Following up my Snow Leopard client article, here’s a rundown of the changes expected in the upcoming server release. We’ve seen that Apple has already pushed the desktop environment to it’s logical conclusion given the current state of the art. Macs are starting to enter the enterprise in larger and larger numbers and it’s the logical moment to start trying to taking some ground in the entrenched server market currently dominated by Microsoft and Linux.
In the current Leopard 10.5 server we’ve seen the first generation of a suite of standards based communications and collaboration tools (the wiki/blog service was completely rewritten in this release so I’m counting it as a v1 product). Note that all of this is my take on things based on publicly available information, as I don’t have an NDA copy to play with. And if I did I’d be under NDA so I wouldn’t be able to talk about it anyway.
Mail Server
Apple already has a very nice IMAP implementation based on open source components (cyrus, postfix). The Snow Leopard objective is to work on the overall performance in order to bring it up to ISP and large enterprise capacities. Other planned tools include server-side email rules and vacation messages. This transforms the current product into an enterprise oriented mail server instead of just a (relatively) basic mail service.
While not mentioned it would appear obvious that push mail is included in this mix using the standard IMAP4 IDLE protocol. This would be directly linked to the ability for an iPhone to integrate perfectly with OS X Server Mail services.
Address Book Server
Another standards based initiative directly targeting the functionality of the Exchange Global Address List. Based on the emerging CardDAV standard, which is in turn largely based on the CalDAV standard, this is a logical extension to the capabilities of OS X Server as a fully featured enterprise server platform.
We already have some basic consolidated centralized address book services built into the Open Directory Service, but by pushing this out using the CardDAV standard we can expect to see push updates and bi-directional live updates for both iPhone and regular desktop clients. Currently live lookups can be done via LDAP (which works very well), but active distribution and updates are currently not available in 10.5.
iCal Server 2
Leopard Server includes a calendaring server based on CalDAV today, but it’s a little quirky and certainly not up to challenging Exchange on it’s home turf. Obviously this needs some more work.
Apple is leveraging its development in the consumer space with MobileMe which is clearly based on the same technologies and brings simplified management of group and shared calendars, push notification and standards based invitations so you can invite people that use other calendaring solutions.
Now there’s a huge difference is putting together the protocol layer pieces and packaging this up for use by your average systems administrator or general end user to building an online service. So between now and release, I’m sure that the primary goal of Apple’s engineers will be putting a simple management front end on these functions so that any Apple user will be able to manage them effectively. On the other hand, by using MobileMe as a testing ground, we can be sure that the implementation will be scalable to just about any size.
Messaging Services and Windows
The trojan horse in this arena is the Windows client for MobileMe. Up until now the free Calendar Synchronization tools for connecting Outlook to CalDAV have been ugly at best and certainly far from robust. There are probably a few commercial offers that work better, but I haven’t tested them. Apple’s MobileMe consumer test bed will let them validate and debug this technology for Windows clients in the wild. Then all that’s left to do is repackage this as the Enterprise Outlook Conduit and let people choose the server they will connect to instead of going directly to the MobileMe service.
The combination of an upgraded Mail server, AddressBook management and iCal Server 2 starts looking an awful lot like a serious Exchange competitor. Server-side rules, well developed secure web interface, integrated shared calendars…
On top of the technology advances there are also a few interesting things that change the competitive landscape :
-
There are no per user licences as opposed to Exchange CALs, so basically you get all of this for the price of the hardware instead of buying the machine and then buying the Exchange licences
-
Everything is standards based, permitting all client platforms to participate fully with no artificial barriers to competition
-
They’ll have tested the technology on an Internet scale deployment with MobileMe
-
The killer piece: Outlook will be able to participate as a full client
The one things I’m hoping to see here is an Exchange Migration Tool. A drop-in Exchange replacement for the cost of the xServe will be hard to argue against. Heck, in many smaller companies an iMac would do the job just fine for running OS X Server.
On a side note, there’s nothing limiting this architecture from integrating with a Blackberry environment which would be a huge coup.
Collaboration
Leopard Server provided businesses with the power of online group collaboration through the use of wikis, blogs, mailing lists, and RSS feeds. The tools are incredibly simple to manage and use and take advantage of the local authentication mechanisms, so building a secured shared workspace is easy and efficient. However the current incarnation doesn’t play nicely with Mobile Safari which is a shame. Snow Leopard Server extends on this collaboration suite with wiki and blog templates optimized for viewing on iPhone (basically reworking the CSS and XSL templates on detection of Mobile Safari), content searching across multiple wikis and attachment viewing in Quick Look for desktop clients.
It also introduces My Page, which gives users one convenient place to access their web applications, receive notifications, and view activity streams. Basically this would appear to be an integrated RSS aggregator similar to Google Reader, but optimized for internal use. What’s not mentioned is whether it will be possible to build and develop applications on top of the blog and wiki platform in order to permit more structured data handling and workflows. This will be something that many Sharepoint folks will be looking for. But from a practical perspective, the basic tools in Leopard work even better and are more accessible than a Sharepoint portal.
The Leopard version of these tools was a complete rewrite from the original blog service based on Blojsom. This is important since it allows them to use the same engine for managing both the blog and wiki content with common RSS feed management, tagging, and file management. One really nice thing about these tools is that they use a very simple pragmatic data structure that doesn’t require that you add in an additional database server, and you can write your own scripts for interrogating the data either via the online APIs like Atom and MetaWebLog, or you can go directly to the filesystem. Aple has an excellent document that explains the structure used for storing the data, including the versioning of modified wiki pages. With minimal effort I was able to write a script that scans the wiki and blog pages to generate a sitemap file used by Google and other search engines.
The new architecture permits Apple to redesign and repurpose this data quickly and relatively easily by adding in XSL and CSS stylesheets.
Remote Access
Leopard already comes with built-in VPN service and IPFW for the firewall. Here it looks like their planning on ensuring that you can easily give access to your remote users with special attention to push based services as well as the current reverse proxy services for secured access to internal services without requiring a VPN. Nothing really new here other than the optimisations necessary for keeping multiple IP connections open for the push based services.
Multicore
This aspect was addressed in the client article, but is even more important on the server side where we see a very high consolidation of services and processes that need to efficiently share multiple cores concurrently.
ZFS
There has been much noise regarding ZFS as the next generation filesystem and it would certainly be a valuable add-on for OS X Server. A dynamically extensible filesystem that allows you to add disks to your server as part of an existing filesystem is a really useful tool. Now you don’t have to worry about partitioning the same way you do today. Assuming that OS X will be able to boot from ZFS, you’ll create one filesystem to start with, and then just add storage as your needs grow for the simplest possible solution to storage. ZFS can efficiently handle disks based on hardware RAID or you can use individual disks or a JBOD and let ZFS handle the redundancy requirements by designating them as RAID 1 or RAID 5 groups. Other pieces that are also useful are things like the ability to snapshot a filesystem so you can roll back to a previous state if an update goes wrong, and both local and remote replication.
Podcast Producer 2
A bit of a niche market, but rapidly becoming more and more present in a lot of enterprises for both internal and customer facing communications. Some of the features appear clearly targeting the education market, like the ability to handle dual-video source capture with both the speaker and the slides for podcasting lectures and other sales presentations.
But as the price of fuel continues to climb and airline tickets get more expensive this kind of on demand presentation/training solution is looking more and more interesting to a variety of markets. Most of these companies don’t have much in the way of internal multi-media resources that know how to prepare this kind of presentation though, so tools that simplify and automate have a real value here in getting new media initiatives off the ground.
64-bit
Duh. The most important message here is going to be with regards to the maximum number of concurrent network connections, especially once we start seeing push services deployed on a large scale.
OpenCL
See the client article
What’s not mentioned
iSCSI: I was disappointed to see this one missing from Leopard and I really want to see it integrated natively in Snow Leopard. More and more storage solutions are offering iSCSI as a viable solution for storage consolidation without the cost of traditional fiber channel SANs. Also, iSCSI plays well with ZFS on both the target and initiator side. Both Microsoft and Linux offer native iSCSI client implementations and Apple is lagging behind here.
Other than that - who knows ? Don’t forget that there are Snow Leopard sessions happening at WWDC that are under NDA. I’d love to be there to hear all about it, but in the meantime, we’ll just have to guess.