RHEL5:即将上演（Scott专访）

　我最近有机会采访到Scott Crenshaw，作为红帽公司的RHEL产品管理和市场部门的高级主管，我为此感到非常荣幸。我和他谈到了红帽公司即将在2007年年初发布的RHEL5，关于该产品的性能，尤其是和竞争对手Novell，还有就是最近冒出来的、气势汹汹的Oracle。　　我们谈话的重点是在虚拟化方面，比如红帽公司如何看待虚拟化技术，以及如何部署、如何向市场推广。
　　当被问及最近Oralce公司宣布Linux支持方案时候，Scott Crenshaw先生的回答使得我印象深刻。他说：“其实Oracle的举动也很平常，谁都可以这么坐，但是坐的怎么样就不知道了。红帽最关心的是客户，而不是对手，Oracle宣布进入Linux市场确实造成了不小的恐慌，但是很快，红帽公司就回到了它原来的轨道。”
　　他还举例说，对客户的关注渗透到了红帽公司工作的每个方面。就像最新的RHEL5一样，完全是为用户度身定制的，比如在虚拟技术和方案方面，由用户来选择。同样，在其他的功能要求上，我们也提供同样的可选择的服务，一切由用户决定，我们负责提供服务，公司的IT架构由客户来打造。
　　更多的内容阅读：
　　I was fortunate to do a Q&A session today with Scott Crenshaw, Senior Director of Product Management and Marketing for the Red Hat Enterprise Linux product. We talked about a range of things related to the early 2007 release of RHEL 5: product features, competition with Oracle and Novell, and other things.
　　We spent the most time, however, talking about Red Hat's views on and plans for virtualization and how Red Hat gets product to market.
　　On the latter topic, I was most impressed with something Scott told me when I asked how the Oracle announcement had affected Red Hat:
　　This will sound trite, because every vendor says it, but Red Hat is focused on customers, not competitors. Oracle's announcement shook things up around here for a day or so, and then we got back to work.
　　This focus on the customer permeates everything we do. For example, RHEL 5 was designed by customers, for customers. The customers decided on the right virtualization approach. They decided on the other features, as well. Red Hat ships it, but customers build the roadmap.
　　It would sound trite but for the fact that I've worked with (or against) Red Hat for six years now. Its focus on the customer can at times be frustrating (like if you're trying to get them to speak at your conference... :-), but it's a tangible, driving force within the company. I honestly have not experienced the same level of intensity on customer issues than I have with any other company.
　　It is the reason, for example, that Red Hat has yet to ship its GA release of the Xen virtualization server. As Scott noted,
　　The Red Hat model is first built around product requirements, then quality requirements, and then the schedule. We don't ship anything until it's stable, secure, and high performance.
　　With this in mind, I asked Scott to comment on RHEL 5 generally. His response (keep in mind that I was typing as fast as I could, but may not get his comments exactly right. Any errors are my own.):
　　Red Hat is providing the future of how enterprises consume software, starting with virtualization. Virtualization is at the heart of this vision, which is why the main driver for RHEL 5 is virtualization. Vendors talk about server consolidation as a benefit of virtualization, and that's true. It is. But Red Hat believes virtualization's most important benefits are availability and flexibility. Open source virtualization democratizes IT, allowing enterprises to better utilize spare capacity, already inherent in their systems, to provide increased availability to applications under resource crunches.
　　All fine and good (btw, Red Hat gives other benefits on its site), but why open source virtualization? Scott's reply:
　　Most of the use cases that make virtualization worthwhile are too expensive with proprietary software. Wall Street CIOs tell us that at $6,000 - $9,000/server, as is the case with VMware, the economics don't work for virtualization. All you can do at that point is consolidate servers to replace multiple machines with one big, expensive machine.
　　Open source virtualization makes the economics of virtualization work, but also gives the enterprise access to a rich development ecosystem, as with the Linux ecosystem.
　　I wanted to know more about the pricing for Red Hat and virtualization, and pushed a little harder. Scott informed me that it's important to recognize that memory, I/O, and processor capacity mitigate against people running thousands of instances of an OS on a machine. They technically can do it, but not practically. With this as context, he indicated that the RHEL 5 comes with virtualization and a certain number of guest instances included for free. It sounds like RHEL 5 customers will get enough included in the base price to get up and running, and will pay for more advanced implementations.
　　Consistent with Red Hat's emphasis on value, however, Red Hat's focus is not on price, but on customer experience. For that reason, Scott said that
　　the RHEL Virtualization Platform goes far beyond providing the giblets. The intent is to give users an out of the box experience with virtualization with the primary use cases ready to go out of the box. For example, the platform comes witha fully integrated storage solution (because the ability to move between machines, if you don't have storage persistence, doesn't do you much good). Customers can choose to roll their own or to take RHAT's prefab'd, integrated virtualization.
　　With all this as context, I asked Scott to talk about a few of Red Hat's competitors, starting with
　　Novell
　　Novell's approach is to provide technology components as early as available, because they want to appear to be the innovators. Red Hat does the same thing through Fedora. With our RHEL product, however, our concern is with providing a rock solid platform. The early Xen code base had performance, stability, and security weaknesses, so RHAT chose to wait. RHAT's schedule is driven by quality concerns. We won't ship something until we believe it will deliver real customer value.
　　Microsoft
　　No idea. Microsoft's virtualization is just a press release at this point.
　　VMware
　　Red Hat works closely with VMware to provide support and software to our mutual customers. But RHAT believes strongly in the open source innovation model and feels storngly that model will win over time. Closed source can't compete with the open source innovation machine, nor its cost.
　　And then, in a roundabout way, I asked about Oracle. I asked if Red Hat invested engineers in the Xen project in the same way it does with the Linux kernel. He responded:
　　Red Hat has 30-40 of its best engineers working on the Xen project, helping to drive its innovation. The original Xen team did a great job creating the foundation of the project. Red Hat and others (IBM, Intel, etc.) then came together to help push Xen to the next level. Memory management, scheduler, IO, changes in the OS to support virtualization, etc.: Red Hat has been instrumental in all of these fundamental changes.
　　And Oracle? How many developers does Oracle have working on the Xen project?
　　Oracle has approximately zero developers on Xen.
　　For this reason, (as well as others noted by Paul Cormier), it's very, very hard to out-support and outperform Red Hat on its own turf, much less with a muddied mixed source message.) It will be very hard for Oracle (or anyone else) to support software that it doesn't play an active role in developing.
　　Anyway, that's the meat of the discussion Scott and I had. In addition to virtualization, of course, there will be a range of other innovations in RHEL 5, among them:
　　* SELinux MLS support (EAL4+/LSPP)
　　* SELinux Troubleshooter - Greatly simplifies administration
　　* Network storage ( Autofs, CacheFS / NFS persistent local cache, iSCSI )
　　* Integrated directory & security
　　* Desktop (GNOME, X.Org 7.1, Laptop)
　　* Stateless Linux (Desktop/Server/ Virtualized )
　　* New Driver Model (better support for 3rd party drivers)
　　* Enhanced development tools (SystemTap, Frysk)
　　* Enhanced large SMP support (big kernel lock removal, etc.)
　　* Multi-Core beyond Dual
　　* Single Node GFS
　　* Kexec / Kdump ( replacing Diskdump and Netdump)
　　* Installer improvements
　　* RHN support for virtualization
　　* Broad range of new HW support
　　* IPv6 support and conformance enhancements
　　* IPSEC enhancements
　　* I/O-AT &＃65533; Intel's network accelerators
　　* Improved ACPI support, suspend to disk
　　* Block device data encryption support
　　* Root device MPIO support
　　* Dynamically switchable per-queue I/O schedulers
　　* Enhanced pipe buffering (circular buffers)
　　* IPv4/IPv6 fragmentation offload & optimized buffer management
　　* Integrated multi-media support
　　* Smart card login - with PKI/Kerberos authentication
　　* Enhanced plug and play hardware support (cameras, etc)
　　* Enhanced graphics using AIGLX/Compiz (e.g. fading, transparency, etc)
　　* Network Manager - automatic network configuration
　　* Samba - improved Microsoft Active Directory integration
　　* Audit - powerful new search/reporting tools; unique real time interface
　　Looks like a good time to be buying Linux.