Suppose you're a data-hungry enterprise IT pro, lying awake at night, haunted by the thought of cloud storage latency stealing precious milliseconds from your mission-critical workloads. Microsoft has heard your silent weeping—and graciously responded by turbocharging its Azure Managed Disks with a new “Performance Plus” tier, swooping in to hand you both peace of mind and, ideally, a much fatter bandwidth pipe for your virtual machines.
Last week, Microsoft unveiled general availability for Performance Plus—a free upgrade (because who doesn’t love free stuff?) to its Premium (v1) Azure Managed Disk tier, targeting disks larger than 1TB. The headline perk? Disk performance is now doubled for the beefy “P40” 2TB disks, compared to the old “P30” 1TB recommendation, thanks to the enhanced bandwidth and a meatier helping of input/output operations per second (IOPS).
Sweet deal, right? More throughput, no sudden sticker shock… unless you count the fine print, which, in Microsoft’s true love-hate fashion, definitely exists.
Cue dramatic movie trailer voice: “In a world where nothing comes without caveats…”
First, to drink from the Performance Plus fountain, you need to create new disks—no upgrade-in-place magic available here. Sure, you can use a snapshot as a basis, but if you’re running a hard-hitting database workload, expect at least brief downtime or the need to stage everything via extra virtual machines. For those with compliance officers breathing down their neck, that might be as fun as a root canal.
And, notably, once you switch Performance Plus on, there’s no switching off—you're in for the whole ride. Also, Azure Site Recovery (ASR) won’t hold your hand if you’re living your best Plus life. Ditto for Managed Disk V2 and Ultra Disk. It’s starting to look suspiciously like there’s a “you must be this slow to ride” sign somewhere in ASR’s codebase.
This deep dive leans on synthetic IO testing tools: Microsoft’s own DiskSpd (for Windows fans) or FIO (for Linux loyalists). Simulated OLTP and data warehouse workloads can be checked with HammerDB for the truly ambitious, but for this head-to-head, straight-up DiskSpd was the weapon of choice.
Glenn Berry’s famous DiskSpd scripts provided the starting line, dialed up for clouds with big 200GB working files. (Yes, we know, that’s overkill, but let’s not talk about the time when test results all fit into RAM and everyone thought the SAN was fast.) The mix? 25% write, 75% read, mirroring even heavy OLTP patterns. Realism: not just for movies with giant robots.
Here’s where things get fun. In the cloud, your choice of disks is only half of the story—the other half is the size (and capability) of the VM you attach them to. Virtual machines have published caps on storage IOPS and bandwidth. For this test, the storied EDBSv5-16 VM was chosen, with plenty of headroom: 70,000 IOPS and 1,800MB/s for premium storage, even higher for Premium V2 and Ultra.
Why so much surplus? Because throttling, dear reader, is always waiting in the wings. The moment you hit your VM or disk limits, Azure will gently apply the brakes to your workload. Or, if you’re unlucky, stomp on them with both feet and crank up your latency. That leads to our golden rule: when aiming to understand raw disk performance, always stay below every possible quota and threshold.
With both Premium Plus and Premium V2 disks set up for 16,000 IOPS and 600MB/s bandwidth, repeated DiskSpd runs revealed a tale as old as cloud: consistency is king.
Premium V2—Microsoft's newer disk class—still came out on top for delivering the most predictable, low-latency I/O under heavy, steady workloads. The “everyday” latency was impressively flat: sub-5ms for the vast majority (with the 50th to 99th percentile between 4.5ms and 5ms).
But, climb up to the once-in-a-blue-moon “nines” territory, and strange things start to happen: outliers spiked into the tens, then hundreds of milliseconds. Whether that’s Azure gently limiting noisy neighbors, a bad node, or just a sleep mode prank, no cloud storage is truly immune. Still, for the vast majority of real-world cases, V2 remains the gold standard for steady latency.
Performance Plus, for its part, closely followed—but with just enough variance to matter for those running ultra-strict transaction processing or latency-sensitive analytics. For most cloud mortals, however, this upgrade is a clear improvement: double the performance at no extra cost, with only the classic migration hurdles to worry about.
Here’s a little nugget for the cynics: Ultra Disk—Azure’s show pony with its dedicated hardware platform—is still the champ if you’re chasing maximum throughput at any cost. But since that cost is roughly four to five times higher than mere mortals care to pay, most workloads will cheerfully live with Premium Plus or V2.
Even with the sexiest, most finely-tuned disks, throttling lurks below the surface—applied by Azure to keep everyone playing nicely in the same virtual sandbox. Drift too close to your VM’s limits, or your disk’s, or just make too much noise for too long, and wham: latency soars, throughput tumbles, and those database SLAs start sweating.
It doesn’t stop there. The same hidden resource governors that protect other tenants can also become accidental saboteurs, especially under sudden, varying loads.
For IT pros, the implication is clear: when architecting storage in Azure (or any public cloud), don’t just look at the headline IOPS/bandwidth. Map your disk performance to your VM’s maximums—and leave enough headroom for Azure’s throttle fairy to do her work without ruining your weekend.
If you’re lucky, this is a few clicks and just a bit of downtime—or a cleverly orchestrated shell game with extra VMs so users don’t notice. If you’re unlucky (or your data volumes are vast), be ready for service interruptions and some squirming during cutover.
And let's have a moment of silence for those who depend on Azure Site Recovery for disaster recovery peace-of-mind. Performance Plus, like its V2 and Ultra Disk friends, isn’t eligible. It’s a limitation with a whiff of “cloud era’s unfinished business”—because apparently, high performance and bulletproof recovery are star-crossed lovers in the Azure storage universe.
But, as ever, the devil is in the details:
The caveats are real, but so are the improvements. Microsoft is courting the high-I/O crowd with more flexibility than ever. But, if you want every last resilience bell and whistle (like ASR support), the advice remains: check under every menu option, ask around, and expect a few growing pains.
There's also the inescapable “cloud gravity” principle: moving large, latency-sensitive data in and out of Azure’s ecosystem remains a non-trivial affair. Optimizing inside the cloud gets easier every year, but you’ll want to keep tight controls on what triggers a wholesale migration or workload expansion.
Is this the last we'll see in Azure storage innovation? Absolutely not. The cycle will repeat: more tiers, more features, more acronyms. Each wave aimed at making the next swath of applications “cloud native”—until, one day, your most finicky database doesn’t even remember the difference between a SAN and a hyperscale VM.
For the skeptical IT pro, this is classic cloud: more performance, more complexity, and just enough caveats to make sure nobody gets too comfortable. But hey, at least now we can all rest a little easier, knowing that if the storage gods ever smile, they’ll do it with double the bandwidth… and just a hint of throttling.
Let’s just hope the next upgrade comes with a toggle switch—and maybe, just maybe, ASR support. If not, see you all in the change management war rooms, with our best-laid migration plans and, inevitably, a few grumpy end-users.
And so the Azure disk spins on—with a little more pep in its step, and a lot more performance in its tank.
Source: Redmondmag.com Microsoft Boosts Azure Disk Efficiency With 'Performance Plus' Tier -- Redmondmag.com
What’s New With Performance Plus?
Last week, Microsoft unveiled general availability for Performance Plus—a free upgrade (because who doesn’t love free stuff?) to its Premium (v1) Azure Managed Disk tier, targeting disks larger than 1TB. The headline perk? Disk performance is now doubled for the beefy “P40” 2TB disks, compared to the old “P30” 1TB recommendation, thanks to the enhanced bandwidth and a meatier helping of input/output operations per second (IOPS).Sweet deal, right? More throughput, no sudden sticker shock… unless you count the fine print, which, in Microsoft’s true love-hate fashion, definitely exists.
Cue dramatic movie trailer voice: “In a world where nothing comes without caveats…”
First, to drink from the Performance Plus fountain, you need to create new disks—no upgrade-in-place magic available here. Sure, you can use a snapshot as a basis, but if you’re running a hard-hitting database workload, expect at least brief downtime or the need to stage everything via extra virtual machines. For those with compliance officers breathing down their neck, that might be as fun as a root canal.
And, notably, once you switch Performance Plus on, there’s no switching off—you're in for the whole ride. Also, Azure Site Recovery (ASR) won’t hold your hand if you’re living your best Plus life. Ditto for Managed Disk V2 and Ultra Disk. It’s starting to look suspiciously like there’s a “you must be this slow to ride” sign somewhere in ASR’s codebase.
Under the Hood: Methodology, Benchmarks, and That Familiar Old Throttle
Whenever someone announces “faster cloud storage,” IT pros collectively raise one eyebrow. Claims are great, but performance benchmarking is where the rubber meets the virtual road.This deep dive leans on synthetic IO testing tools: Microsoft’s own DiskSpd (for Windows fans) or FIO (for Linux loyalists). Simulated OLTP and data warehouse workloads can be checked with HammerDB for the truly ambitious, but for this head-to-head, straight-up DiskSpd was the weapon of choice.
Glenn Berry’s famous DiskSpd scripts provided the starting line, dialed up for clouds with big 200GB working files. (Yes, we know, that’s overkill, but let’s not talk about the time when test results all fit into RAM and everyone thought the SAN was fast.) The mix? 25% write, 75% read, mirroring even heavy OLTP patterns. Realism: not just for movies with giant robots.
Here’s where things get fun. In the cloud, your choice of disks is only half of the story—the other half is the size (and capability) of the VM you attach them to. Virtual machines have published caps on storage IOPS and bandwidth. For this test, the storied EDBSv5-16 VM was chosen, with plenty of headroom: 70,000 IOPS and 1,800MB/s for premium storage, even higher for Premium V2 and Ultra.
Why so much surplus? Because throttling, dear reader, is always waiting in the wings. The moment you hit your VM or disk limits, Azure will gently apply the brakes to your workload. Or, if you’re unlucky, stomp on them with both feet and crank up your latency. That leads to our golden rule: when aiming to understand raw disk performance, always stay below every possible quota and threshold.
Reality Check: Consistency vs. Peak Throughput
So, what happened when theory met the relentless monotony of automated benchmarking?With both Premium Plus and Premium V2 disks set up for 16,000 IOPS and 600MB/s bandwidth, repeated DiskSpd runs revealed a tale as old as cloud: consistency is king.
Premium V2—Microsoft's newer disk class—still came out on top for delivering the most predictable, low-latency I/O under heavy, steady workloads. The “everyday” latency was impressively flat: sub-5ms for the vast majority (with the 50th to 99th percentile between 4.5ms and 5ms).
But, climb up to the once-in-a-blue-moon “nines” territory, and strange things start to happen: outliers spiked into the tens, then hundreds of milliseconds. Whether that’s Azure gently limiting noisy neighbors, a bad node, or just a sleep mode prank, no cloud storage is truly immune. Still, for the vast majority of real-world cases, V2 remains the gold standard for steady latency.
Performance Plus, for its part, closely followed—but with just enough variance to matter for those running ultra-strict transaction processing or latency-sensitive analytics. For most cloud mortals, however, this upgrade is a clear improvement: double the performance at no extra cost, with only the classic migration hurdles to worry about.
Here’s a little nugget for the cynics: Ultra Disk—Azure’s show pony with its dedicated hardware platform—is still the champ if you’re chasing maximum throughput at any cost. But since that cost is roughly four to five times higher than mere mortals care to pay, most workloads will cheerfully live with Premium Plus or V2.
The Cloud’s Dirty Secret: Throttling, Spikes, and the Law of (Data) Gravity
No discussion about cloud storage would be complete without revisiting our favorite bogeyman: throttling.Even with the sexiest, most finely-tuned disks, throttling lurks below the surface—applied by Azure to keep everyone playing nicely in the same virtual sandbox. Drift too close to your VM’s limits, or your disk’s, or just make too much noise for too long, and wham: latency soars, throughput tumbles, and those database SLAs start sweating.
It doesn’t stop there. The same hidden resource governors that protect other tenants can also become accidental saboteurs, especially under sudden, varying loads.
For IT pros, the implication is clear: when architecting storage in Azure (or any public cloud), don’t just look at the headline IOPS/bandwidth. Map your disk performance to your VM’s maximums—and leave enough headroom for Azure’s throttle fairy to do her work without ruining your weekend.
Migration Musings: Why Moving to Performance Plus Feels a Lot Like Dieting
Picture this: You want Performance Plus, but you’re already up to your elbows in production data. You can’t add the feature to existing disks. So, what’s next? You create fresh disks, possibly using a snapshot as your starting point.If you’re lucky, this is a few clicks and just a bit of downtime—or a cleverly orchestrated shell game with extra VMs so users don’t notice. If you’re unlucky (or your data volumes are vast), be ready for service interruptions and some squirming during cutover.
And let's have a moment of silence for those who depend on Azure Site Recovery for disaster recovery peace-of-mind. Performance Plus, like its V2 and Ultra Disk friends, isn’t eligible. It’s a limitation with a whiff of “cloud era’s unfinished business”—because apparently, high performance and bulletproof recovery are star-crossed lovers in the Azure storage universe.
Real-World Implications for IT Pros
In practical terms, Performance Plus expands Azure’s storage playbook in a positive direction—especially for organizations that can’t (or won’t) pay for Ultra Disks, but need more oomph than Premium V1 provided. Databases, analytics, even struggling ERP systems will all benefit from the extra headroom.But, as ever, the devil is in the details:
- No in-place upgrade: Every migration is an opportunity for things to go sideways. Plan the move, test religiously, and buy your favorite sysadmin a coffee.
- No Azure Site Recovery: If site-to-site disaster recovery is non-negotiable, either architect around this, push your friendly Microsoft TAM for features, or stay on older disk types.
- One-way street: Once a disk is Performance Plus, it stays Performance Plus—like a tattoo, but with less street cred.
- Beware the throttle: As soon as you start pushing IOPS or bandwidth anywhere near your allocated limits, Azure’s “dynamic gentle throttling” kicks in, turning your cloud Ferrari into a family sedan. Monitor everything; trust nothing.
- VM matters, too: All the disk performance in the world won’t help if your VM can’t keep up. Always size disks and VMs together as a matched set.
- Cost/Benefit: For most, Performance Plus is “just right”—the Goldilocks of cloud storage for IOPS junkies who want value. But if you’re running something so latency-sensitive that every microsecond counts, pony up for Premium V2 or Ultra. Or maybe, just maybe, keep a beloved on-premises SAN on hand for critical workloads.
The Bigger Picture: Where Microsoft’s Storage Strategy Is Headed
Azure’s steady expansion of disk tiers—V1, V2, Ultra, now Performance Plus—reflects the cloud arms race to sweeten the enterprise deal. On-premises performance still tempts the hardcore, but the public clouds know that every doubled IOPS or streamlined migration removes one more excuse not to jump ship.The caveats are real, but so are the improvements. Microsoft is courting the high-I/O crowd with more flexibility than ever. But, if you want every last resilience bell and whistle (like ASR support), the advice remains: check under every menu option, ask around, and expect a few growing pains.
There's also the inescapable “cloud gravity” principle: moving large, latency-sensitive data in and out of Azure’s ecosystem remains a non-trivial affair. Optimizing inside the cloud gets easier every year, but you’ll want to keep tight controls on what triggers a wholesale migration or workload expansion.
Witty Wisdom: The Circle of Storage Life (Now With Plus)
At the end of the day, Azure’s new Performance Plus is a solid, if slightly imperfect, step forward—one that will quietly help those whose workloads were just starting to chafe against the old disk’s guardrails.Is this the last we'll see in Azure storage innovation? Absolutely not. The cycle will repeat: more tiers, more features, more acronyms. Each wave aimed at making the next swath of applications “cloud native”—until, one day, your most finicky database doesn’t even remember the difference between a SAN and a hyperscale VM.
For the skeptical IT pro, this is classic cloud: more performance, more complexity, and just enough caveats to make sure nobody gets too comfortable. But hey, at least now we can all rest a little easier, knowing that if the storage gods ever smile, they’ll do it with double the bandwidth… and just a hint of throttling.
Let’s just hope the next upgrade comes with a toggle switch—and maybe, just maybe, ASR support. If not, see you all in the change management war rooms, with our best-laid migration plans and, inevitably, a few grumpy end-users.
And so the Azure disk spins on—with a little more pep in its step, and a lot more performance in its tank.
Source: Redmondmag.com Microsoft Boosts Azure Disk Efficiency With 'Performance Plus' Tier -- Redmondmag.com
