Archives

photo credit: anitakhart

I will always think of one word in particular when I think of this recent memorial day weekend…….Spare.

I hosted 25 people for a BBQ on Sunday, and things were going great until the gas ran out as I was cooking the appetizers. Thankfully I always have a spare gas tank, and 5 minutes later we were back in action.

Then. on Monday afternoon I received some SMS and email alerts from our monitoring system that a few of our app servers had stopped responding to queries. Since it was a few app VM’s which failed I went straight to our HP based SAN, which would not respond to either SSH or web access, even ILO showed it as being up. I began to get very nervous thinking about the risk of data corruption on our multi terabyte array…..

Share/Bookmark

This is part 2 of a multi part series on “What to Look for in a Hosting Environment.” The first installment was on monitoring and the post can be found here.

As we all know backups should be an integral part of any companies IT disaster and recovery planning, and just because your applications are running at a 3rd party hosting facility does not mean that it’s “not your problem.” If it’s important to your business, its important to ensure that it fits your companies backup/recovery/archival strategy.

Most hosting companies provide backup services, and many of these can be tailored to your needs. Here’s how to ensure the hosting company can meet your backup needs, and the questions to ask to ensure that your needs can be met before the contract is signed.

When a hosting company says that they will provide backup services, ask them the following:

1. What does the standard backup provide? OS configuration as well as data (for full server recovery), or data only?
2. Are backups stored on site, or are they also stored off site (for full disaster recovery)?
3. Is there a dedicated backup network (so the backups cannot impact server/service network performance)?
4. How are your backups performed? Just file level backups, or are there application backups such as MySQL dumps or MSSQL backups?
5. How often are backups performed?
6. Can there be different snapshot frequencies for certain data types (e.g. database every 30 minutes?)
7. How many GB are included in the standard backup package? What is the cost of overage?

If you’re using dedicated servers then regular backups using enterprise software such as Commvault or Veritas should be fine. A full server backup should be performed weekly, together with daily incremental backups (certain critical directories may need to be backed up on an hourly basis, or whatever schedule you’re comfortable with). VMWare (or other virtualization) Virtual Servers have the additional benefit of backup via the integrated backup utility on the Virtual Host Server.

If you have large storage requirements then local backups by the hosting company could be expensive. One option for consideration (depending upon the security needs of the files to be backed up) is to use a cloud storage service (such as Amazon’s S3) for these files, At only 15 cents per GB (plus data transfer fees) it’s a very cost effective option for important information. We were able to realize large savings in this area for one of our clients by storing large image files (totaling around 15 TB per year) on Amazon’s S3 service as the hosting company was charging over 50 cents per GB for a saving of over $5000 per month on backups alone.

As always we recommend redundant hardware (servers/storage devices) for all applications to minimize downtime in case of hardware failure, as no matter how good the backup strategy is, recovery of a large system will take time. And time is money. Your money.

Share/Bookmark

Having built and managed data center environments, provided hosting and managed 3rd party hosted solutions, I have seen a very broad spectrum of hosting options, and the benefits and pitfalls of each. Sometimes there are deals that appear to be too good to be true, and many times they are.  Once you drill down into the details you can usually find where the corners have been cut to market to a certain budget.

These cuts and sacrifices are not necessarily a bad thing, but you need to ensure that you understand their impacts, and that if you are fine with them, you can save some $’s, but not necessarily avoid the headaches. You will find that you generally get what you pay for, and that if you make fully informed choices you will get the exact environment you need, for a reasonable budget.

Not many people have the time to drill down into the nitty gritty of each hosted solution, so I thought I would discuss what to look for, why it’s important, and more importantly (to save your valuable time), what to ask in order tho get that information.

We’re going to look at five main areas to consider when evaluating a hosting environment:

1. Monitoring
2. Backup strategies
3. Hosting environment hardware and options
4. Security
5. Other topics of interest not covered by the above mentioned.

We’re not going to tackle cloud computing or specialized hosting, each of which deserve their own discussions.

Lets start with monitoring.  I hate getting late night calls for support. I’d rather either foresee a problem coming at me and avoid it in the first place, or at least know exactly what the cause of the issue is if I am called. A good monitoring system will provide all this and much, much more.

Share/Bookmark

In our infrastructure practice we’ve seen an uptick in the number of people implementing Oracle RAC on Linux. It’s a good solution and takes some of the headache out of capacity planning, but there are certain hardware pitfalls of which you need to be aware. First among these is the choice of the Intel Itanium platform.

One issue with the Itanium is that the driver support for RAID cards and SAN’s is well behind those of other platforms. Consider RHEL4 with Oracle. As of this writing, you cannot find an officially certified RAID card for the Itanium. You’re stuck with software RAID or unsupported hardware.

Or consider Hitachi multipathing software for HP OEM’d SANs. First off, Hitachi hasn’t built their HDLM software to run on a 64 bit Itanium (the AMD Opterons and Intel Xeon 64bit x86 architectures are supported, of course), and you can’t run Oracle on 32 bit RHAS as that
isn’t certified (don’t even mention the performance hit). So your choices for a decent SAN solution have just been slashed.

On paper, Itanium has great potential, but the much cheaper and better (full current X86 64 bit support!) AMD Opteron and Intel Xeon 64 bit solutions have come into the market and crushed it. Performance benchmarks show that an Opteron solution can run circles around Itanium at a fraction (30 – 40%) of the cost.
As a result, Itanium has never really gained traction, has very low market share, and is falling further behind. Software and hardware support for it lags way behind support for the AMD Opteron and Intel Xeon systems.

I predict that this trend of worsening support will continue, and the future upgradeability and supportability of the system will go from poor to highly suspect.

This scratches just the surface of the issues we’ve seen with Itanium and they stretch far beyond just implementations of Oracle RAC. The pain is real. If you’re considering Itanium, think again.

Share/Bookmark

Trying to keep your web site content up to date can be incredibly challenging when you have a large site with constantly changing data. However, not keeping it up to date can lose you business. That was the case with the behemoth known as HP.

I recently performed a compatibility analysis on the software and hardware for a large client project. If you’ve done this sort of thing, you know that the conflicting information provided on many vendors web sites can be a source of frustration. When you’re as large as HP, the gap between good and bad information can be especially large.

I was trying to determine if there was a RAID card on the HP site that is compatible with the HP Integrity 8620 series Itanium servers (running RedHat Advanced Server 4.0). The answer was yes. Or it was no. Or it was maybe.

l found 3 different documents; one saying that there was no RAID card with Linux support, one saying that one card was supported, but only the RHAS 2.1, and one showing all of the linux drivers available, which apparently excluded all but Redhat Enterprise linux 4.0.

So I decided to try our friendly neighborhood HP rep. He said that they do support RHAS 4.0, but has not yet provided confirmation. Most likely he’s searching the same mess of contradictory information on the HP site.

In the end l had to tell the client that the server they selected could not be certified in time due to the conflicting information on its ability to support hardware RAID. As a result HP will miss out on  a million-plus for the 3 servers the client needs.

Lessons learned for hardware and software vendors?

• Plan your site design to make it easier to keep content synchronized and updated. Make sure that only product managers and their teams can publish specs and docs on their products.

Lessons learned for system integrators?

• When analyzing compatibility, don’t trust the 1st document you see. Like a journalist, try to find support for it in other documents.

• Look at each item’s driver/software support to see whether it lags behind the competition. This should help you determine the useful life of the device. The more recent the driver support, the higher the likelihood it will maintain that support into the future.

Share/Bookmark

Many new network infrastructure management appliances are coming on the market, including several convergent devices which integrate many components together in one killer app(liance).

An example of this convergence is the new “F5 Big IP series” load balancing appliance that touts not only load balancing of server and application requests, but also firewall packet filtering and layer 7 application security, traditionally the domain of stand alone firewalls such as Checkpoint FW1.

While l see this convergence as a good thing for ease of management (one appliance to manage vs many) and lower rack space and power consumption, it comes at a much higher cost and misses some of the critical components the single focus applications/appliances provide. Either way it makes tech buying a more complex process trying to understand where each new device may fit in your organization.

When purchasing additional network functionality, in addition to standalone appliances tech buyers should evaluate convergent devices which may ultimately be a better fit for them. With the profusion of new appliances out there, understanding where the new offerings fit in your network requires a more knowledgeable and sophisticated buyer and VAR.

l will be writing an article soon discussing specific convergent devices, highlighting which offerings are suitable for which organizations.

Share/Bookmark