oracle's Sun server line but none are mentioned here. Any reader would notice this overly-short section and question "why indeed?"

I was excited to read this whitepaper, as blade servers and converged networking are a special interest for me. But I was frustrated throughout, as the discussions of the benefits of blade servers and the capabilities of the Oracle NEM failed to enlighten me. I finished the paper more confused than I started, and I feel that it could be vastly improved with some technical assistance and visual aids.

The discussion of consolidation and blade servers starts well enough, but quickly derails with a seemingly-outdated discussion of increased cooling for blades and the statement that "each blade requires its own NIC." As demonstrated in this paper, neither of these statements are true of Oracle/Sun, and every other blade vendor has overcome these obstacles as well. An informed reader would know this and it would set them on the defensive before even beginning the paper!

The real meat of the paper is located in the next section, "Efficient Servers, Virtual Networking". This seems to be a mash-up of two separate sections, one focused on Intel's Xeon 5500 and QPI and the other describing the Oracle/Sun NEM technology. It would be best to separate these two mostly-unrelated sections.

The ensuing technical discussion of NEM is simply frustrating. It is not specific enough to really inform about what NEM is, and yet repeats statements haphazardly. After reading and re-reading, I deduce that a blade chassis can have up to 2 NEMs and each NEM sports 2 10 GbE links. A diagram would have been very helpful here, as would some pruning of the text.

Once one grasps the really quite-simple NEM technology, the discussion is even more exasperating. We learn that "as many as 10 server modules" can share a 10 GbE NIC, then are informed that "some estimate" the cable reduction to be "10 to 1." I imagine every attentive reader would be fairly certain of this, no estimate required!

We are never fully informed of the switching infrastructure required, though a reduction is called out. Why do we need less switching? Does NEM have internal switching like some competitors? Or is this simply a way to say that fewer links require fewer switch ports?

As someone with a storage background, I also found the short mention of serial-attached SCSI (SAS) confusing. How does a 10 GbE NEM "improve the integration" with SAS?

In summary, this paper would be much-improved by a refresh with additional technical expertise, some thoughtful editing, and a graphical representation of the blade chassis and NEM.

Since my background includes a great deal of focus on enterprise storage, this whitepaper was well within my comfort zone. It does a good job of laying out the case of using SSDs and includes the use cases for them, which I appreciate. There are a few areas that could be improved, but it stands up better than most of the others in this series.

My first suggestion is to be more specific. The performance of SSD is mentioned repeatedly with objective numbers (e.g. "35,000 IOPS") but comparable numbers for HDDs are not included and would have strengthened the case. Similarly, claiming that hard disk drive performance improvement since 1996 has been "modest" does not tell the same story as a "175x" performance gain for CPUs. Readers notice that the comparisons are one-sided, and this is unfair since a stronger argument could be made if numbers had only been included.

Like the rest of the series, this whitepaper is all text. The addition of some call-outs, boxes, tables, or graphics would have been helpful. This is especially true of the section of SSD use cases, where a "first", "second" list is mentioned in the text but not highlighted in a box.

In all, this is an effective paper that could be improved with a few touches.

This paper starts off strong but falls down by the end. I was engaged by solid writing and a good problem statement, and found myself agreeing with the discussion of the impact of new server technologies and consolidation efforts on the old five-year retention strategy.

But the numbers just don't hold up. Claiming that server consolidation ratios are directly proportional to operational headcount reductions is simply wrong. If we achieve these consolidation ratios by implementing server virtualization, we have added a new and complex technology to manage, potentially increasing operational demands. Even if we simply run multiple applications or instances on a single physical server, we cannot assume we will need proportionally less management personnel. Although some headcount reduction is warranted, it is much less than suggested here, so this should not be the foundation of this paper's claims.

The final section was simply puzzling. Introducing new technologies like Java and thermal management in the conclusion was surprising and did not support the paper's claims any more than the operational ratios. And once again, no mention is made of Oracle's place in this picture (presumably since this paper pre-dates the acquisition). This paper should not be used without a major overhaul.

This paper is easily the best of the bunch. It is packed with solid information and the authors seem to have a solid understanding of the subject at hand. It is also up to date regarding Intel's new Xeon 5600 CPU technology and the Oracle acquisition, two elements missing from all the rest of the papers. The Solaris paper really needs no improvement to stand on its own.