Thanks Tony. =)
A few things I can offer.
First off, a lot of people throw the argument of parry-haste and ZOMGtheyllpwnurfaceyounub around, but not a lot of people seem to actually understand how parry haste works, so let me explain it as best I can.
Bosses (you can practically ignore most other smaller mobs since they'll have a whopping 4-5% parry chance) have a set swing time. Many of them are squarely at roughly 2 seconds, there are some who swing substantially faster but these same bosses (Mother Sharazz and Patchwerk for example) are incapable of parrying for this very reason. So, when the boss parries your attack his swing timer is reduced to 20% of the maximum value regardless of where it is. It cannot be moved to less than that. So, what this means is, if he JUST attacked (100% of swing time left) and he parries you, he just ignored 80% of his swing. If he is down at the last 15% of his swing, the parry does not speed the attack up. If he is half way through his swing timer and he parries, he cuts 30% off his swing time. This is typically, statistically, averaged out to roughly 40%, which is the number that gets thrown about by the knowledgeable.
So here's some simple numbers to give an illustration:
2.0 sec swing timer, 4000 damage per hit (after all your mitigation)
=> zero haste and this will be 2000 damage per second
Let's say over a 60 seconds of a boss fight, pure auto-attacks, no expertise:
<<With nothing else happening the boss will hit 30 times for a total of 120,000 damage>>
Case #1: Warrior with a typical tanking sword (1.5 sec swing time)
60 seconds is 40 attacks by the warrior, if the boss parries at only 10%, he'll parry 4 of those attacks. If every parry is the worse case scenario, then the boss gets 4x0.8 more swings, or 3.2 more swings, round down to 3, and add that to the 30 swings normally, so we have 33 swings or 132,000 damage (a 10% increase). Remember, best case scenario is zero haste, if we use the 40% average, that's only 1.6 more swings, with conservative rounding up, that's 2 more swings or 128,000 damage (7% increase).
Case #2: Bear Tank, feral attack speed is 1.0 sec without modification
In 60 sec that is 60 attacks, 10% parry means 6 swings will be parried, worst case scenario is 6x0.8, or 4.8 rounding up to 5, that's 35 attacks for 140,000 damage (17% increase). 40% average would be 2.4 more, rounded off to 2 is 32 attacks and 128,000 damage (7%).
Case #3: Death Knight with a big 2-hander and a slow speed of 3.5
This will be roughly 17 attacks, 10% parry chance, we'll round up to 2 parries. Worst case scenario would be 1.6 additonal swings, or rounded off, 32 or 128,000 damage (7%). Average of 40% and we have rounded up, 1 more swing or 124,000 damage (3%).
Case #4: Death Knight dual wielding with two classical tanking weapons (1.5 sec)
This would be 80 attacks, 10% parry translates to 8 parries. Worst case scenario is 8x0.8=6.4 round down to 6 extra swings, for 144,000 damage (20% increase). 40% average is 3.2, round to 3 extra swings, for 132,000 damage total (10% increase).
Case #5: Death Knight dual wielding two slow tanking weapons (2.4 sec)
This would be 48 swings, 10% parry, round up to 5 parries. Worst case that's 4 extra attacks for 136,000 damage (13% increase). 40% average translates into 2 extra swings, or 128,000 damage (7%).
So, to sum up (using 40% haste averages, with better numbers):
Warrior (classical tank weapon) = ~5% increase
Bear Druid (feral attack speed) = ~8% increase
DK (slow 2-hander) = ~2.2% increase
DK (dual wield classical tank weapons) = ~10.7% increase
DK (dual wield slow tank weapons) = ~6.4% increase
DK (dual wield slow main hand, fast off hand) = ~ 8.5%
I added the last scenario because this is the setup I'd likely use, I'll get more into that below. This is based solely on auto-attacks, but auto-attacks are only half the picture. There is also the matter of special abilities.
For warriors, the standard rotation will be made primarily of Shield Slam, Devastate, and Revenge. Shield Slam and Devastate can be parried, Revenge can not. You will be using these abilities on the GCD, so every 1.5 seconds. In that 60 second window that is 40 GCD's. Let's do some practical consideration, if a warrior has a total 60% chance to dodge/parry/block to proc Revenge, they'll receive 18 chances to Revenge (and our warrior never misses that option). We'll also say he Tclaps twice to keep the debuff up, and demoralizing shouts twice as well for the same reason. So, 22 if the lucky warrior's CD's are unparriable. 18 of those attacks can be parried, 10% rate means an additional 2.4% damage increase, and realistically there won't be THAT many actual revenges, but we'll use the conservative number, add it to the warrior's auto-attack for a grand total of 7.4% increase.
For druids, they have swipe, lacerate, and mangle that will be the main use of cooldowns, all of which can be parried. We'll say 2 demoralizing roars for the debuff, and 2 faerie fires for the same. This means a whopping 36 moves can be parried for an additional 4.8% damage increase. Added to the auto-attack value and you're looking at a 12.8% increase.
For DK's it gets a bit more complicated. The moves you use and the number you can fire off aren't just restricted by the GCD, particularly while tanking, and there are more than a few moves that are not susceptible to parrying. One additional move that limits our vulnerability is Rune Strike. Each time we dodge or parry we get a Rune Strike that will replace our next attack with an added damage and threat move that cannot be parried. In other words, this means our avoidance actually is doing double duty at reducing our parry haste vulnerability, except unlike revenge where it makes our next special move safe, it makes our next auto-attack safe.
So, let's consider a big 2-hander with an Unholy tanking spec. Let's take a rigid rotation (ick) where we're fighting a boss (single target, no aoe needed). If we keep DnD up that would be DnD > IT > PS > BS (> UB/DC's slipped into cracks) in set 1 (10 sec), ScS > ScS > BS > BS on set 2, PS > IT > ScS > BS > BS on set 3, DnD > ScS > BS on set 4, PS > IT > ScS > BS > BS on set 5, ScS > ScS > BS > BS on set 6. DnD and IT are not parriable, so 20 attacks out of 25 attacks are parriable resulting in a 2.7% increase. Alternately if we go single target only moves, we'll say, IT > PS > ScS > BS > BS on odd sets, then ScS > ScS > BS > BS on evens. That's 24 of 27 parriable, or an 8% increase. Now, if we give the DK a 50% avoidance (entry level tank gear say, with Blade Barrier up, misses don't count), then 15 of the boss attacks will generate a Rune Strike opportunity, though sadly our slow swing speed means we can only use half of them, so 7 rune strikes instead of 7 of the 17 auto-attacks. 10 auto-attacks means a 1.3% increase in boss damage. Combined with our DnD rotation that puts us at a 4.0% damage increase, or single target strikes, an 9.3% increase.
Now let's consider a deep frost spec with my preference, a slow weapon in the main hand (to maximize the instant weapon damage strikes), and a fast weapon in the off-hand (to encourage procs, of say Killing Machine). Now, with the frost spec, the special rotation will look like IT > PS > HB > BS > BS on odd sets, and we'll do a FUFUFU combo of OB > HB > OB on the even sets, with lots of FS's thrown in the middle. Now, take note, IT, HB, and FS cannot be parried. That means that a total of 15 moves are parriable, resulting in a 2% damage increase. Now because we're dual wielding, with that same 50% avoidance, we get to use all of the Rune Strikes! 15 of our 64 auto-attacks are now no longer parriable. The auto-attack vulnerability is now down to 6.5%. Combined with our specials we're looking at a total of 8.5% increase.
So, to sum up this section, we're looking at:
Warrior = 7.4%
Bear = 12.8%
DK (2-hand Unholy) = 9.3%
DK (dual wield Frost) = 8.5%
(bearing in mind that the spec plays a major part for the DK but with the same set of gear the 2-hand unholy knight has a larger vulnerability than the frost dual wielder. The bear actually has an even higher chance than either.)
This is where we start to consider the nature of each sort of tank. Bears will have the most armor and the most health, and probably about 45% avoidance (entry level dodge+miss). Warriors will have less armor, less health, and only 35-40% avoidance at equivalent levels, but they have shield block to functionally add ~5% damage reduction to their armor, they also have Def stance to offer 10% damage ignore to physical damage. Death Knights will have a margin more armor and health than a warrior, they'll have about 55-60% avoidance (dodge+parry+miss with Blade Barrier) at the same level gear.
So Bears will suffer the largest damage increase due to parry haste at the base, but they can dodge and eat more than a warrior. The warrior will take more of the hits but take less damage from a given hit. And the Death Knight will avoid more hits than any other, though he may take the hit a small margin harder than the warrior, unless of course he has one of his CD's up.
Now the final major factor to consider on the survival front: expertise. 32.8 expertise rating relates to a 1% parry neglect. On our calculations, that is a 10% reduction in the % posted above (i.e. expertise will have a bigger effect on damage reduction the larger your % increase is). In reality it is slightly less, closer to 7-8% reduction in the vulnerability. Currently, Blood can offer 11 straight Expertise (independently worth 2.75% parry neglect), while Frost and Unholy can each offer 5 (1.25%). This makes also brings Blood into a more reasonable range to be dual wielding, though it will be more vulnerable than frost due to the reliance on more parriable attacks. Just for shits and giggles:
Blood, dual wield slow main fast off, 64 auto-attacks. IT > PS > OB/DS > HS > HS on set 1, HB > HB > HB > HB > OB/DS on set 2, IT > PS > HS > HS > HS > HS on set 3, OB/DS > OB/DS > HS > HS on set 4, the repeat 1 and 2. That's 27 parriable, or a 3.6% damage increase. We get our 15 rune strikes, leaving our auto-attack parry haste at 6.5%, for a grand total of 10.1%. Remember this spec gets more than double the talented expertise base of the frost spec, so can nearly even them out. Maybe not as much better as we might of hoped, but the expertise talents certainly have their value now don't they.
End result? For purposes of survival, you're looking at DK's in the middle of a +/- 2% damage increase on bosses due to parry haste. this is a gross simplification since the nuance of this damage increase will vary in implication for each type of tank. Namely warriors will get hit more, but for less, death knights will be hit less, but for more, and bears will get more parry hasted ouchies but they'll see the smaller % damage increase because of their larger health pool and heavier armor.
(/bow, math nerd at your service)
Now to shift gears, consider why we would want to dual wield as a Death Knight while tanking. For starters, with the current state of tanking gear, it's not just possible, it is reasonable to be hit capped in normal tanking gear once you get past the entry level and into the epics (though it may mean you're hunting more specific pieces depending on your luck with drops), that means you'll never miss with your specials regardless of your weapon choice. For auto-attacks (white damage), with equivalent item level you can get 2 one-handers with better combined dps than a 2-hander. This is regulated a bit with the increased miss chance. The upshot is that your white damage will be higher with 2-handers, but because of the nature of many of our DK moves (again depending on spec) we'll lose other dps for the decrease in instant weapon damage. This isn't quite as bad as you might think, let me give some real examples:
Broken Promise (2.5 speed, 392 avg damage, 1-hand)
Death's Bite (3.4 speed, 693 avg damage, 2-hand)
Both same item level. The 2-hander has a bit better than half again the base damage of the one hander, and most of the strikes based on weapon damage use only 30-60% of that damage in part of the strike damage. Incidentally for Frost, IT and HB two major damage contributors don't scale with weapon damage at all, similarly Unholy's DnD, disease damage, and UB are not at all affected by weapon damage. Arguably Blood would be most susceptible since Heart Strike is 60% and Oblit is 100%.
So, in DPS tests, which I have done a lot of, a smartly geared and spec'd dual wielder can pretty much match the dps of a 2-hand wielder with a different spec, but here the major variable is the player, since they need to select a smart spec for their weapon setup, pick smart talents, and gear to use that spec well. Certain specs gain a lot with dual wielding, such as Killing Machine with some buffed auto-attack crit. So, if we consider that an equal footing (allowing that person to person you'll have preferences and perform better with certain specs and styles), then the last thing to consider is the weapons themselves!
What can we get from dual wielding, and what can we get from 2-handers. Well, currently there are no 2-handers with defense/parry/dodge on them, there are however plenty with piles of strength and stamina. A quick perusal of equivalent 1-hand tanking weapons versus smart DK 2-hand tankers is that you can get roughly the same amount of stamina from either, but the pair of 1-handers will get you less strength and more direct tanking stats (def/parry/dodge), and 2-handers will get you more Strength and possibly other combat stats (crit/hit/exp, or icky haste/armor pen) but lack the defensive values.
Really, to me, it's all a wash. You can dual wield or 2-hand, it's all perfectly feasible. If you dual wield it is not a signed death warrant, spec will determine your vulnerability and it is possible (as shown above) to dual wield with one spec at lower risk than a 2-hander with a different spec.
As always, with our glorious DK class, what wins out at the end of the day is a smart player who selects their talents, their gear, and their move priorities smartly. Be adaptive, be creative, and you can achieve far more than the sheep who do one thing or another on the direction from someone else without really understanding it themselves.