Advertisment

No Benchmark Business Here

author-image
DQI Bureau
New Update

Sunil Kapoor is the CIO of an upcoming IT-savvy hospital near

Chandigarh. His mandate–the creation of a paperless environment. As Kapoor

goes about selecting hardware, he is subjected to marketing spiel from almost

all vendors operating in this segment. He has heard enough about benchmarks and

figured that all vendors, somehow, top the list in their respective categories

with respect to some parameter or the other. Kapoor has found a way to cut

through the hard sell. He has devised two simple rules: for systems with a

smaller applications budget, price is the key determinant while for bigger and

core applications, he tells the vendors, "Forget the benchmarks, lets run

the application in my environment." And the decision is based on these

results. No bench mark business here, declares Kapoor. However, it is difficult

for vendors to operate without benchmarks. More on that later.

Advertisment
Benchmarking:

How Can CIOs Tackle It?
  • Complete results: Vendors usually give benchmarks only on

    products that top the list. To have a fair comparison, as for the

    complete results of the benchmark test either from the vendor or from

    the rating agency’s website.
  • Define internal benchmarks: Does a SPECint95 rating of 60 of

    one vendor compared to 50 of another make a difference to your

    operations? Define internal benchmarks and balance them with price and

    performance. Make sure that vendors provide them without additional

    charges.
  • Interaction with the technical team: Remember Dilbert, the

    engineer, and the usual marketing team dilemma of taking him to

    clients. The reason — Dilbert talks the ‘real truth’ much to the

    marketing team’s discomfort. The same happens in the real world.

    After the initial talks with the marketing team, insist on interacting

    with the technical team. The response - usually honest and to the

    point. No frills added and no false assurances. Very Dilbert like and

    of course to the consternation of the marketing people.
  • In-house testing: Benchmarks are in a simulated environment.

    The only way to make sure that you get correct information is to get

    it yourself. Do it in-house and seek to match your internal price

    performance benchmarks. This will check the application in your

    environment handled by your IT staff.

India is one of the fastest-growing markets for servers and

all the major vendors want to have the biggest share of the cake. In FY 2000-01,

servers saw highest growth in value terms and revenues shot up from Rs1023 crore

to Rs1805 crore growing at 76% (DQ estimates). According to the DQ CIO

(September 15, 2001) CIOs say investment will continue in security, database and

network applications. To that add the discounts. The market remains hot for

servers and benchmarks are here to stay.

Benchmarks and vendors

Advertisment

CIOs need to differentiate between fact and the marketing

brouhaha being created by various vendors. While vendors need to address the

core issue of creating the necessary mind share by advertising, this is usually

at a broader level like IBM’s eServer campaign or Sun’s earlier ‘dot in

the dotcom’ campaigns. However, to demonstrate the product superiority over

others given the same processors and similar operating systems (OS) and

applications, the only option available is benchmarking.

Comments M Ganesh, Country Manager, Enterprise Server Group,

"Benchmarks ensure that the marketplace has a fair and useful set of

metrics to differentiate candidate systems. Also, benchmarking is important for

mid-range and high-end systems like Unix servers and storage systems, since they

are high value products with sophisticated technology and made for specific

computing applications." Adds Kamal Dutta, country sales manager, Unix

systems, HP, "Benchmarks provide us with the key to differentiate our

products from others and place before CIOs."Multiple benchmarks indicate

how a particular vendor’s system is likely to perform compared to another. The

most common benchmarks are- Processors like the SPEC (Standard Performance

Evaluation Corporation), generic ones like TPC (Transaction Performance

processing Council) or application specific performance. eg SAP.

Benchmarks and the CIO

Advertisment

So, how important are benchmarks in the CIO’s purchase

decision? The unanimous answer is ‘not very.’ According to the CIO’s we

spoke to, the key problem with benchmarks is that ‘somehow all the vendor’s

products top the results.’ It is true that most of the benchmarks are endorsed

by independent third parties and hence have a high degree of credibility.

Theoretically, it should help the CIO’s’ in comparing the performance of

competing products. But in reality, the CIO’s problem is a number of

benchmarks and vendors keen to talk about only those benchmarks that fit in with

their marketing compulsions.

For example, the SPEC standard has about seven to eight broad

benchmarks each with sub divisions. SPECint is subdivided into SPECint95 and

SPECint2000. This makes it easier for the vendors to publish only categories

where its products are superior. Adds Venky Mahadevan, VP-IS, Coca Cola India,

"While benchmarks are important , we did not use formal benchmarking tools

to evaluate server performance."

Rajeev Arya, Director — IT and Business Systems, Amway

points out that benchmarks are being done in a ‘controlled environment".

Comments Sunil Kapoor, Head-Information Technology, Fortis Healthcare, "We

don’t go with the results of benchmarking as we want to know the system

performance in our environment." In this respect, MNC’s with globally

defined systems and platforms are more fortunate. Ravi Parasuram, GM MIS,

Planning and Control Division, Sony India says, "Globally, Sony is on the

AS400 platform. So there is no discussing benchmarks with vendors".

Advertisment
Some

Prominent Benchmarks
  • Transaction Performance Processing Council: The TPC is the

    widest used benchmark. The benchmarks look at the transaction

    processing and database. A typical transaction, as defined by the TPC,

    would include the updating a database system for things like inventory

    control (goods), airline reservations (services), or banking (money).

    The TPC produces benchmarks that measure transaction processing (TP)

    and database (DB) performance in terms of how many transactions a

    given system and database can perform per unit of time, e.g.,

    transactions per second or per minute. In TPC-C, the benchmark is

    centered on the principal activities (transactions) of an order-entry

    environment and the TPC-H looks at decision support benchmarks. More

    info available at www.tpc.org.
  • Standard Performance Evaluation Corporation: The SPEC

    benchmarks methodology provides a standardized suite of source code

    based on existing applications that has already been ported to a wide

    variety of platforms by its members. The SPEC has evolved into an

    umbrella organization encompassing three diverse groups namely the

    Open Systems Group (OSG) focussing on benchmarks for high-end

    workstations and servers running open systems environments, the

    High-Performance Group (HPG) benchmarks targeting high performance

    system architecture, such as symmetric multiprocessor systems,

    workstation clusters, distributed memory parallel systems, and

    traditional vector and vector parallel supercomputers and the Graphics

    Performance Characterization Group (GPC) in the area of standardized

    graphics performance tests (for more info www.spec.org).
  • Application-specific Benchmarks: At the final layer are the

    application specific benchmarks like SAP, Lotus Notes and Oracle

    Application Standard Benchmarks. These are undertaken by organizations

    to evaluate a range of cross platforms for a particular application

    (refer to respective application vendors for details).

Says Ganesh, "To effectively use a benchmark, it is

imperative to carefully evaluate its characteristics against those of the target

workloads." CIOs need to take the flaws of any testing methodology into

account. Testing in a simulated environment is likely to be done by users who

are familiar with the working of the system as against new users in the actual

environment.Results may be skewed due to this. In house testing can be the best

tool available or alternatively define your required metrics and make the vendor

accountable to deliver them. Where does that leave benchmarking? Millions are

spent on getting top rating benchmarks and flashing them in marketing campaigns.

Well, CIOs could use them to narrow down the field and shortlist technologies

that meet their needs best.

Yograj Varma In New

Delhi

Advertisment