You are currently browsing gdrawdy's articles.
Command Line administration…for those times with a good GUI just won’t work!
esxcfg-* Commands
· esxcfg-advcfg Sets advanced VMkernel options
· esxcfg-auth Authentication configuration
· esxcfg-boot† Configure boot, including PCI allocation
· esxcfg-dumppart Set, activate, deactivate, list potential and current VMkernel dump partitions
· esxcfg-firewall Configure firewallingoptions.
· esxcfg-info Prints information about the service console, the VMkernel, various subsystems in the virtual network and storage resource hardware
· esxcfg-init† Perform initialization steps in the initrd
· esxcfg-linuxnet Converts vswifdevices to eth when booting into Linux debugging mode
· esxcfg-upgrade† Upgrade from ESX Server 2.X to 3.0
†Should not be run unless instructed to do so by VMware Technical Support Representative
· esxcfg-mpath Multipathingconfiguration
· esxcfg-nas Add, delete or manage NAS file systems
· esxcfg-nics Presents physical NIC information
· esxcfg-rescan Rescan LUNs on SCSI device
· esxcfg-resgrp Create, delete and list resource groups
· esxcfg-route Enable or disable routing for vmknics
· esxcfg-swscsi Configures software iSCSIadapters
· esxcfg-vmhbadevs Maps COS device files to vmhbanames
· esxcfg-vmknic Create & configure VMkernel NICs
· esxcfg-vswif Create & configure vswifsfor the COS
· esxcfg-vswitch Create & configure virtual switches and port groups
NETWORKING
The biggest issue I see with networking is on the initial setup when the service console is not accessible from the network. The following will help correct this so that it can be accessed via the GUI where you can use the GUI to configure as ultimately needed.
When service console networking problems require manual virtual switch reconfiguration from the physical console use the esxcfg-vswitch command to reassign physical NICs to virtual switches
· Display current configurations:
o # esxcfg-vswitch-l
· Detach a physical NIC from a virtual switch:
o # esxcfg-vswitch–U <pnic> <vSwitch_name>
· Link a physical NIC to a virtual switch:
o • # esxcfg-vswitch–L <pnic> <vSwitch_name>
Modifying Service Console Networking
· # esxcfg-nics–l
Name PCI Driver Link Speed Duplex Description
vmnic0 02:02.00 tg3 Up 1000Mbps Full Broadcom …
vmnic1 02:02.01 tg3 Up 1000Mbps Full Broadcom …
vmnic2 0a:01.00 tg3 Up 1000Mbps Full Broadcom …
· # esxcfg-vswitch–l
Switch Name Num Ports Used Ports … Uplinks
vSwitch0 64 3 … vmnic0
PortGroupName Internal ID … Uplinks
Service Console portgroup0 vmnic0
· # esxcfg-vswitch–U vmnic0 vSwitch0
· # esxcfg-vswitch–L vmnic2 vSwitch0
· # esxcfg-vswitch–l
Switch Name Num Ports Used Ports … Uplinks
vSwitch1 64 3 … vmnic2
PortGroupName Internal ID … Uplinks
Service Console portgroup1 vmnic2
Service Console Firewall
The service console in ESX 3.0 is protected by an iptablesfirewall use esxcfg-firewall to administer
By default, only services required by VMware Infrastructure are enabled; all other services must be activated by the administrator via the VI Client or CLI
Service ports are opened either by known service name or by individual port number
Opening a port by service name:
· # esxcfg-firewall –e smbClient
·
Opening a port by service name:
· # esxcfg-firewall –o 123,udp,out,ntp
iptablesKnown Services
AAMClient*
CIMHttpServer*
CIMHttpsServer*
CIMSLP*
commvaultDynamic
commvaultStatic
ftpClient
ftpServer
LicenseClient*
nfsClient
nisClient
ntpClient
smbClient
snmpd
sshClient
sshServer*
swISCSIClient
telnetClient
TSM
veritasBackupExec
veritasNetBackup
vncServer
vpxHearbeats*
* Indicates the service is enabled by default
Networking Diagnostics
Diagnostics: VMKernel TCP/IP Stats
# cat /proc/vmware/net/tcpip/ifconfig
ping command uses service console TCP/IP Stack
vmkping uses VMKernel TCP/IP stack
# vmkping -D -v
Diagnostics: Collecting Network Traces
Run tcpdump/ethereal/netmoninside the guest or in the service console
Traffic visibility depends on the portgrouppolicy settings
Portgroup with VLAN id 0 (No VLAN)
• Sees all the traffic on the virtual switch without VLAN tags
Portgroup with VLAN id ‘X’ (1-4094)
• Sees all the traffic on the virtual switch with VLAN id ‘X’
Portgroup with VLAN id 4095
• Sees all traffic on the virtual switch
• Traffic is captured with VLAN tags
Promiscuous mode
• Accept: All visible traffic
• Reject: Only traffic matching the client MAC address
Storage
esxcfg-dumppart
esxcfg-mpath
esxcfg-nas
Service Console Administration
esxcfg-info
esxcfg-advcfg
esxcfg-resgrp
esxcfg-vswif
esxcfg-vswitch
esxcfg-nics
esxcfg-swiscsi
esxcfg-vmhbadevs
These pics were taken at a corporate datacenter that I work with. This is a hugh install all managed using IBM Director.
Start tftp up on a server and put the boot and os flash files in the tftp landing directory
The firmware for the Nortel Networks L2/3 GbESM is contained in two files: one is a boot image file and the other is the OS image file. Use the following steps to upgrade the firmware on the Nortel Networks L2/3 GbESM via the Telnet session:
1. Type /boot/gtimg.
2. Enter where the new image file will be placed. We are upgrading the boot image file, so enter boot. That is the location for the boot image file.
3. Enter the IP address of the TFTP server.
4. Enter the fully qualified path name for the boot image file that is on the TFTP server.
5. The switch reports the current version of the boot kernel on the switch and ask if you wish to replace it with new file. If you wish to continue, enter y.
6. When the download is finished, go back to Step 1, and repeat the process for the OS image file. In step 2, enter image1 or image2 as the location to store the new image file.
7. If the download location is the same as the location for the currently loaded OS image, the switch warns you that a failed download could result in an inoperative switch. If the download location is different from the location of the currently loaded OS image, the image file downloads. After the download is finished, the switch asks whether you want to use the old location or the new location. Figure5-6 on page37 shows a successful download of the OS image to image2.
8. Type /boot/reset to reset the switch and reboot with the new firmware files.
I needed a way to run TFTP on my MacBook Pro and this is the procedure I went though below. I used it to flash the boot and firmware on a IBM Blade Center’s Nortel Layer 2/3 Ethernet switch.
If you plan on making any changes in the default settings for the standard tftp service that is standard (but not necessarily running) you need to back up the default tftp.plist using this command:
cp /System/Library/LaunchDaemons/tftp.plist ~/Desktop/tftp.plist
The tftp.plist has only one argument: -i.
You should refer to the tftpd man pages for additional arguments. In 10.4, The directory /private/tftpboot exists by default in version 10.4 so you dont really have to do anything to start it up. The commands to start and stop the tftp daemon are:
# sudo launchctl load -w /System/Library/LaunchDaemons/tftp.plist
# sudo launchctl unload -w /System/Library/LaunchDaemons/tftp.plist
If it doesnt work verify no syntax errors in the tftp.plist file and that the correct ports are open on your firewall.
VMware ESX has a feature called HA which basically restarts downed Virtual Machines after a VMWare physical host has gone down due to a hardware failure.
While in the industry HA, at least in the UNIX world, means assurances to a certain absolute degree of operational continuity during a given measurement period…. regretfully in VMware if it is used…it means you have experienced down time!
But what if we could some how use VMware’s vmotion to move the running virtual machines to a different physical host long before the hardware actually failed?
What if you could find a piece of software that extended the use of VMWare….that was free….and where you could have direct manufacture support 24 hours a day 365 days a year?
What if we could spec that the software must have these features:
- A Single console to manage both physical and virtual systems
- Administer multiple virtualization technologies from a single console
- Easy to install and use
- Proactive virtual machine migration using VMotion based on predictive failure alerting from power supplies, CPUs, Memory, Disk Drives, Voltage Regualtion Modules, PCI slots, Fans, etc
- Drive VMware VMotion using physical hardware status information
- in the event of an actual failure, server must near instantaneously notify the manufacture of the failure so repair parts and assistance can be
- A built-in topology view to help with understanding of resource linkages
- Integration with an enterprise maintenance console to increase serviceability by migration of virtual machines to a standby server during a service window
By using the IBM Director Extension Virtualization Manager you can have all of this.
Imagine this scenario:
You have 5 VMware physical hosts each supporting 30 virtual servers. Your hardware is now 2.5 years old as you deployed your VM infrastructure beginning back in 06. You are running IBM Director and Virtual Center Server. IBM Director is monitoring the IBM BMC or RSA components of your 5 physical hosts. You also have deployed the free Service and Support Manager plugin of IBM Director. On one of the hosts the voltage regulation modules for CPU01 begins accumulating counts at an increasing rate as monitored by the hardware’s BMC. The hardware’s BMC determines at 5:00 am on Wednesday that a failure is eminent and flags the VRM with a predictive failure alert. The predictive failure alert causes 2 action plans to occur:
- The IBM Service and Support Manager plugin to IBM Director receives the PFA and sends a notification to IBM hardware support in Atlanta Ga. The engineering staff there receives the alert, validates that the server is under maintenance, calls your corporate help desk to validate the ship to address and to schedule a IBM CE on site if needed. Your help desk validates the PFA alert on the server and send you an email letting you know that a new IBM component will be on site after you finish your first coffee of the day.
- The IBM Virtualization Manager extension to IBM Director receives the PFS simultaneously and notifies VMware that the physical host is going into maintenance mode. Using predefined actions all running virtual machines residing on the physical host with the failing VRM are live migrated (vmotioned) off to one of the 4 remaining physical vmware hosts thereby assuring zero down time for your virtual server farm. After the failed component is replaced and the server is taken out of maintenance mode all domiciled virtual machines will be live migrated (vmotioned) back to the repaired server
The alternative to these automated actions steps include down physical servers, down virtual servers, calls from internal and external customers, missed coffee, middle of the night wake up calls, missed SLAs, frustration, anger, etc.
AND……….
These extensions are FREE!
PLAN PLAN PLAN and then PLAN some more.
- Make the Plan, Work the Plan, Ride the Wave – Glenn Drawdy
- A good plan today is better than a perfect plan tomorrow – General George S. Patton
- Always plan ahead. It wasn’t raining when Noah built the ark. – Richard C. Cushing
- Plans are nothing; planning is everything. – Dwight D. Eisenhower
- By failing to prepare, you are preparing to fail. – Benjamin Franklin
Plan Your Deployment Before You Install -
- What will your load mix be?
- What will be the density of your virtual machines?
- Will you use VSMP?
- Will you have heterogeneous or homogeneous work loads?
- WIll you need 64bit virtual machines?
- What type of licensing structure will you use?
- Is your active directory going to integrate into your virtual infrastructure?
- Is your AD structure optomized for VMware?
- Is your network strong enough and partitioned to allow for appropriate security to the virtual infrastructure?
- Is your shared storage designed in a way to allow for the correct IOPS when they are needed?
- Have you taken steps to prevent your the service consoles root file system from filling up?
————
View Gotchas:
View Client with Offline Desktop cannot be installed on the same machine as VMware Workstation, VMware ACE, VMware Player, or VMware Server
View Client, the View Client with Offline Desktop, and View Agent cannot be installed on the same machine
Synchronize time – Make sure that you have synchronized time across View Connection Server and desktop virtual machines. By default, virtual machines pick up the time from the ESX host at boot and you might have to synchronize time on the ESX hosts through the ntpd service
Make sure your virtual network is configured correctly for the desired maximum number of machines. The default value for ports in a vswitch in esx is 24
Upgrade View Client machines to use Microsoft Remote Desktop Connection (RDC) 6.x
View does not support using link-local (169.254.x.x) addresses for virtual desktops – Configure virtual desktops to use DHCP or static IP addresses
Install the VMware View Client with Offline Desktop if you are going to use Offline Desktop
VMware Sets Capacity Record Running Microsoft Exchange on IBM System x3850 M2 Servers. This was done in Feb of 2008. As usual after IBM and VMWare worked through and resolved the issues surrounding running 16,000 heavy mailbox users on a single 4 processor server….Dell took their research and configuration and did likewise. I guess this is what stumps me about technology….there are so many me too companies in the crowd but very few with the imagination to actually invent or engineer something original. Anyway here is the first paragraph of the article….the link below provides access to the entire article if you are interested.
———–
Microsoft Exchange Virtualized by VMware More than Doubles Native Capacity of Mailboxes Running on 16-core Physical Servers CANNES, France, February 26, 2008 — VMware, Inc. (NYSE: VMW), the global leader in virtualization solutions from the desktop to the datacenter, has set a record in system capacity and resource utilization for running Microsoft Exchange. VMware deployed Microsoft Exchange Server 2007 on VMware Infrastructure 3 (VI3) and successfully supported 16,000 heavy-user* Exchange mailboxes on a single 4-socket multi-core IBM System x3850 M2 server. Running Microsoft Exchange on VMware software increased by more than 100% the number of supportable Exchange users as compared to Microsoft Exchange’s prescribed recommendations for running natively in a non-virtualized environment.** VMware virtualization software enables enterprises to take full advantage of multi-core hardware servers to run the most demanding enterprise applications much more efficiently.
From time to time I talk with customer’s who use Dell servers. As a former Dell customer I am fascinated with those who are still Dell customers so I typically (if given a chance) ask questions to see if I can determine how and why they use Dell. I can summarize these Dell situations into a few basic groups using basic observations and absolutely without malice. These groups are:
- purchased a Dell laptop and liked it, so once in a position to influence the business IT direction insisted on Dell
- attended school (high school, college, university) where Dell was a key player, became comfortable on the Dell equipment and do not have time or desire to learn of equipment from other vendors
- believe Dell web site offers easier purchasing experience
- received a Dell laptop as part of a university program for free (I’m not calling this indoctrination but it sure sounds familiar)
- started working in a organization where Dell is already entrenched and do not have the time to research other technologies or solutions
- I like Dell and I will not or can not tell you why (sort of the old Chevy vs Ford or BMW vs Mercedes thinking process)
- We have Dell and don’t want to mix in any other vendors because we are afraid they will cause issues
- Dell has great advertising and consistently beats IBM and HP (I’ll have a post coming up on this in the future!)
For these groups I decided to do a comparison of the new Intel Nehalem processor based servers from Dell and IBM. Hopefully you will see why Dell is being replaced in most data centers.
This information was all obtained from the respective manufactures web sites (www.dell.com and www.ibm.com).
- On the Dell small and medium business section of their web site the Dell R710 starts out at $1,899 with a instant internet savings of $470 for starting price of $1,429
- On the IBM web site the IBM System X 3650 m2 starts out at $1,316
With a web page open to each manufactures site I configured these two systems as identical as possible to function as VMWare VI3 Servers running ESX Enterprise and using iSCSI for the VM Storage (many of the differences I point out below and are based on different decisions the manufactures have taken in either assembling parts or in their engineering).
- Coming in at $10,415 (after the instant internet rebate) the Dell R710 with two Intel® Xeon® Quad Core X5570, 2.93Ghz processors, 8M Cache, 32GB Memory (8×4GB), 1066MHz Dual Ranked RDIMMs (Note that I had to use the 4Gb RDIMMS because Dell’s design forces the smaller DIMMS to run at a much slower clock rate of 800Mhz), rack rails and cable management arms (I list that here because unlike with IBM these are extra), PERC 6/i SAS RAID Controller at RAID 5, 3 73GB 10K RPM SAS 2.5″ Hot Plug Hard Drive, Redundant 870W power supplies (note that the base 570watt power supplies had to be scrapped because of the faster processors and memory foot print), Bezel (IBM includes this in the base price), iDRAC6 Enterprise (IBM includes this in the base), Embedded NICS with TOE and iSCSI enabled (IBM includes this in the base), RW DVD ROM (IBM includes a RW optical drive in the base), 24/7/4 hour on site maintenance
- Coming in $863 less at $9,552 the IBM 3650 m2 with two Intel® Xeon® Quad Core X5570, 2.93Ghz processors, 8M Cache, 32GB Memory (8×4GB), 1333MHz Dual Ranked RDIMMs (Note that IBM uses faster 1333MHz memory than Dell), rack rails and cable management arms (I list that here because even though IBM includes these with the base Dell does not), ServeRAID-MR10i RAID Controller at RAID 5 (note that this RAID controller supports RAID 0, 1, 1E, 5, 6, 10, 50, and 60 – the Dell iDRAC6 does not support all of these RAID levels), 3 73GB 10K RPM SAS 2.5″ Hot Plug Hard Drive, Redundant 675W power supplies (note that this is 195 watts less than the Dell per Power Supply), Bezel (IBM includes this in the base price), embedded remote management, Embedded NICS with TOE and iSCSI enabled (IBM includes this in the base), Ultra-slim CD-RW DVD ROM Combo, 24/7/4 hour on site maintenance
The absolute strangest thing I found during this comparison was that of the power supplies. I know that there is a big push out there to go GREEN….but it seems that Dell is having a hard time at it so they are changing the rules and the unsuspecting customer will be paying the price! The base IBM 3650 M2 server comes standard all the power it needs unless the customer wishes to have redundant power. The server uses 1 675 watt power supply and the customer can a redundant power supply for $299. This power supply is designed to provide this chassis all the power it needs for years into the future and through all the speed and core bumps that Intel will be throwing at it. In Dells case, especially with the power hogging 3.5″ drives, they offer their base unit with a 570 watt power supply. A untrained technology buyer will do that simple math and deduce that the Dell is more power efficient. THAT IS NOT TRUE! Dell is offering these power supplies in hopes that you will do that simple math and not dig any further. The processor in Dell base unit is an Intel E5502. This is a 45nm DUAL CORE processor. In the Dell unit as you add more features into the server you are forced to upgrade to the 870 watt power supplies which use nearly 200 watts more per server than the IBM 3650 M2. In a comparably equipped system, like for like, apples to apples, the R710 will NOT be more energy efficient than the 3650 M2. But you mark my words….Dell will be moving to advertise that way because of the “teaser” power supply they put in this units standard model. Image if you will your server room storage closet with 100 extra power supplies stored because you had to upgrade your 50 servers to run more memory or more hard drives. IBM doesn’t force customers into that situation! I cant find exactly what the magic number of components is for the Dell server to require the upgrade but when I created a balanced server on their web site with 2 quad core processors and anything over 24 gigs of ram requires the upgrade of the power supply. Seems to me this is absolutely a bait and switch play for Dell and they are trying to hook customers with the lower rated power supplies only to have them find out later that they need to upgrade.
Both systems have the same Intel Xeon E5502 processor BUT the base Dell system is using unbuffered Dimms while IBM is using registered and buffered Dimms. This is a performance and availability issue here. Dell should be using the more expensive memory for their customer’s sake. The Dell offering also doesn’t include Rack Rails or cable management arms which are included with the IBM 3650 M2 standard. The Dell offers only raid 0 or 1 standard where the IBM 3650 M2 offers those plus RAID 1E. In Dell’s document “Cabling the Dell PowerEdge R710″ we read
- “The CMA (Cable Management Arm) on the Dell PowerEdge R710 is optional. Without the CMA installed, the system must be powered down and all cables disconnected before it can be removed from the rack.”
While they say removed from the rack…what they actually mean also includes the pulling of it forward so they can do maintenance on an internal failed component. Now tell me why would a manufacture offer a SERVER that has to be un-cabled before it can be pulled out for maintenance? Desktops I can understand…but a SERVER?
I did notice that in this NEW server model Dell seems to be offering old style 3.5″ disk drives. This is very bad news for their customers as the industry is absolutely moving away from these drives. Dell will buy up loads of them and dump them on their unsuspecting customers. I am sure that Dell is doing this because they are pushing internal storage and the 3.5″ drives go up to 1 Terabyte vs trying to assist their customers with a move to a SAN or NAS solution.
IBM normally includes the DVD drive in their price but in this specific model it wasn’t included on line. A CD-RW/DVD-ROM Combo can be added for $129. On the Dell web site it indicates their DVD ROM is included in price. It is not the RW model most people are wanting and has these words in small print next to it “May delay your PowerEdge R710 ship date”. Dell offers the ability to upgrade to a RW model for an additional $69.
————————————
If you are doing VMWare on these chassis IBM absolutely has the coolest support offering. IT is call VMWare RTS plus Base. What this does is give you one throat to choke! You call IBM for hardware support, VMware support, and virtual machine support whether the VM operating system is MS, RedHat, Susie, etc! You can dump your other support contracts and deal directly with IBMs support engineers in Atlanta and Raleigh….vs India with Dell…but wait…Dell doesn’t have anything like this offering! Its price is right on track to being only a couple of $100 more than the standard RTS support. Ask your Business Partner about it….or contact me and I can hook you up.
IBM Active Energy Manager available on the 3650 M2 enables power usage to be managed, monitored, and alerted via IBM Systems Director EVEN when the server is turned off. This feature allows customers to see power trends with power and heat over time. It also allows customers to place a cap on how much energy the IBM 3650 M2 power supplies are allowed to draw. If capped, the system will throttle performance to the new power window that has been selected. I see customers using this to cap power during none peak usage times periods, evenings, and or weekends to lower their total power bill through lowered power draw and lower HVAC requirements. Dell does not have anything remotely close to this capability….thus they do the head fake shenanigan of offering a incomplete power supply. The IBM System X 3650 M2 supports the ServeRAID-MR10is VAULT encryption adapter card to encrypt data on the hard drive. This feature allows customers to do hardware based encryption without regard to software or to whether the drive is lost or stolen. The data is secure regardless. Since the encryption is hardware based it does not cause performance issues like software based solutions do.
IBM has chosen to go with the next generation BIOS which as of now Dell and HP are scoffing at. I am sure when the feature hits their servers their marketing machines will go in reverse to let everyone know how awesome it is. But for now as they try to get it running they are slamming it. This new next generation BIOS is Unified Extensible Firmware Interface (UEFI) is a common management interface that reaches across all of IBMs new System X servers. It is a replacement to the old BIOS technology that is still available on both HP and Dell servers. UEFI provides enhanced functionality, a easier user interface, and much easier management. I will do a post on it in the future.
In terms of PCI slots, the 3650 M2 comes with 4 PCI Express x8 slots which can be converged to 2 x16 slots. The R710 comes with 2 x8 slots and 2 x4 slots.
In terms of PFA IBM monitors Power Supplies, Processors, Memory, Hard Drives, Voltage Regulator Modules, Power Supplies and Fans where Dell only has PFA on hard drives and memory. This is BAD especially in a VMware environment where a failed processor can down a server. You need this PFA and IBM Systems Director to tie into VMware to enhance the HA features of VMWare HA. The PFA feature provides additional functionality when the servers are unplugged in the case of a failed component by the usage of light path diagnostics. Light path diagnostics is a feature where small on board LEDs illuminate when a mother board based button is pressed. The LED illuminates on any failed component to allow fail proof identification of fault components. These LEDs function with or without power being applied to the servers \ blades. Dell does not have this feature.
Dell is pushing their customers to use UDIMMS vs RDIMMS in their memory slots. The less expensive UDIMMS that Dell is offering do not support the memory capacities which is again absolutely needed for server consolidations using products like VMware or Xen. These two sentences were taken from Dell’s technical white paper titled Dell PowerEdge Servers 2009 – Memory:
- UDIMMs should be purchased by customers who need a limited amount of memory
- RDIMMs should be purchased by customers who need large amounts of memory (up to 8 GB DIMMs), a broader future memory expansion roadmap (due to the ability to achieve three DIMMs per channel), and the latest RAS features (address parity)
If customers are virtualizing and concerned about power, the larger RDIMMS from IBM offer performance and power gains not available in the UDIMMS. IBMs memory, depending on configuration runs either at 1066 or 1333 Mhz while dells UDIMMS start at 800 Mhz. In the models where customers can upgrade to RDIMMS in the Dell they only run at 1066 Mhz thereby causing performance degradation in the same memory components in the IBM server. IBM also offers Memory Mirroring while Dell only offers Spare Bank Memory. Again it looks like Dell has gone down the wrong path at least for now on the UDIMMs being used in Servers. I find that they are mainly used for Desktops since the largest udimm available right now is a 2G DIMM where the RDIMMs go up to 16G.
The network controllers are the same on both servers with the 3650M2 offering dual gigabit ethernet ports with TOE (TCPIP over Ethernet) and the option of 2 additional ports. Dell charges $276 to activate the TOE and iSCSI features.
The IBM 3650 M2 has 12 2.5″ sas/sata/SSD (Solid State Disk – i.e. disks made from memory chips) drive bays as well as an option for a internal tape drive while Dell offers up to 6 3.5″ drives or 8 2.5″ drives. Dell has no option for a internal tape drive.
IBM offers Hot Swap fans. Dell does not.
Well hopefully you see some of the differences in these two models. If performance is your thing your probably not looking at Dell anyway but I will post performance numbers in the future.
Stay Informed!
This post is for all those who love Dell. I love them to….because they keep me in business on the side where I do odd jobs to repair, restore, or recover data lost on them!
Anyway if you are actually wanting to do a comparison of Dell vs IBM consider this to be the place to get your information. When looking at the Dell machines it is important to not get hypnotized by the low “teaser” prices that always reflect a stripped down model. YOU need to be knowledgeable enough to know what options you need so that the server you buy will do the job you need done. This isnt always a simple thing to do. You need to know CPU speeds and feeds, what memory configuration is the right one for performance, what raid level you need for the correct number of iops and for the protection of your data. You also need to know about building a balanced system so that you are not over loading the cpu, memory, disk, i/o, or even the fsb.
You see many of Dells teaser models do not come with the standard components that many of the IBM servers come with. For example: RAID controllers (RAID 0 and 1), a CD-RW/DVD drive, an advanced management processor, rack rails, as well as hot swap redundant power and fans. Of course this list doesn’t include the items that are not even offered in the line.
Dell stresses that their PowerEdge servers deliver high levels of reliability….and uptime….that meet the needs of “industry-standard server customers”. This is Dell secret code for anyone who is looking for the absolute cheapest server made with no thought given to performance or stability. The vast majority of companies today who use Dell servers do in fact have servers that have features such as hot-plug hard drives, RAID levels of some sort, ECC and spare-bank memory. When looking at the marketing material from Dell it is often easy to overlook the fact that Dell’s marketing machine uses words such as TCO, TCA, “next generation” ROI, high availability, resiliency, savings, etc to give the impression that they are the way to go. I can not stress enough that this is a big mistake. Please make sure that you know what you need before you go down the Dell path!
If you take a casual look at the IBM product line you will see that IBM doesn’t just go out and buy Intel or AMD parts and slap them in a package and hope that software vendors can provide a way of managing the finished product. IBM engineers in Raleigh have been steadily putting in their IBM Stealth Black boxes many mainframe inspired pieces of technology. Items such as Enhanced Predictive Failure Analysis (EPFA), Light-Path Diagnostics, memory resiliency, advanced remote-management capabilities, embedded RAID controllers, hardware based data encryption, and a free management solution that is hands-down one of the most amazing products I have used.
In terms of systems management IBM and Dell are like the old Levi Jeans tag where two horses were hitched to a pair of Levi Red Tag jeans and each pulling in opposite directions. Dell has established an framework where they expect third-party vendors such as Microsoft and Altiris to supply key pieces of the solution. In contrast IBM has the IBM Ssytems Director Management software that it supplies free with each IBM server. IBM Systems Director is a product that can run standalone or can be used to add increased value to third party management products if your company is already using them. If not IBM Systems Director is absolutely a must do project for your data center. The main problem with Dell’s approach to systems management is its general lack of potential and the inadequate amount of integration between modules. In Dell’s management solution when something does go wrong you have an extremely difficult time trying to identify the module and responsible party. In comparison IBM Systems Director and all of its components are totally owned, architected, and engineered by IBM. This ownership allows all of the components to be tightly integrated and tested before deploying. In Dell’s case a standard patch from one of their third party products could yield your management system totally down. IBM Systems Director out of the DVD\CD sleeve will allow you to preform miracles in your server environment that you never thought were possible before. When combined with VMWare Virtual Center you will be able to extend VMWares HA to the point to where you never experience any downtime. Additionally as I said above IBM Systems Director can be extended to work up or down link with with any of the following products if you already have them installed….but they are by no means a requirement:
Tivoli, Microsoft Ssytems Center Operations Manager, OpenView, BMC Performance Manager, and CA NSM
Bottom line is that Dell doesn’t offer anything unique in its systems management approach.
From time to time I will hear from customers commenting that Dell offers more internal drives than IBM comparable System X Servers. The reason for this is that IBM has made a general decision on many of their system x rack servers to limit the number of drives. IBM made this decision after extensive consultation with both industry performance experts, server optimization experts, and a wide range of customers. Dell and HP both know this and as a result it is one of the key areas they like to point out as one of their strengths when in fact what they are pointing out is a engineering decision that IBM is capitalizing on. IBM found that the vast majority of customers tend to go with external storage via sas, fibre, or iscsi and has developed storage options to that end. In a few cases such as the 3650m2 we do offer up to 12 server mounted disks. Although Dell doesn’t offer this many IBM doesn’t point this out as a competitive advantage because we do feel that massive internal storage is not the way for data centers to go. I believe that the internal storage should be for booting an operating system (unless you are booting from SAN which is a very popular option especially for Disaster Recovery purposes) or running a small SMB based application. What I see is that as customers move to virtualization, using internal storage will tie their hands in the future. Do not be side tracked by this tactic of Dells!
I grin when I hear Dell’s claim about their PowerEdge servers matching System X in terms of serviceability. I guess their goal is to match….but even in trying to match they fall way short! All IBM servers are equipped with Predictive Failure Analysis (PFA). PFA monitors all server hardware components and notifies the Systems Manager (BaseBoard Management Controller BMC or Remote Supervisor Adaptor RSA) when a component is nearing failure. The PFA monitors bad sectors, voltage glitches, errors on chips etc and will nearly eliminate all unplanned downtime by giving notice up to several days in advance of a failure. As soon as the PFA light is triggered, IBM will be notified via the Electronic Service Agent feature of IBM Systems Director (REMEMBER THIS IS FREE). Once IBM receives this notification THEY will call YOU to let you know of the failure. On that call they will work out the details of getting your system back to 100% health BEFORE it fails! Dell does have a neat LCD on the front of their servers but wouldn’t you rather have the manufacture call you and let you know of an impending failure….thats what I call serviceability! Note that IBM SYstems Director manages all of IBMs hardware (System X (intel \ amd), System P (Unix), System i (AS400), Storage, and System Z (Mainframe)). Check out this demo:
Dell has a big push going on now where they are touting they are the right choice for Small and Medium Business Customers (SMB). Dell is very active in offering cheap PCs, servers, and storage to this customer base. I make a lucrative amount of income on the side working with these customers who bought product from Dell but did not have the expertise to install or manage the products. I find that most SMBs lack internal staff to meld both the hardware and software into complete solutions. In working for my firm I like to learn all of what I can about a companies business and their short and long term goals for both their business and their IT investments so that I am putting together a total solution that is long lasting. Dell doesn’t have the staff for this and will typically come in at the last minute to offer a “low price bid”. If the customer goes with the low price bid my side business flourishes. If they go IBM I will typically see them 3-5 years after project implementation.
gdrawdy