Current & Future Linux on Sysetm z Technology Deep Dive

A presentation at NSA Open Source Software User Group in January 2010 in Fort Meade, MD, USA by Shawn Wells

Slide 1

Slide 1

Current & Future Linux on System z Technology Deep Dive NSA Open Source User Group Meeting Friday 29-JAN-2010 UNCLASSIFIED

Slide 2

Slide 2

Introductions ● ● Shawn Wells (sdw@redhat.com, 443-534-0130) U.S. Intelligence Community Sales Manager – Former NSA-er (S31121, Data Management) – Former Tech Director, led Red Hat’s Global System z Business Unit until 1-DEC-2010 – RHCE Jim Stann (jstann@redhat.com, 703-703-626-8580 Senior Solutions Architect – Formally SUN from 1996-2009 – Moved to Red Hat as a technical pre-sales engineer – RHCE

Slide 3

Slide 3

Agenda ● ● ● Current technology through RHEL 5.4 relating to both x86 and s390x platforms – RHEV / KVM – System z Hardware Enablement – Performance Numbers Roadmap moving forward (RHEL 5.5, RHEL6) – Ext4 – Kernel patches & improvements – Performance Open Dialog

Slide 4

Slide 4

Development Model

Slide 5

Slide 5

Development Model

Slide 6

Slide 6

Development Model

Slide 7

Slide 7

Red Hat Confidential - NDA required Release Roadmap Highlights: RHEL 5 ■ RHEL 5.3 released January 20, 2009 ● Virtualization ● ● ● ● ● Improved scalability (Xen-based) ● 126 sockets & 1TB memory per host ● > 4 nics and > 16 disks per guest ● 2MB page table support ● AMD RVI support Virtio drivers to run 5.3 on KVM hosts libvirt-cim CIM/WBEM provider Timer improvements FasTr ack Platform enablement ● ● ● ● ● ● ● ● ● Nehalam enablement and testing OFED 1.3 SystemTap updates GDB debugger refresh FCoE enablement for specific configurations Broad driver refresh GFS2 fully supported Ext4 TechPreview DHCPv6 Support ■ FasTrack 7% ● ■ Hardware Enablement - 7% ● ■ New chipsets and processor feature support New Features - 21% ● ■ Early release of low impact fixes Feature requests from customers and partners Other - 65% ● ● Bugfixes Documentation 7

Slide 8

Slide 8

Red Hat Confidential - NDA required Release Roadmap Highlights: RHEL 5 contd. ■ RHEL 5.4 released August 2009 ● Virtualization technology refresh ● ● ● ● ● ● Platform enablement ● ● ● ● ● ● ● ■ Addition of KVM in parallel to Xen VT-D / IOMMU support for KVM & Xen 32 pv-guest on 64 bit hosts supported with Xen Improved timer support for FV guests (kvm & VMWare) Libvirt refresh RAS, EDAC OFED 1.4 Hugepage core dump support Storage improvements (fcoe-utils, IO accounting in iostat, iSCSI offload, driver updates, XFS) Network improvements (IO/AT, IPv6, drivers) IPMI update SystemTap updates RHEL 5.5 planned for early Q2 2010 ● ● ● Regular minor release with hardware enablement, reduced content compared to 5.4 Intel Boxboro-EX platform, AMD Magny-Cours processor PCIe AER support 8

Slide 9

Slide 9

Red Hat Confidential - NDA required Release Roadmap Highlights: RHEL 6 ■ ■ Planned for early 2010, development currently in Fedora 12 Planning focus ● ● Virtualization – making RHEL an optimized host & guest Performance & scalability optimized ● ● ● ● Manageability ● ● ● ● ● Integration with RHEV Management AMQP messaging base (for use in virt, MRG) IPA enablers – for audit, smartcard, desktop integration, MRG & virt – enabling centralized administration Open standard management enablers Power Management ● ● ● Qumranet thin client enablers Resource containment, guest resource isolation & security containment Device optimization For virt, bare metal, laptop Hardware level as well as dynamic system service startup and suspend Scalability ● Larger max CPU, mem, IRQs, filesystem – both virt & bare metal ● RAS (Hotplug, data integrity) ● Hardware Enablement ● ● UEFI – new bios boot loader interface, when legacy mode goes away Currency – leveraging RH’s & community innovation post-RHEL5 codebase 9

Slide 10

Slide 10

RHEL 5.3 Current Technology

Slide 11

Slide 11

RHEL 5.3: Overview GA on January 20, 2009  ~150 additions, ~3,400 BugZillas  FasTrack 7%  Early release of low impact fixes  Hardware Enablement - 7%  New chipsets and processor feature support  New Features - 21%  Feature requests from customers and partners  Other - 65%  Bugfixes  Documentation  FasTrack Features Hardware Other

Slide 12

Slide 12

RHEL 5.3: Networking  Provision of several selectable TCP congestion modules (2.6.13)  Ref: http://lwn.net/Articles/128681/  IPV6 - Support several new sockopt / ancillary data in Advanced API (2.6.14)  IPv4/IPv6: UFO (UDP Fragmentation Offload) (2.6.15)  Offloads IP fragmentation functionality of large UDP datagram to hardware  Improves performance  Add nf_conntrack subsystem: (2.6.15)  Common IPv4/IPv6 generic connection tracking subsystem  Allows IPv6 to have a stateful firewall capability (not previously possible)  Increased security  Enables analysis of whole streams of packets, rather than only checking the headers of individual packets

Slide 13

Slide 13

RHEL 5.3: Networking  IPv6  RFC 3484 compliant source address selection (2.6.15)  Add support for Router Preference (RFC4191) (2.6.17)  Add Router Reachability Probing (RFC4191) (2.6.17)  Generic segmentation offload (GSO) (2.6.18)  Available in place of TSO (TCP Segmentation Offload)  Performance improvements for large packet transfers without hardware assistance  SELinux per-packet access controls  Replaces old packet controls  Add Secmark support to core networking  Allows security subsystems to place security markings on network packets (2.6.18)  Inclusion of DCCPv6 – Datagram Congestion Control Protocol (2.6.16)

Slide 14

Slide 14

RHEL 5.3: Storage Management  RAID 4/5/10 support added to dm-raid.  Full support for software iSCSI target.  Full support for LVM cluster mirror (cmirror).  Add the ability to prioritize paths on HP MSA/HSV active/passive storage controllers.  Reduce boot time by improving lvmcache, to reduce the amount of device scanning.  Enhanced disk partition statistics

Slide 15

Slide 15

RHEL 5.3: File System / Storage Mgmt  Block device encryption support, including support for /root partition, including configuration in anaconda installer.  ext4 tech preview  samba: rebased from 3.0.28 to 3.0.32 for bugfixes  Now supports Windows Vista and 2008  fixes for DC functionality (interoperability with Citrix and Domain trusts)  Ecryptfs fixes (tech preview)

Slide 16

Slide 16

RHEL 5.3: System Services  Rebased version of CUPS print server, now fully Kerberized  dhcpv6 support  ktune, a service that sets several kernel tuning parameters to values suitable for specific system profiles. Currently, ktune provides a profile for large-memory systems running disk-intensive and network-intensive applications. New package, tech preview.  Package upstream rebases to the following utilities:  ksh, lm-sensors, lftp, net-snmp, openIPMI-tool, openldap, openmotif, pythonurlgrabber, openPegasus, VNC  RPM to Fedora 9 version, which includes numerous bugfixes  yum and yum-utils primarily for speed improvements  totem, rb, and gstreamer rebased to enable modular codecs addition  Numerous wireshark security fixes

Slide 17

Slide 17

RHEL 5.3: Security Enhancements  pkinit clients can now be configured to use keys for client certificates which may not contain Kerberos-specific extensions & interoperability fixes  nss_ldap now configured with support for paged results extension  SELinux: enablement of New NetworkManager and Audit functionality.  SELinux: Hundreds of AVC denial fixes.  Improved Audit and Logging  TTY input audit support  Remote audit logging via unencrypted connection

Slide 18

Slide 18

RHEL 5.3: System z Specifics BugZilla ID Summary 46327 stage1: sshd error loading shared lib: libfipscheck.so.1 184770 LTC18425-62140: (big) xDR system Initialization for LPAR Clients 472788 rhel 5.3 snapshot3 scsi mpath install failed on z9bc lpar 439479 LTC:5.3:201474:Include gcc 4.3 as Add-On for latest z10 instruction set support 439440 LTC:5.3:201160:Long Random Numbers Generation 439441 LTC:5.3:201158:Selective Logging of ECKD DASD devices 439482 LTC:5.3:201542:FCP - Enhanced Trace Facility 447379 LTC:5.3:200994:Linux CPU Node Affinity 463917 unable to find DASD drives to install 439484 LTC:5.3:201490:Libica Library: Integration of Icainfo 43946 LTC:5.3:201360:OSA 2 Ports per CHPID Support - Installer Enhancements 466474 [RHEL5.3] *** glibc detected *** /usr/bin/python: double free or corruption (!prev): 0x000 0000080d55e90 *** 466305 cosmetic error message: failure in nl_set_device_mtu 466291 anaconda silently omits uninitialized disk

Slide 19

Slide 19

RHE 5.4

Slide 20

Slide 20

RHEL 5.4 : Overview  The official GA release of Red Hat Enterprise Linux 5.4 (kernel-2.6.18-164.el5) was released on 02-Sep-2009  Generic (not s390 specific) updates include  Virtualization: full support for the Kernel-based Virtual Machine (KVM) hypervisor only on x86_64 while Xen only x86 & Power based virtualization is still available & supported  Network: Kernel & Userspace update to support Generic Receive Offload (GRO) which increases the performance of inbound network connections by reducing the amount of processing done by the CPU. Furthermore Netfilter Framework & Bind Updates  Storage: Support for the XFS file system has also been added to the kernel as a Technology Preview.  Tools: SystemTap is now fully supported, and has been re-based to the latest upstream version.

Slide 21

Slide 21

RHEL 5.4: File System / Storage Mgmt  Add integrity check to cryptsetup-luks, in order to meet FIPS-140 requirements.  Ext4 - refreshed the backport for our tech preview to bring in bug fixes and support for delayed allocation.  File system freeze/quiesce interface added to support hardware snapshots for file systems.  Full support for FUSE and libfuse to allow end users to more easily install and use their own user space FUSE file systems.

Slide 22

Slide 22

RHEL 5.4: System z Specifics BugZilla ID Summary 475556 [LTC 5.4 FEAT] DS8000 support: Large volume support (userspace) [201738] 475569 [LTC 5.4 FEAT] Shutdown actions tools [201755] 461288 [EMC 5.4 feat] Require kernel support to issue Control I/O to CKD dasd on EMC Symmetrix arrays 474688 [LTC 5.4 FEAT] Automatic IPL after dump (kernel) [201169] 475346 [LTC 5.4 FEAT] Improve checking mechanisms and workflow of Linux on System z Anaconda install process [201676] 475670 [LTC 5.4 FEAT] Program directed IPL support - no XML in system dumper [200782] 475552 [LTC 5.4 FEAT] FCP - Performance data reports [201730] 488496 [LTC 5.4 FEAT] 201173:Crypto Hardware Enablement Device Driver Support - toleration 475564 [LTC 5.4 FEAT] Shutdown actions interface (userspace) [201748] 474679 [LTC 5.4 FEAT] Dynamic CPU hotplug daemon for System z [201132] 475345 [LTC 5.4 FEAT] Change list of Anaconda network alternatives to indicate supported devices on System z [201679] 475551 [LTC 5.4 FEAT] TTY terminal server over IUCV (kernel) [201734]

Slide 23

Slide 23

475563 [LTC 5.4 FEAT] Shutdown actions interface (kernel) [201747] 474664 [LTC 5.4 FEAT] System z support for processor degradation BugZilla ID Summary [200975] 475556 [LTC 5.4 FEAT] DS8000 support: Large volume support (userspace) [201738] 475569 475334 [LTC 5.4 FEAT] FCP - Performance Data collection (kernel) [LTC 5.4 FEAT] Shutdown actions tools [201755] [201590] 461288 feat] Require kernel support to issue Control I/O to dasd on EMC Symmetrix 475572 [EMC [LTC5.4 5.4 FEAT] HiperSockets Layer3 support forCKD IPv6 [201751] arrays 474688 475346 475670 475552 488496 475548 [LTC 5.4 FEAT] FCP - Performance data collection (blktrace) [LTC 5.4 FEAT] Automatic IPL after dump (kernel) [201169] [201729] 5.4 FEAT] Improve checking mechanisms and workflow of Linux on System z Anaconda 477189 [LTC [LTC 5.4 FEAT] Pick up latest version of s390-tools install process [201676] 475558 [LTC 5.4 FEAT] TTY terminal server over IUCV (userspace) [LTC 5.4 FEAT] Program directed IPL support - no XML in system dumper [200782] [201735] [LTC 5.4 FEAT] FCP - Performance data reports [201730] 474646 [LTC 5.4 FEAT] Kernel NSS support - kernel part [200790] [LTC 5.4 FEAT] 201173:Crypto Hardware Enablement Device Driver Support - toleration 475564 475333 [LTC 5.4 FEAT] FCP - Performance Data collection & analysis [LTC 5.4 FEAT] Shutdown (userspace) [201591]actions interface (userspace) [201748] 474679 FEAT] Dynamic CPUimage hotplugdump daemonon forDASD System [201752] z [201132] 475571 [LTC [LTC5.45.4 FEAT] Large 475345 [LTC 5.4 FEAT] Change list of Anaconda network alternatives to indicate supported devices on System z [201679] 475551 [LTC 5.4 FEAT] TTY terminal server over IUCV (kernel) [201734]

Slide 24

Slide 24

475530 [LTC 5.4 FEAT] Extra kernel parameter via VMPARM [201726] 475557 [LTC 5.4 FEAT] DS8000 Disk Encryption [201740] 475563 [LTC 5.4 FEAT] Shutdown actions interface (kernel) [201747] 474942 [LTC 5.4 FEAT] Add vmconvert option to vmur tool [201758] 474664 [LTC 5.4 FEAT] System z support for processor degradation BugZilla ID 475570 Summary [LTC 5.4 FEAT] Provide service levels of HW & Hypervisor in [200975] LinuxDS8000 [201753] 475556 [LTC 5.4 FEAT] support: Large volume support (userspace) [201738] 475334 [LTC 5.4 FEAT] FCP - Performance Data collection (kernel) 468172 FEAT: 201085: cio_ignore entry in generic.prm for LPARs 475569 [LTC 5.4 FEAT] Shutdown actions tools [201755] [201590] [LTC 5.4 FEAT] Device Driver of Thin Interrupts 461288 475572474700 [EMC feat] Require kernel Crypto support to issue Control I/Ouse to dasd on EMC Symmetrix [LTC5.4 5.4 FEAT] HiperSockets Layer3 support forCKD IPv6 [201751] [201174] arrays 475548 [LTC 5.4 FEAT] FCP - Performance data collection (blktrace) 475350 [LTC 5.4 FEAT] Dialog defaults for Linux on System z specific 474688 [LTC 5.4 FEAT] [201729] Automatic IPL after dump (kernel) [201169] Anaconda [201677] 475346 [LTC 5.4 FEAT] Improve checking mechanisms and workflow of Linux on System z Anaconda 477189 install [LTC 5.4 FEAT] Pick up latest version of s390-tools 475820process [LTC[201676] 5.4 FEAT] Linux to add Call Home data [201167] 475558 [LTC 5.4 FEAT] TTY terminal server over IUCV (userspace) 475670 [LTC 5.4 FEAT] directed IPL support - no XML dump in system dumper [200782] 484296 [LTC Program 5.4 FEAT] Automatic IPL after (userspace) [201757] [201735] 475552 488496 [LTC 5.4 FEAT] FCP - Performance data reports [201730] 474646 [LTC 5.4 FEAT] Kernel NSS support - kernel part [200790] [LTC 5.4 FEAT] 201173:Crypto Hardware Enablement Device Driver Support - toleration 475564 475333 [LTC 5.4 FEAT] FCP - Performance Data collection & analysis [LTC 5.4 FEAT] Shutdown (userspace) [201591]actions interface (userspace) [201748] 474679 FEAT] Dynamic CPUimage hotplugdump daemonon forDASD System [201752] z [201132] 475571 [LTC [LTC5.45.4 FEAT] Large 475345 [LTC 5.4 FEAT] Change list of Anaconda network alternatives to indicate supported devices on System z [201679] 475551 [LTC 5.4 FEAT] TTY terminal server over IUCV (kernel) [201734]

Slide 25

Slide 25

S390-tools package rebased to Version 1.8.1 The s390utils package has been rebased to version 1.8.1.  This package provides the essential tool chain for Linux on System z. It contains everything from the boot loader to dump related tools for a system crash analysis .   News Features (excerpt)  DASD related tools: Add Large Volume Support for ECKD DASDs Ipl_tools: Can be used to change the reipl & shutdown behaviour  ziomon tools: Set of tools to collect data for zfcp performance analysis.  lsluns: List available SCSI LUNs depending on adapter or port.  lszcrypt: Show information about zcrypt devices and configuration.  chzcrypt: Modify zcrypt configuration.  cpuplugd: Daemon that manages CPU- and memory-resources based on a set of rules. Depending on the workload CPUs can be enabled or disabled. The amount of memory can be increased or decreased exploiting the Cooperative Memory Management (CMM1) feature.  chchp: Tool to modify channel-path states  lschp: Tool to list information about available channel-paths.  mon_procd: Daemon that writes process information data to the z/VM monitor stream.  vmur: Tool to work with z/VM spool file queues (reader, punch, printer).  zfcpdump_v2: Version 2 of the zfcpdump tool. Now based on the upstream Linux kernel 2.6.23.  Plus various bug fixes 

Slide 26

Slide 26

Kernel Control Program Identification (CPI)  If your RHEL5.4 Linux instance runs in LPAR mode, you can now use the extended control program identification (CPI) module, sclp_cpi and the sysfs interface /sys/firmware/cpi to assign names to your Linux instance  The names are used, for example, to identify the Linux instance on the HMC.  This feature is only available while running in LPAR  Extra kernel parameter via VMPARM  Modify the IPL records to append extra parameters specified with the z/VM VMPARM option to the kernel command line.  Support for processor degradation  Adds support for processor degradation, which allows processor speed to be reduced in some circumstances (i.e. system overheating). This new feature allows automation software to observe the machine state.  TTY terminal server over IUCV  Provide central access to the Linux console for the different guests of a z/VM.  The terminal server connects to the different guests over IUCV.  The IUCV based console is ASCII based.  Fullscreen applications like vi are usable on the console. 

Slide 27

Slide 27

Virtual Server  Named Saved Segments (NSS)  Using NSS the z/VM hypervisior makes operating system code in shared real memory pages available to z/VM guest virtual machines.  With this update, Linux guest operating systems using z/VM can boot from the NSS and be run from a single copy of the Linux kernel in memory. z/VM Expanded Storage Linux A mem=xxxx Shared Memory vi bash Virtual Disk Virtual Disk Linux Guest Linux Guest Linux A Guest Storage z/VM Central Storage Linux B Guest Storage Linux NSS z/VM Page Space 0 MB Linux B mem=xxxx

Slide 28

Slide 28

Networking  HiperSockets Layer3 Support for IPv6  How IPv6 support for HiperSockets devices running in layer 3 mode is available  IPv6 is supported on:  Ethernet interfaces of the OSA-Express adapter running in QDIO mode.  HiperSockets layer 2 and layer 3 interfaces  z/VM guest LAN interfaces running in QDIO mode.  IPv6 is not supported on the OSA-Express Token Ring and ATM features.

Slide 29

Slide 29

RAS      Multi volume dump support for DASDs  Added the ability to dump on multiple ECKD DASD devices, which can be necessary, if the system memory size is larger than the size of a single DASD device. Service Levels of Hardware & Hypervisor  A new Interface which provides service levels of hardware and z/VM service-levels to the Linux userspace. Interface: /proc/service_levels Lstape support for SCSI Tapes  With this feature it is now possible to list installed FCP-attached tape devices (SCSI tapes) besides channel attached tapes using the lstape command Shutdown Actions Interface  The new shutdown actions interface allows to specify for each shutdown trigger (halt, power off, reboot, panic) one of the five available shutdown actions (stop, ipl, reipl, dump, vmcmd).  A sysfs interface under /sys/firmware is provided for that purpose.  Possible use cases are e.g. to specify that a vmdump should be automatically triggered in case of a kernel panic or the z/VM logoff command should be executed on halt. Automatic IPL after dump  The new shutdown action dump_reipl is introduced. It combines the actions dump and re-ipl, first a dump is taken, then a re-ipl of the system is triggered

Slide 30

Slide 30

Storage FCP performance data collection & reports:  Fibre Channel Protocol (FCP) performance data can now be measured. Metrics that are collected and reported on include:  Performance relevant data on stack components such as Linux devices, Small Computer System Interface (SCSI) Logical Unit Numbers (LUNs) and Host Bus Adapter (HBA) storage controller information.  Per stack component: current values of relevant measurements such as throughput, utilization and other applicable measurements.  Statistical aggregations (minimum, maximum, averages and histogram) of data associated with I/O requests including size, latency per component and totals.  DS8K Encryption Support  This feature enhances s390-tools to be able to display if the Storage has its disk encrypted or not.  Kernel support to issue Control I/O to dasd on EMC Symmetrix arrays  Support has been added to the kernel to issue EMC Symmetrix Control I/O. This update provides the ability to manage EMC Symmetrix storage arrays. 

Slide 31

Slide 31

Future Linux on System z Technology

Slide 32

Slide 32

Advanced Virtualization Dynamic Memory Add/Remove (kernel 2.6.27)  Enable to attach and use standby memory that is configured for a logical partition or z/VM guest.  Memory Attach & Detach requires running Linux on System z as a VM-guest requires z/VM 5.4 plus the PTF for APAR VM64524.  Standby CPU activation/deactivation (kernel 2.6.25)  Allow standby CPUs to be activated / deactivated  Suspend / Resume (kernel 2.6.31)  With suspend and resume support, you can stop a running Linux on System z instance and later continue operations.  When Linux is suspended, data is written to a swap partition. The resume process uses this data to make Linux continue from where it left off when it was suspended.  A suspended Linux instance does not require memory or processor cycles. 

Slide 33

Slide 33

Storage Support  HyperPav (kernel 2.6.25)  HyperPav is addressing the need to access more data with good performance and high availability!  This feature, which required a IBM DS8000™ disk storage system in average leads to a higher utilization, resulting in I/O transfer rates.  Activated automatically when the necessary prerequisites are there (DS8000 with HyperPAV LIC, z/VM 5.3). Transparent for the Linux on System z guest  DASD Large Volume Support (> kernel 2.6.29)  Large Volume Support is a feature that allows to use ECKD devices with more than 65520 cylinders. This features is available with DS8000 R4.0  High Performance FICON (HPF) (kernel 2.6.29)  Added HPF support to the DASD Device Driver  HPF is an extension to the FICON architecture and is designed to improve the execution of small block I/O requests.  HPF streamlines the FICON architecture and reduces the overhead on the channel processors, control unit ports, switch ports, and links by improving the way channel programs are written and processed.

Slide 34

Slide 34

Usability & Serviceability  Automatic IPL After Dump (kernel 2.6.30)  Extension to the shutdown action interface which combines the actions dump and reipl, first a dump is taken, then a re-ipl of the system is triggered  Compiler Improvements (gcc 4.3/4.4) – The latest compiler enhancements allow a customer to recompile existing applications which can be optimized for the latest hardware generation without any changes to the source code. – This can lead up to a > 10 % performance improvement.  Large Page Support (kernel 2.6.25) – Support for a new access method to allocate larger chunks of memory, resulting in performance improvements, especially in Java based environments – This feature exploits z10 hardware features and provides a software emulation for older systems.

Slide 35

Slide 35

Miscellaneous  STP/ETR Support (kernel 2.6.27)  Support for clock synchronization using the server time protocol (STP) or an external time reference (ETR).  Kernel vdso support (kernel 2.6.29)  Kernel provided shared library to speed up a few system calls (gettimeofday, clock_getres, clock_gettime)

Slide 36

Slide 36

Red Hat Confidential - NDA required Filesystems ■ ■ ■ Ext4 – main, default filesystem. Incremental scalability (ie 100TB files) & perf enhancements over ext3. SSD device tuning & testing NFS ● ● ● ■ ■ Fuse kernel portion as enabler. GFS2 ● ● ● ■ NFS4.0 – as default NFS4.1 – referrals, delegation & failover. Secure NFS, selinux labels over the wire – enabler for secure virt – timeframe likely post-GA update Targeting HA clusters of up to 16 nodes. Primary incremental focus is stabilization and performance Possible improvements in ability to shrink/defragmentation BTRFS – next generation, enhanced data integrity, ease of use, & scalability. Unlikely to be production ready for RHEL6GA. Likely tech preview, non root/boot. 36

Slide 37

Slide 37

Red Hat Confidential - NDA required Power Management ■ Kernel ● ● ● ● ■ System services / daemons ● ● ● ● ■ Tickless kernel – fewer interrupts, more idle time to drop to lower power states ASPM – (Active State Power Management) – PCI Express reduced power states on inactivity ALPM (Aggressive Link Power Management) – SATA links in low power mode when no I/O pending Kernel enhancements tracking hardware capabilities for deeper low-power states and virtualizatized guest lower power states Intelligent drive spin down Reduced wakeup count (audit / profile package set and adapt) New system startup infrastructure, upstart, facilitates dynamic service initiation and suspension. Have more services start on demand. Dynamically power down unused peripherals when idle – ie, network, bluetooth Power monitoring tools ● ● GUI tools for tuning & monitoring power policy APIs for power monitoring and policy configuration 37

Slide 38

Slide 38

Red Hat Confidential - NDA required Kernel Scalibility ■ Scalability limits – Maximum values ● ● ● ● ● ● ■ Max CPUs – dynamic allocation of CPU structs (2.6.29) allows supported limit of 4096 and theoretical limit of 64K. Max IRQs – dynamic allocation (2.6.28) allows limit of 256 Max memory 48-bit addressing (256TB) requires pending patch incorporation Max # processes – 4 million on 64 bit kernels Max threads per process remains at 32000 (same as RHEL5) Filesystem limits for ext4 – 100T (practical target – bounded by fsck time) Scalability features ● ● ● Split LRU VM – different eviction policies for file backed vs swap backed Ticket spinlocks fixes NUMA starvation CFS scheduler – better NUMA balancing 38

Slide 39

Slide 39

Red Hat Confidential - NDA required IPA Client (Identity, Policy, Audit) ■ IPA client in core RHEL for centralized security management ● ● ● ● Kerberos authentication with host based access control Provides central storage of extended user attributes Enables centralized policy for applications, including SELinux policy Audit log aggregation services & search capabilities 39

Slide 40

Slide 40

Red Hat Confidential - NDA required Security enhancements ■ Virtualization isolation in conjunction with SELinux ● ● ● ● ■ NSS crypto ● ● ● ● ■ Login, Firefox, Thunderbird port RHEL5 smartcard enablers. Volume encryption ● ■ Broaden the core services which utilize NSS crypto libraries Allows cheaper implementation of new features, ie TPM & centralized key store Incremental targeted conversion of: Openswan, openldap, glibc Add new crypto GUI for key import & establishment of trust Smartcard integration ● ■ May be post-GA update Labeled NFS for filesystem isolation Cut-and-paste window controls per authorization level Guest confinement via SELinux policy enhancements Basic operation already present in RHEL5 – incremental centralized key management for RHEL6 Sectool – compliance checking / intrusion detection utility – validates system admin config, ie file permissions, valid UIDs, reasonable passwords, etc 40

Slide 41

Slide 41

Red Hat Confidential - NDA required Desktop – product themes ■ ■ General Desktop technology update Large scale desktop deployment: ● ● ● ● Power management Mobility Hw enablement Monitors / docking / projectors 41

Slide 42

Slide 42

Q&A

Slide 43

Slide 43

Red Hat Enterprise Linux Performance Update Industry Benchmark Performance Records on Nehalem Aug 31, 2009 Product Marketing, Platform BU, Red Hat

Slide 44

Slide 44

Red Hat Confidential - NDA required RHEL on NEHALEM means … Best Performance on web apps  Highest SPECWeb2005 result ever  Performance: Top 11 SPECWeb(c)2005 results based on Red Hat Enterprise Linux  Six new 2-socket HPC benchmarks  Manufacturing Industry: 4 on commercial computer simulation apps broadly used  Performance: 70 - 115 % faster than previous Intel Xeon 5400  Energy Industry: Eclipse simulations for days and months to optimize their tactics to extract oil  Performance: 150 % faster than previous Intel Xeon 5400.  Cisco and Red Hat benchmark  Performance: Top spot for 2-socket systems with new blade server from Cisco  Cisco Unified Computing System exceeded the previous Intel mark in SPECopmMbase by 154%  Oracle Applications and Red Hat benchmark  Top result for 2-Socket system with Oracle’s E-Business Suite 44

Slide 45

Slide 45

Red Hat Confidential - NDA required Best Performance … Floating Point SPEED Performance on SPECfp*_base2006 benchmark 45 RHEL 5.3 39.3 40 35 SLES 10 26.5 30 25 20 15 MS Windows Server 2003* Ent. SP1 11.7 SLES 10 18.3 10 5 0 SPECfp*_base2006 (Intel Xeon) (3.8 GHz, 2MB L2, 800 MHz FSB, single-core) 5160 (3.0GHz, 4MB L2, 1333MHz FSB,dual-core) x5482 (3.2 GHz, 2x6MB L2, 1600 MHz, quad-core) X5570 (2.93 GHz, 8 MB L3, 6.4 GT/s, quad-core) 45

Slide 46

Slide 46

Red Hat Confidential - NDA required Best Performance … Web Server Performance on SPECweb2005 benchmark RHEL 5.2 75023 80000 70000 60000 50000 RHEL 5.2 29591 40000 30000 RHEL 5.2 15193 20000 10000 RHEL 4U2 4555 0 SPECweb2005 (Intel Xeon) (3.8 GHz, 2MB L2, 800 MHz FSB, single-core) 5160 (3.0GHz, 4MB L2, 1333MHz FSB,dual-core) x5460 (3.16 GHz, 2x6MB L2, 1333 MHz, quad-core) X5570 (2.93 GHz, 8 MB L3, 6.4 GT/s, quad-core) 46

Slide 47

Slide 47

Red Hat Confidential - NDA required Best Performance … Multiphysics Finite Element analysis using ANSYS* RHEL 5.3 RHEL 5.3 RHEL 5.1 RHEL 5.1 47

Slide 48

Slide 48

Red Hat Confidential - NDA required Best Performance … Crash Simulation analysis using LS-DYNA* (single node) RHEL 5.3 RHEL 5.3 RHEL 5.3 RHEL 5.3 RHEL 5.3 RHEL 5.3 48