Storage Winners and Losers: Hedging Your Bets Page 2


Want the latest storage insights?

Download the authoritative guide: Enterprise Data Storage 2018: Optimizing Your Storage Infrastructure

HBAs – A Case Study

In September 1997 I got to test one of the first Fibre Channel HBAs attached to a FC RAID that was available with a working driver for Solaris. The company that produced the product was called Jaycor at the time and is now known as JNI. They quickly came to dominate the HBA market, but today they command less than 20% of the market, with Emulex and Qlogic controlling over 80% of the market according to the Gartner group.

That raises the critical question of how do you go from complete market domination, which JNI had in the late 1990s (and was still the market leader even as recently as 2000), to a company struggling to maintain a much smaller market share?

There are a number of points that are significant to note in this and similar cases, as such a fall from grace is common throughout our industry today and has existed in the past and will exist in the future.

  1. JNI depended on other companies for its HBA parts. Specifically, JNI used the HP Tachyon Fibre Channel chipset, which meant the following:
    1. HP possessed the ability to produce its own HBAs and compete with JNI (and did)
    2. At the time, JNI depended on HP to move forward with full fabric and other features. This, of course, meant that if HP was late with a feature addition, JNI was as well.
    3. Any issues within the HP fibre chipset also became JNI issues. For example, you could not make I/O requests over 8 MB, as the chipset could not DMA more than that amount

  2. OEM relationships were beginning to be developed by other vendors with major server vendors. SGI teamed with Prisa (and then re-teamed with Qlogic when the company acquired Prisa), Emulex teamed with IBM, and Qlogic teamed with Sun. JNI did get an agreement with EMC and other storage manufacturers, but not with any server vendors. Did this happen because:
    1. As the first vendor to market, JNI had a virtual monopoly, and they figured they did not have to adjust their prices for OEM agreements while newcomers did?
    2. Did JNI believe the best market strategy was to have a direct sales force?

  3. Who was going to use the technology, and with what components? Back in the early days of Fibre Channel, many things did not work together, and certification of RAIDs, Fibre Channel hubs, and early switches was a requirement. Figuring out which storage vendors to work with, what worked, and who needed that storage on what operating system were big issues. If I only have X number of good device driver software engineers, I need to choose where to spend my time given the current and anticipated market directions.

  4. How can you compete with better technology if the standard limits you? From what I have seen, once something becomes a standard, it becomes, by default, a commodity part. Once Fibre Channel standards were released and were working, as a seller the question then became what is our value add compared with the competition? You cannot go faster than the standard, you cannot change the standard, and you cannot be different than the standard. It therefore becomes a race to see who can manufacture the standard commodity product at the lowest cost and keep up with the standards as they are released.

Given all of the industry happenings, what happened to JNI is not at all uncommon in this industry. Qlogic developed its own chipset, which became used in RAID controllers, Fibre Channel switches, and other HBAs as an alternative to HP Tachyon. JNI now develops its own Fibre Channel chipsets as well, but I do not see them in the broad switch and/or HBA market selling their chipsets. HP is now Agilent and is getting out of the HBA business.

What will happen to Emulex and Qlogic, as they are now the two dominant players? I will leave it up to you to do your on analysis. Will we have only one dominate vendor, will we continue to have two, or will we see something different emerge?

Page 3: Conclusions

Submit a Comment


People are discussing this article with 0 comment(s)