Clouds and Chips

The IT world has provided us with a number of interesting developments this week, starting with a Google suit filed over a proposed Department of the Interior messaging system award to Microsoft.  Google feels that its own Apps could have been used for this, and that they should have been given the opportunity to demonstrate their compliance with federal security requirements and bid on the contract.  Thus, the lawsuit.

Some in the DoI have suggested to us that the problem with Apps is the same one that’s a problem for users of Google’s online competitors to Office; the features Google provides are a subset of those already in use rather than the full set.  What’s not totally clear is whether the missing features are actually used at DoI, but in some ways you have to be sympathetic with the department; how easily could they find out whether all or some features were used?  The suit may thus be an important one for cloud services in general.  Many (probably most) cloud-based alternatives to popular installed software tools are functionally more limited than the stuff they’re intended to replace.  That’s also true with most open-source tools.  I’ve tried Google’s document tools and they won’t properly process either our spreadsheets or our presentations, and they create problems with some publication/paper styles as well.  Same for OpenOffice.  But there’s no question that you could do most of what I do in either Apps or OpenOffice if you started from scratch.  So cloud applications would be promoted if they were deemed acceptable if they offered relatively full functionality, even if differently, or if they offered at least some way of doing what buyers actually did rather than what they could do.  Without that kind of ruling, it may be hard to promote the cloud version of many apps unless the cloud providers step up and fully duplicate capabilities.  Frankly, that’s what they should do.  You can’t sue your buyer into submission as a long-term business strategy.

The other interesting development is in the chip space, and the two vendors making the news were Intel and Oracle.  Intel abandoned a long practice of keeping its fab to itself by doing a deal with an FPGA chip specialist Achronix for 22nm capacity.  The actual volume of fabrication here is small, but what may be interesting is that Achronix is perhaps the speed king of FPGAs, which are field-programmable chips that can be used for fast responses to market needs or applications where volumes won’t justify a custom ASIC.  It’s not hard to see that such chips might be very valuable in the consumer device market, which could mean either that Intel may want a stake in Achronix later on, or that it may itself be thinking about getting into the consumer space on a larger scale.  Recall that Intel has its own mobile OS and that it’s often been said to have aspirations of being a player in a retail space.  What better one than devices?

Oracle’s move is if anything even more interesting; they’ve taken a stake in Mellanox, who is one of the key providers of chips for InfiniBand data center switches.  They’ve been a partner with Sun and also provide stuff for Oracle’s storage appliances, but as we’ve noted before, Oracle is the only data center player with no position in networking, and nowhere is that position more important than in the data center.  InfiniBand is a superior technology to at least the current generation of Ethernet in terms of latency and capacity, and were Oracle to be planning to do a big flat fabric for the data center, Mellanox would be a likely player in their decision.  It’s also interesting to note that the deal includes Mellanox supporting Solaris as one of its host OSs.  That suggests that Oracle may be planning to continue to field Solaris as an alternative to Linux.  We think that’s smart; Solaris has a good following, and for specialty applications like OLTP we think it’s the best OS out there.  Could Oracle be planning a major data center move?  It certainly could be.

Leave a Reply