Azure B Series VM: Burstable Sizing Guide

Introduction

Azure B-series VMs can cut compute costs for workloads that idle most of the day and need short bursts of CPU. The catch is credits. Pick a size without understanding them and you get throttling, slow builds, and a bill that doesn't look like a bargain anymore.

This guide is for Azure administrators, DevOps teams, and FinOps practitioners. It covers how B-series behaves, how CPU credits work, how to size one, where Standard_B4ms fits (it's now a legacy SKU), how B compares with D and other families, and how to monitor and evaluate total cost.

Key Takeaways

  • B-series suits low-to-moderate sustained CPU with intermittent bursts, not constant high demand.
  • Credits build below the baseline and drain above it; empty credits mean baseline-level performance.
  • Size on memory, disk, network, latency, and burst duration, not vCPU count or hourly price.
  • Validate with Azure Monitor data, load tests, and current Microsoft documentation.

What Are Azure B Series VMs?

What makes an Azure VM burstable?

B-series VMs are general-purpose sizes that run on CPU credits. Credits accumulate while the VM runs below its baseline and are spent when it runs above it.

Microsoft's B family documentation lists web servers, proofs of concept, small databases, dev builds, microservices, and dev/test as intended workloads.

The tradeoff is straightforward: you commit to a smaller sustained CPU level and keep burst headroom for short spikes. Always check current pricing for your region.

How to read B-series size names

Take Standard_B4s_v2. B is the family, 4 is the vCPU count, and the suffixes carry meaning:

  • s: Premium SSD compatible
  • a: AMD processor
  • p: Arm processor
  • l: low memory
  • m: memory intensive
  • v2: series version

Current B-family series are Bsv2 (Intel), Basv2 (AMD), and Bpsv2 (Arm). Region, OS, and licensing change availability and price, so confirm the exact SKU in the Azure portal or pricing calculator.

Which workloads fit, and which don't

Good candidates:

  • Dev and test environments
  • Low-traffic web servers
  • Small application servers
  • CI/build agents
  • Lightweight microservices
  • Bastion or utility hosts

Poor candidates:

  • Sustained CPU-heavy jobs such as continuous video or data processing
  • High-throughput application tiers
  • Latency-sensitive production services
  • Workloads with unpredictable or prolonged bursts

Those belong on D-series or a compute-optimized family. The dividing line is sustained versus intermittent demand, not production versus nonproduction.

How Azure B-Series CPU Credits and Burstable Performance Work

Understanding the baseline and credit bank

Each size has a documented baseline CPU level and a credit bank with a ceiling. Idle capacity fills the bank; bursts drain it. Using Microsoft's figures, a Standard_B4s_v2 has a 40% baseline, earns 96 credits per hour while idle, and banks up to 2,304.

Percentages need care. The older presentation scaled to vCPU x 100%, so legacy B4ms showed a 90% baseline and 400% maximum. Microsoft's 2024 update normalized to 0-100%, where B4ms is 22.5%. The underlying metric did not change, so always check which scale a source uses.

Earning and spending credits

Microsoft's CPU credit model gives this formula for credits banked per minute:

((Base CPU performance * vCPUs) - (Percentage CPU * vCPUs)) / 100

Microsoft's own example: a Standard_B2ts_v2 with a 20% baseline running at 10% CPU banks ((20*2) - (10*2))/100 = 0.2 credits per minute.

Below baseline, the balance rises. At baseline, it holds. Above, it falls. More vCPUs and a bigger gap between baseline and peak speed up both accumulation and depletion.

Azure B-series CPU credit accumulation and depletion cycle diagram

What happens when credits run out?

The VM throttles back to baseline until credits rebuild. Symptoms include:

  • Slower request handling
  • Longer builds
  • Growing queues
  • Missed performance targets

A cheap hourly rate doesn't make up for repeated throttling. If a B-series VM hits empty regularly, price a larger B size or a D-series VM against the cost of the degraded service.

Credit persistence and lifecycle events

Microsoft's CPU-credit and Bv1 documentation specifies initial credits at deployment but does not spell out what happens to balances after stop/deallocate, restart, redeploy, or host maintenance. Don't assume credits survive. Treat the balance as a runtime signal, and check it after maintenance, scaling, or recovery.

Azure B Series VM Sizing Guide

Start with workload evidence, not the VM catalog

Collect these before shortlisting anything:

  • Average and peak CPU
  • Burst frequency and duration
  • Memory working set
  • Disk IOPS and throughput
  • Network traffic and concurrency
  • Latency objectives and OS requirements

Review Azure Monitor history across busy and quiet periods. That's how you tell a short burst from sustained demand.

Use current B-series specifications as a reference

Representative four-vCPU sizes from Microsoft Learn (confirm in the portal, since availability varies by region):

SKU vCPUs RAM GiB Baseline Credits/hr idle Max bank
B4ls_v2 4 8 30% 72 1,728
B4s_v2 4 16 40% 96 2,304
B4pls_v2 4 8 30% 72 1,728
B4ps_v2 4 16 40% 96 2,304
B4ms (legacy) 4 16 22.5% 54 1,296

Azure B-series four-vCPU specifications comparison chart

The v2 four-vCPU sizes list no temp disk, up to 8 data disks, 6,400 uncached IOPS at 145 MBps, and burst to 20,000 IOPS. Disk bursting can last up to 30 minutes, which says nothing about CPU burst time.

Standard_B4ms as a practical checkpoint

Standard_B4ms is a Bv1 SKU that Microsoft marks End of Life. It offers 4 vCPUs, 16 GiB RAM, a 32 GiB temp disk, 2,880 uncached IOPS, and a 22.5% baseline. It is not another name for B4s_v2.

An East US pay-as-you-go sample from November 2026 listed $0.1660 per hour on Ubuntu and $0.1820 on Windows without Azure Hybrid Benefit. Legacy availability may be restricted, so treat these as reference points only.

Warning signs the size is wrong:

  • Too small: credits hit zero during normal peaks, or memory pressure appears.
  • Too large: credits sit at the ceiling for weeks.
  • Unsuitable: the workload needs sustained CPU or higher disk performance.

A repeatable selection process

  1. Establish resource requirements from telemetry.
  2. Shortlist two or three SKUs.
  3. Estimate credit accumulation and depletion.
  4. Compare against D-series.
  5. Run a representative test.
  6. Review production telemetry.

Six-step Azure B-series VM selection process flow

A smaller size spends credits often and costs less, but performance varies. A larger size has a higher baseline, steadier results, and a higher bill. Customer-facing services usually justify the steadier option.

Validate before production

Load test normal utilization, expected spikes, maintenance events, and worst-case burst length. Define the thresholds that trigger a resize, such as credits below a set level for a set period.

Document the resize and rollback steps. One low-utilization snapshot proves nothing.

B Series vs Other Azure VM Families

B-series versus D-series

Standard_D4s_v5 has the same 4 vCPUs and 16 GiB footprint, with no temp disk and 6,400 uncached IOPS. It delivers consistent CPU with no credit model. Choose D when:

  • Utilization is sustained
  • Latency is customer-facing
  • You need predictable memory-to-vCPU behavior
  • Credit depletion is an operational risk

How the major families differ

Family Positioning
B General purpose, credit-based CPU
D General purpose for enterprise apps, dev/test, web servers
E Memory optimized: databases, caches, in-memory analytics
F Compute optimized: batch, analytics
N GPU: AI/ML, rendering, simulation

"Types" can mean family, generation, or specialty capability, so compare actual SKUs rather than the family letter.

Virtual desktop scenarios

Azure Virtual Desktop is Microsoft's cloud service; VDI is the broader model. Microsoft's B-family list includes virtual desktops, and its session-host guidance gives D4s_v5 examples for medium workloads (4 vCPUs, 16 GiB).

Those figures are starting points that need validation. B-series can fit light or intermittent desktops, but not sustained interactive use.

Pricing, Monitoring, and Rightsizing

Evaluate total cost, not only hourly price

Compare B and D candidates under the same service levels, including:

  • Compute for your OS, region, and purchase model
  • Managed disks and snapshots
  • Windows licensing (Azure Hybrid Benefit can help)
  • Egress, monitoring, and backups
  • Reservations or savings plans
  • The cost of throttling

Azure VM total cost evaluation factors comparison diagram

Deallocated VMs stop compute billing, but disks and some networking keep charging.

Monitor performance and credit health

Azure Monitor exposes CPU Credits Remaining and CPU Credits Consumed on B-series VMs. Watch them alongside:

  • Percentage CPU
  • Memory pressure
  • Disk latency and throughput
  • Network activity
  • Application latency

Azure B-series VM monitoring metrics dashboard overview

Set alerts for sustained credit depletion, repeated throttling, and rising latency, with tighter thresholds in production.

Azure Advisor also flags burstable candidates when average CPU sits below the candidate baseline and P95 stays under twice that baseline, using a default 7-day lookback.

Keep VM sizing and storage optimization connected

A B-series VM fixes compute economics, not storage waste. Unattached disks keep charging after a VM is deleted, and provisioned disks are often far larger than needed. A FinOps Foundation paper on Azure managed disks describes free space and low IOPS utilization relative to provisioned capacity.

Lucidity complements VM rightsizing. Its platform gives real-time visibility across Azure volumes and autonomously optimizes block storage without infrastructure changes.

Lumen scores each disk's tier against IOPS, throughput, latency, and cost. It flags unattached, reserved, unmounted, and zero-I/O disks, which can represent up to 70% of unused block storage spend.

Review disks separately from CPU credits, then reassess the full workload cost and performance profile.

Frequently Asked Questions

What are Azure B series VMs?

Burstable, economical VM sizes with a lower CPU baseline for workloads with occasional spikes. CPU credits earned while idle fund the bursts.

What are the different types of Azure VMs?

Families map to workloads: B burstable, D general purpose, E memory, F compute, N GPU. Compare the exact SKU and generation.

What is Standard_B4ms?

A legacy Bv1 SKU (4 vCPUs, 16 GiB, 22.5% baseline) marked End of Life. Verify details and pricing in Microsoft documentation and the pricing calculator.

Are AVD and VDI the same thing?

No. AVD is Microsoft's Azure-hosted desktop service, while VDI is the broader approach. Either can use different VM families.

How do Azure B-series CPU credits work?

Credits accumulate below baseline and are consumed above it. When they run out, the VM throttles to baseline.

When should I choose B-series instead of D-series?

Choose B for intermittent, bursty CPU demand. Choose D for sustained, predictable performance or when credit depletion is unacceptable.