This month’s article will review TZO’s High Availability service. TZO provides Global Load Balancing as a service through DNS delegation, lowering site failover scenarios down to seconds rather than hours or in some cases days.
I remember a time when a person’s business phone and email system was something that was at the office, not in the car or corner cafe. Today in an ever expanding global workplace we expect to be able to communicate anytime, anywhere, from any device, and under any circumstance, suffice to say the rules have changed. Large companies need to consider how they can provide email and a slew of other web services to their employees and customers 24 x 7.
The problem is how does a company load balance or failover critical Enterprise web services across two or more different physical locations? Some companies deploy very complex, not to mention expensive, globally load balanced architectures using numerous appliances at multiple sites, expending lots of resources and time. Others try to leverage DNS by round-robin, using multiple DNS records to distribute customers across sites. The issue with a round-robin design is many websites, like OWA (Outlook Web Access)make multiple DNS queries during a session and customers could be randomly connecting to all of your sites during a singular session, throw in authentication issues and you will quickly finding yourself needing a more mature solution. Another approach some companies use is to advertise DNS records with very short TTLs (Time To Live), this allows the DNS record in a client’s cache to expire very quickly so that clients and recursive DNS servers, are always making authoritative requests for new records. This is a great plan up until the moment your primary DNS server goes up in a mushroom cloud and you are now reliant on working with the ISP to redirect authoritative requests to another primary DNS server.
So what’s the solution? In a word……Outsource.
(Insert Mad Loyal Reader) OK Tony now you’ve gone too far and used the “O” word. People lose jobs over that word, are you crazy? I can think of half a dozen appliances that do load balancing without having to give up authoritative control of our DNS zones. What makes TZO so special?
Alright everybody calm down I’m not suggesting outsource you jobs, just DNS. While I understand we have been snowballing IT management for years about the great mysteries and complexities of DNS we all need to come clean and explain that DNS, (insert admission gulp), is just a text file that get’s copied from server to server. Would anyone really care if that text file, which (come on admit it) has maybe 20 entries, was outsourced? I personally doubt it, in fact by outsourcing your DNS you have the potential to gain much more from increased flexibility to update records anywhere at any time, like in a disaster, and provide your customers with global redundancy.
If the word “outsource” is not allowed in your organization please consider the following: “Software as a Service”, or its popular acronym “SAAS”, or maybe the recently coined “Cloud Computing”. Trust me everyone I get it, I’m in the same boat, I’m going with “Cloud Computing” it has that mysterious quality that Senior Executives love and chicks dig.
I recently had a chance to talk to TZO Director of Sales & Marketing Christopher Cook about their Global Failover and Load Balance offering and asked if he could describe some of the key benefits to delegating DNS to TZO.
Chris explained that by delegating authority of your DNS zones to TZO you immediately become part of a global DNS infrastructure that extends through “North America, Western Europe, and soon the Pacific rim”. The benefit of this design is that company’s can decrease DNS propagation delays down to seconds and minutes. This is critical in a DR scenario where traditional DNS architectures can have propagation delays that are upwards of half a day or more.
OK, so to be fair I asked Chris why outsource, couldn’t companies install their own Global load balanced solution?
Chris’ answer was easy, TZO is about “25% of the cost of hardware solutions” and you don’t have to setup or configure anything and “if your business needs change your not [stuck] with unwanted hardware”. The other benefit according to Chris is that “when a company is ready all they have to do is change their [DNS delegation] to our servers”, it is that easy.
I told Chris many companies just won’t feel comfortable delegating their entire authoritative DNS zone to TZO, is it possible to only delegate a sub-zone?
Chris explained that if all a company wanted to do was load balance one website they could just delegate a sub-zone for that website. For example, WWW.company.com would just need to delegate the sub-zone “WWW”. The beauty of this design is that the company remains in full control of their “company.com” authoritative space except for the “WWW” subzone. When customers make an authoritative request for WWW.company.com TZO will answer. And if you are load balancing across two different physical sites customers could be redirected to any “A” record in that zone, such as SITE1.www.company.com and SITE2.www.company.com.
How does TZO monitor site availability?
TZO monitors site availability using a Multi-Point Monitoring architecture. They monitor each site from two different geographic locations within their architecture and both locations have to agree a site is down before automatically redirecting traffic to the alternate site, reducing false positives. Wow! Imagine the cost of trying to configure that kind of reliability on your own?
If a customer connects to my website at Site1 and during the session they make another DNS request are they redirected to Site2?
No, the answer is that TZO can be configured to make sessions persistent so that customers are not crossing the Internet to another physical site every time a DNS request is made during a session.
How flexible is the TZO HA service?
TZO allows customers to manage their DNS zones using a very user friendly web interface. Customers can customize the percentage of DNS requests that resolve a particular site. An Enterprise could answer 70% of WWW.company.com to SITE1 and 30% to SITE2. If you wanted to you could make one site 100% and the other 0%, might come in handy during scheduled maintenance. For companies that have secondary Co-Locations this service could be the cornerstone of their redundant architecture.
The final point to all of this is to remind Enterprises that even though DNS is easy, redundancy can be complex. Companies need to evaluate their core competencies and decide what fundamental services need to be highly available in an organization. To successfully load balance multiple physical sites, most organizations are going to need a little more than a DNS round-robin design or short TTLs. If your plan is to white wash management with the idea that your ISP is going to be a part of your failover design, good luck. Experinece has shown that if by some miracle you can get a person on the phone that can actually spell DNS I wouldn’t bank on propagation times under a day. So please consider your load balance and DNS propagation plans carefully, the last thing anyone wants is a short TTL on their “J-O-B” record.
-Tony
To find out more information about TZO please visit https://www.autofailover.com
1 response so far ↓
Rich Kucera // November 7, 2008 at 2:58 pm
Read it. Very interesting Tony. I should get back to my UI customization work… when will that go away? TOGAF has called for de-customization in its core principles, I say “go ahead, make my day”.
Oh yeah, also have to get back to my humble system access interface that has become Grand Central Station for every business process…